PyPI · #805 most downloaded on PyPI

great-expectations

1.23.219M downloads/mofivetran/great_expectations ↗New releases in your daily digest, changelog included.

Always know what to expect from your data.

Last release 9 days ago
25 Sep 2026
Ships on a steady schedule
a new release about every 2 weeks
Nearly every release is documented
notes for 60 of the last 60 stable releases
11 versions withdrawn
withdrawn after publishing
9 years old
369 releases · first in 2017

Release timeline

369 releases since 2017

20172019202120232025

Releases

1.23.225 Sept 2026Deprecated2 sources agree
The changelog entries for releases 1.17.1 through 1.23.0 are rewritten in the structured format, and the deprecation timeline gains rows for the gx-re…
Compatibility: sqlalchemy now <2.1 (extras snowflake, databricks)

Highlights

Fixes GX on SQLAlchemy 2.1 — SQLAlchemy 2.1.0, released 2026-09-24, broke GX 1.23.1 and earlier on Python 3.11+, where every SQL extra resolves it by default. Depending on the backend, import great_expectations failed whenever snowflake-sqlalchemy was installed, every Databricks query failed, driverless postgresql:// URLs could not load a driver, BigQuery queries comparing against a float failed, SQL Server reported mixed-case and upper-case tables as missing, and expect_column_values_to_be_of_type(type_="Numeric") failed on float columns. 1.23.2 fixes all of these: the snowflake and databricks extras stay below SQLAlchemy 2.1 until their dialects support it, and every other SQL extra runs on 2.1. Python 3.10 is unaffected, since SQLAlchemy 2.1 requires Python 3.11. If you can't upgrade yet, pin sqlalchemy<2.1; do the same if you install snowflake-sqlalchemy or databricks-sqlalchemy outside GX's extras. (#12269)

pip install --upgrade 'great_expectations[snowflake]' # include your extras so the SQLAlchemy cap applies

Regex Expectations work on ClickHouse — The four regex Expectations now run on ClickHouse, which does not support regexp_like(). ClickHouse is covered by integration tests for these Expectations. (#12222)

gx.expectations.ExpectColumnValuesToMatchRegex(column="name", regex="^A")

Correct substring matching for regex Expectations on Snowflake — Snowflake's REGEXP operator anchors patterns to the whole value, so unanchored patterns behaved differently there than on other backends. Regex Expectations on Snowflake now use substring semantics that match every other GX backend, while preserving the user's pattern. (#12221)

gx.expectations.ExpectColumnValuesToMatchRegex(column="name", regex="ell")

Nested struct columns supported in Spark value-counts Expectations — On the Spark engine, Expectations that rely on the column.value_counts metric — including ExpectColumnMostCommonValueToBeInSet and ExpectColumnKLDivergenceToBeLessThan — now work for dotted nested struct paths such as address.city, instead of returning an empty result with an unresolved-column error. (#12231)

gx.expectations.ExpectColumnMostCommonValueToBeInSet(column="address.city", value_set=["Springfield"])

Changes

Bug fixes

GX works on SQLAlchemy 2.1, which broke 1.23.1 on several backends: the snowflake and databricks extras are capped below 2.1, a broken snowflake-sqlalchemy install no longer prevents import great_expectations, driverless postgresql:// URLs fall back to psycopg2 when psycopg is unavailable, BigQuery renders Double as FLOAT64, SQL Server reflects mixed-case tables, the Numeric type name matches float columns again, and database URL masking keeps the database and query string verbatim. (#12269)

Validation results containing an infinite Decimal value — for example the maximum, mean or sum of a PostgreSQL numeric column or a pandas column of Decimal values — now serialize as float infinity instead of raising decimal.InvalidOperation. (#12254)

Regex Expectations on Snowflake now match substrings, consistent with other backends, rather than requiring the pattern to match the entire column value. (#12221)

UnexpectedRowsExpectation no longer misreads a query as containing a JOIN when the letters appear inside a string literal, a column name such as join_date, a quoted identifier or a comment; such queries are aliased correctly again and no longer fail with a syntax error on MySQL and SQL Server. (#12249)

Building validators on multiple threads no longer serializes or deadlocks on Python 3.10 and 3.11: concurrent validator construction now proceeds independently per instance while still building each validator only once. (#12232)

Regex Expectations now work on ClickHouse, which lacks regexp_like(); other SQL dialects are unchanged. (#12222)

Type Expectations on ClickHouse now compare and report the underlying SQL type for nullable columns instead of the Nullable(T) wrapper, so ExpectColumnValuesToBeOfType and ExpectColumnValuesToBeInTypeList behave as expected. (#12219)

The Spark column.value_counts metric now resolves nested struct columns such as address.city, so Expectations built on it return results instead of an unresolved-column error. (#12231)

Docs

The changelog entries for releases 1.0.0 through 1.17.0 are rewritten in the structured format, and the previously missing 1.13.1 release now has its own entry. (#12227)

The changelog entries for releases 1.17.1 through 1.23.0 are rewritten in the structured format, and the deprecation timeline gains rows for the gx-redshift extra alias and for CloudDataContext / cloud mode of get_context. (#12225)

Maintenance

Redshift CI jobs are capped to three concurrent runs against the shared cluster, and test teardown now drops every test schema with retries so failed runs no longer leak schemas. No user-visible change. (#12257)

Contributors

Thanks to @adimalkar, @Rayan-and-beyond (first contribution), @nanjeshramesh, @feiiiiii5, @alibro005, @pentaoa (first contribution).

1.23.118 Sept 2026Deprecated

The changelog now carries a deprecation timeline table listing every deprecated item, the version that deprecated it, and the version that removes it.

Highlights

Spark now evaluates each regex independently with match_on="all" — On Spark, ExpectColumnValuesToMatchRegexList with match_on="all" now checks every regex separately against each column value, so patterns anchored at different positions (such as ^A and [0-9]{3}$) both match a value that satisfies them. This matches the behavior already seen on Pandas and SQL. (#12198)
```
gxe.ExpectColumnValuesToMatchRegexList(
    column="id",
    regex_list=["^A", "[0-9]{3}$"],
    match_on="all",
)
```
Each Validator reports results for its own Batch when a datasource is reused — Validators built on the same datasource no longer borrow one another's Batch. Running two validation definitions on threads, or creating two validators from one datasource on a single thread, now evaluates and reports each validator's own data, with the correct batch_id, batch_spec and batch_definition on the result. This fixes a long-standing latent bug made reachable by #12148 in the 1.23.0 release. (#12211)
```
validator_a = context.get_validator(batch_request=request_a)
validator_b = context.get_validator(batch_request=request_b)
# validator_a still validates request_a's batch
result = validator_a.expect_table_row_count_to_equal(value=3)
```
A Spark schema saved in great_expectations.yml reloads correctly — A persisted spark_schema is now read back through StructType.fromJson, so reopening a File Data Context round-trips the schema instead of failing inside PySpark. Values that are not an accepted schema form now raise a validation error naming the field and the accepted types. (#12200)
```
context = gx.get_context(mode="file")
asset = context.data_sources.get("spark_ds").get_asset("my_asset")
assert asset.spark_schema is not None
```
GX config files are read and written as UTF-8 regardless of locale — config_variables.yml and great_expectations.yml, and the .gitignore read while scaffolding a project, are now opened with an explicit UTF-8 encoding. Projects containing non-ASCII values or comments can be created and reloaded on hosts with a non-UTF-8 locale, and a project YAML file that is not valid UTF-8 now raises an error naming the file instead of a bare decode error. (#12182, #12204)
A gallery-wide test tier for data sources — A new gallery support tier asserts a measured test result across the entire shipped expectation library — one case per registered expectation, each pairing a passing and a failing configuration — and nine data sources (pandas in-memory and filesystem CSV, SQLite, MySQL, PostgreSQL, Trino, BigQuery, Databricks and Redshift) now declare it after being measured against the full gallery. (#12150)

Changes

Features

Adds a gallery support tier that asserts a passing test result over the whole shipped expectation library, with nine data sources declaring it, a required CI lane per member, membership and coverage guards, and a measurement mode for evaluating new candidate backends. (#12150)

Bug fixes

ExpectColumnValuesToMatchRegexList with match_on="all" on Spark now evaluates each regex independently, so patterns anchored at different positions no longer fail on values that satisfy them all. (#12198)
A Validator now keeps the identity of the Batches it loaded even when its execution engine is shared, so concurrent validations and multiple validators on one datasource each evaluate and report their own Batch; a configuration naming a batch the engine does not hold now raises instead of silently validating the most recently loaded batch. (#12211)
The remaining config file reads and writes in the serializable data context are pinned to UTF-8, so scaffolding a project with a non-ASCII .gitignore and reading or updating great_expectations.yml work under a non-UTF-8 locale; an unreadable project YAML now raises an error naming the file. (#12204)
A spark_schema persisted in great_expectations.yml is loaded back through StructType.fromJson so the schema round-trips when the context is reopened, and an unsupported value raises a validation error naming the field and the accepted types. (#12200)
The Spark test-connection test, which covers behavior when PySpark is unavailable, is now skipped when PySpark is installed so it no longer fails for contributors with PySpark in their environment. (#12199)
config_variables.yml is now read and written as UTF-8, so saving and reloading a configuration variable containing non-ASCII characters works on hosts with a non-UTF-8 locale. (#12182)

Docs

The changelog now carries a deprecation timeline table listing every deprecated item, the version that deprecated it, and the version that removes it. (#12224)
Documentation fixes: the credential-configuration pages now point at gx/uncommitted/config_variables.yml, a mistagged code fence highlights again, a misspelled snippet name is corrected, and doubled words and spelling errors across the core docs, ADRs, gallery docs and contrib READMEs are fixed. (#12177)
Oracle is now documented on the connection-string reference (oracle+oracledb://...?service_name=...), the compatibility reference with its tested database version, and the data source method reference via add_sql. (#12167)

Maintenance

The pull request title check now requires exactly one current tag at the start of the title, and the contributor docs and template name only the four current tags. (#12223)
The published package metadata now includes project URLs linking to the source repository, documentation and homepage, so PyPI and dependency-tracking services can associate the package with its repository. (#12207)
The metric repository retriever tests are now type-checked, with concrete annotations replacing Any and the module removed from the mypy exclude list. (#12184)
The expectations test suite is now type-checked, including two guards that could never fail being corrected to actually test whether an optional dependency imported, and the directory-level mypy exclusion removed. (#12185)
Five more test modules are removed from the type-check exclusion list. (#12178)

Contributors

Thanks to @lakshayxi (first contribution), @feiiiiii5 (first contribution), @ptimizeroracle (first contribution), @alibro005 (first contribution), @yigitcan-ozturk, @toyeshhm (first contribution), @nanjeshramesh, @p-mandale (first contribution).

Open source →

Additional notes

Known issue: on Python 3.11+, this release resolves SQLAlchemy 2.1 (released 2026-09-24), which it does not support. Upgrade to 1.23.2, or pin sqlalchemy<2.1.

Highlights

Spark now evaluates each regex independently with match_on="all" — On Spark, ExpectColumnValuesToMatchRegexList with match_on="all" now checks every regex separately against each column value, so patterns anchored at different positions (such as ^A and [0-9]{3}$) both match a value that satisfies them. This matches the behavior already seen on Pandas and SQL. (#12198)
```
gxe.ExpectColumnValuesToMatchRegexList(
    column="id",
    regex_list=["^A", "[0-9]{3}$"],
    match_on="all",
)
```
Each Validator reports results for its own Batch when a datasource is reused — Validators built on the same datasource no longer borrow one another's Batch. Running two validation definitions on threads, or creating two validators from one datasource on a single thread, now evaluates and reports each validator's own data, with the correct batch_id, batch_spec and batch_definition on the result. This fixes a long-standing latent bug made reachable by #12148 in the 1.23.0 release. (#12211)
```
validator_a = context.get_validator(batch_request=request_a)
validator_b = context.get_validator(batch_request=request_b)
# validator_a still validates request_a's batch
result = validator_a.expect_table_row_count_to_equal(value=3)
```
A Spark schema saved in great_expectations.yml reloads correctly — A persisted spark_schema is now read back through StructType.fromJson, so reopening a File Data Context round-trips the schema instead of failing inside PySpark. Values that are not an accepted schema form now raise a validation error naming the field and the accepted types. (#12200)
```
context = gx.get_context(mode="file")
asset = context.data_sources.get("spark_ds").get_asset("my_asset")
assert asset.spark_schema is not None
```
GX config files are read and written as UTF-8 regardless of locale — config_variables.yml and great_expectations.yml, and the .gitignore read while scaffolding a project, are now opened with an explicit UTF-8 encoding. Projects containing non-ASCII values or comments can be created and reloaded on hosts with a non-UTF-8 locale, and a project YAML file that is not valid UTF-8 now raises an error naming the file instead of a bare decode error. (#12182, #12204)
A gallery-wide test tier for data sources — A new gallery support tier asserts a measured test result across the entire shipped expectation library — one case per registered expectation, each pairing a passing and a failing configuration — and nine data sources (pandas in-memory and filesystem CSV, SQLite, MySQL, PostgreSQL, Trino, BigQuery, Databricks and Redshift) now declare it after being measured against the full gallery. (#12150)

Changes

Features

Adds a gallery support tier that asserts a passing test result over the whole shipped expectation library, with nine data sources declaring it, a required CI lane per member, membership and coverage guards, and a measurement mode for evaluating new candidate backends. (#12150)

Bug fixes

ExpectColumnValuesToMatchRegexList with match_on="all" on Spark now evaluates each regex independently, so patterns anchored at different positions no longer fail on values that satisfy them all. (#12198)
A Validator now keeps the identity of the Batches it loaded even when its execution engine is shared, so concurrent validations and multiple validators on one datasource each evaluate and report their own Batch; a configuration naming a batch the engine does not hold now raises instead of silently validating the most recently loaded batch. (#12211)
The remaining config file reads and writes in the serializable data context are pinned to UTF-8, so scaffolding a project with a non-ASCII .gitignore and reading or updating great_expectations.yml work under a non-UTF-8 locale; an unreadable project YAML now raises an error naming the file. (#12204)
A spark_schema persisted in great_expectations.yml is loaded back through StructType.fromJson so the schema round-trips when the context is reopened, and an unsupported value raises a validation error naming the field and the accepted types. (#12200)
The Spark test-connection test, which covers behavior when PySpark is unavailable, is now skipped when PySpark is installed so it no longer fails for contributors with PySpark in their environment. (#12199)
config_variables.yml is now read and written as UTF-8, so saving and reloading a configuration variable containing non-ASCII characters works on hosts with a non-UTF-8 locale. (#12182)

Docs

The changelog now carries a deprecation timeline table listing every deprecated item, the version that deprecated it, and the version that removes it. (#12224)
Documentation fixes: the credential-configuration pages now point at gx/uncommitted/config_variables.yml, a mistagged code fence highlights again, a misspelled snippet name is corrected, and doubled words and spelling errors across the core docs, ADRs, gallery docs and contrib READMEs are fixed. (#12177)
Oracle is now documented on the connection-string reference (oracle+oracledb://...?service_name=...), the compatibility reference with its tested database version, and the data source method reference via add_sql. (#12167)

<details> <summary>Maintenance</summary>

The pull request title check now requires exactly one current tag at the start of the title, and the contributor docs and template name only the four current tags. (#12223)
The published package metadata now includes project URLs linking to the source repository, documentation and homepage, so PyPI and dependency-tracking services can associate the package with its repository. (#12207)
The metric repository retriever tests are now type-checked, with concrete annotations replacing Any and the module removed from the mypy exclude list. (#12184)
The expectations test suite is now type-checked, including two guards that could never fail being corrected to actually test whether an optional dependency imported, and the directory-level mypy exclusion removed. (#12185)
Five more test modules are removed from the type-check exclusion list. (#12178)

</details>

Contributors

1.23.010 Sept 2026

[BUGFIX] Cache the SQL execution engine across calls instead of rebuilding it every time

[BUGFIX] Cache the SQL execution engine across calls instead of rebuilding it every time (#12148)
[BUGFIX] query asset SQL ending in a comment fails validation (#12124) (thanks @nanjeshramesh)
[BUGFIX] Wrap a query asset's SQL whole so Oracle does not add FROM DUAL (#12162)
[BUGFIX] TupleFilesystemStoreBackend reads with ambient locale encoding (#12125) (thanks @nanjeshramesh)
[BUGFIX] Make ExpectColumnValueZScoresToBeLessThan pass rather than fail or raise on zero or undefined variance (#12145) (thanks @Star-cloud626, Claude Opus 5 (1M context))
[BUGFIX] expect_column_values_to_be_unique raises KeyError on mixed-case column names (#12180)
[BUGFIX] Report an undefined SQLite standard deviation as None instead of raising (#12168) (thanks @siddharthgaur1, Claude Opus 5)
[DOCS] Update repository links to the fivetran org (#12126) (thanks @iamfeldman)
[MAINTENANCE] Declare metric values, render-content payloads, and metric keys as what they actually are (#12116)
[MAINTENANCE] Generalize the data source declaration record beyond SQL backends (#12110)
[MAINTENANCE] Pin the fluent datasource management API with a per-type CRUD contract suite (#12141)
[MAINTENANCE] Say what the shared fixture column types mean (#12147)
[MAINTENANCE] Type-check tests/test_utils.py, tests/actions/ and tests/checkpoint/ (#12146) (thanks @MannXo)
[MAINTENANCE] Clean up check-actor-permissions and dead CI workflow code (#12159)
[MAINTENANCE] Bump browserslist from 4.28.1 to 4.28.8 in /docs/docusaurus (#12155)
[MAINTENANCE] Bump fast-uri from 3.1.5 to 3.1.7 in /docs/docusaurus (#12151)
[MAINTENANCE] Type-check tests/render (#12152) (thanks @Ryota-Di)
[MAINTENANCE] Type-check tests/core/ #12130 (#12140) (thanks @AnandkumarMall)
[MAINTENANCE] Type-check tests/execution_engine/partition_and_sample/ (#12169) (thanks @siddharthgaur1, Claude Opus 5)
[MAINTENANCE] Bump colord from 2.9.3 to 2.10.0 in /docs/docusaurus (#12170)
[MAINTENANCE] Type-check tests/data_context/ (#12171) (thanks @nanjeshramesh)
[MAINTENANCE] Remove unused checkpoint test fixtures (#12174)
[MAINTENANCE] Bump svgo from 3.3.4 to 3.3.5 in /docs/docusaurus (#12175)
[MAINTENANCE] Bump joi from 17.13.4 to 17.13.7 in /docs/docusaurus (#12176)
[MAINTENANCE] Publish the oracle extra (#12091)
[CONTRIB] Type-check validator tests (#12149) (thanks @yigitcan-ozturk)
[CONTRIB] Remove the dead match_on value key from the not-match-like-pattern-list metric (#12157) (thanks @Star-cloud626, Claude Opus 5)
[CONTRIB] [MAINTENANCE] Type-check the excluded modules under tests/integration/ (#12173) (thanks @adimalkar)

Open source →

Additional notes

Always know what to expect from your data.

pip install great-expectations Copy PIP instructions

Always know what to expect from your data. (See https://github.com/great-expectations/great_expectations for full description).

Project links

Homepage
Download

Key dates

PyPI data Data sourced directly from PyPI's database.

Released: Sep 10, 2026 Latest release

2 maintainers

PyPI data Data sourced directly from PyPI's database.

danklynn josh.stauffer

Credits

Author: The Great Expectations Team

License

Apache Software License (Apache-2.0)

Requires

Python <3.14, >=3.10

Provides Extra

athena teradata singlestore mysql vertica dremio spark oracle redshift cloud gcs snowflake postgresql sql-server bigquery arrow hive spark-connect databricks azure excel gx-redshift pagerduty trino clickhouse test aws-secrets azure-secrets fabric gcp s3

Classifiers

Development Status
5 - Production/Stable
Intended Audience
Developers
Other Audience
Science/Research
License
OSI Approved :: Apache Software License
Programming Language
Python :: 3
Python :: 3.10
Python :: 3.11
Python :: 3.12
Python :: 3.13
Topic
Scientific/Engineering
Software Development
Software Development :: Testing

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages .

Source Distribution

great_expectations-1.23.0.tar.gz (36.9 MB view details ) Uploaded Sep 10, 2026 Source

Built Distribution

Filter files by name, interpreter, ABI, and platform.

If you're not sure about the file name format, learn more about wheel file names .

Copy a direct link to the current filters Copy

File name

Interpreter Interpreter py3

ABI ABI none

Platform Platform any

great_expectations-1.23.0-py3-none-any.whl (5.7 MB view details ) Uploaded Sep 10, 2026 Python 3

File details

Details for the file great_expectations-1.23.0.tar.gz .

File metadata

Download URL: great_expectations-1.23.0.tar.gz
Upload date: Sep 10, 2026
Size: 36.9 MB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for great_expectations-1.23.0.tar.gz Algorithm Hash digest

SHA256 009335ebf49394cfbba646f1f0fe7b46ed6cf74a335723c6a529a1222657886e Copy

MD5 b5c947e2bbcf6a326dc73c88d1ad4b70 Copy

BLAKE2b-256 53195c6d136adc9c3895801a6a5e03db415dc02137fb7c0b88fd179bfdc801f2 Copy

See more details on using hashes here.

Provenance

The following attestation bundles were made for great_expectations-1.23.0.tar.gz :

Publisher: ci.yml on fivetran/great_expectations Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
Statement type: https://in-toto.io/Statement/v1
Predicate type: https://docs.pypi.org/attestations/publish/v1
Subject name: great_expectations-1.23.0.tar.gz
Subject digest: 009335ebf49394cfbba646f1f0fe7b46ed6cf74a335723c6a529a1222657886e
Sigstore transparency entry: 2784397633
Sigstore integration time: Sep 10, 2026 Source repository:
Permalink: fivetran/great_expectations@8210c2e280ebcd7f097e9d8850d81255de3026f6
Branch / Tag: refs/tags/1.23.0
Owner: https://github.com/fivetran
Access: public Publication detail:
Token Issuer: https://token.actions.githubusercontent.com
Runner Environment: github-hosted
Publication workflow: ci.yml@8210c2e280ebcd7f097e9d8850d81255de3026f6
Trigger Event: push

File details

Details for the file great_expectations-1.23.0-py3-none-any.whl .

File metadata

Download URL: great_expectations-1.23.0-py3-none-any.whl
Upload date: Sep 10, 2026
Size: 5.7 MB
Tags: Python 3
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for great_expectations-1.23.0-py3-none-any.whl Algorithm Hash digest

SHA256 ed42eb95ac5c4e7a5b1edb7cb811d635acc0b7b456ef4c8543731f57641ccddb Copy

MD5 f9e520c396f62c60bf0232c744bbd921 Copy

BLAKE2b-256 f9447d2a85d84191764130e078d4fc90024a9a554136095a2cb97304dada8fe3 Copy

See more details on using hashes here.

Provenance

The following attestation bundles were made for great_expectations-1.23.0-py3-none-any.whl :

Publisher: ci.yml on fivetran/great_expectations Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
Statement type: https://in-toto.io/Statement/v1
Predicate type: https://docs.pypi.org/attestations/publish/v1
Subject name: great_expectations-1.23.0-py3-none-any.whl
Subject digest: ed42eb95ac5c4e7a5b1edb7cb811d635acc0b7b456ef4c8543731f57641ccddb
Sigstore transparency entry: 2784397684
Sigstore integration time: Sep 10, 2026 Source repository:
Permalink: fivetran/great_expectations@8210c2e280ebcd7f097e9d8850d81255de3026f6
Branch / Tag: refs/tags/1.23.0
Owner: https://github.com/fivetran
Access: public Publication detail:
Token Issuer: https://token.actions.githubusercontent.com
Runner Environment: github-hosted
Publication workflow: ci.yml@8210c2e280ebcd7f097e9d8850d81255de3026f6
Trigger Event: push

Release history Release notifications | RSS feed

This release

1.23.0 This release

Sep 10, 2026 2 files

1.22.0

Aug 31, 2026 2 files

1.21.0

Aug 19, 2026 2 files

1.20.0

Aug 7, 2026 2 files

1.19.1

Jul 24, 2026 2 files

1.19.0

Jul 13, 2026 2 files

1.18.2

Jun 26, 2026 2 files

1.18.1

Jun 11, 2026 2 files

1.18.0

Jun 2, 2026 2 files

1.17.2

May 14, 2026 2 files

1.17.1

May 5, 2026 2 files

1.17.0

Apr 22, 2026 2 files

1.16.1

Apr 15, 2026 2 files

1.16.0

Apr 9, 2026 2 files

1.15.2

Apr 1, 2026 2 files

1.15.1

Mar 13, 2026 2 files

1.15.0

Mar 11, 2026 2 files

1.14.0

Mar 4, 2026 2 files

1.13.1

Mar 4, 2026 2 files

1.13.0

Feb 26, 2026 2 files

1.12.3

Feb 13, 2026 2 files

1.11.3

Jan 29, 2026 2 files

1.11.2

Jan 23, 2026 2 files

1.11.1

Jan 20, 2026 2 files

1.11.0

Jan 12, 2026 2 files

1.10.0

Dec 18, 2025 2 files

1.9.3

Dec 11, 2025 2 files

1.9.2

Dec 3, 2025 2 files

1.9.1

Nov 20, 2025 2 files

1.9.0

Nov 7, 2025 2 files

1.8.1

Oct 30, 2025 2 files

1.8.0

Oct 23, 2025 2 files

1.7.1

Oct 15, 2025 2 files

1.7.0

Oct 9, 2025 2 files

1.6.4

Oct 1, 2025 2 files

1.6.3

Sep 24, 2025 2 files

1.6.2

Sep 20, 2025 2 files

1.6.1

Sep 16, 2025 2 files

1.6.0

Sep 12, 2025 2 files

1.5.11

Sep 4, 2025 2 files

1.5.10

Aug 27, 2025 2 files

1.5.9

Aug 20, 2025 2 files

1.5.8

Aug 7, 2025 2 files

1.5.7

Jul 31, 2025 2 files

1.5.6

Jul 24, 2025 2 files

1.5.5

Jul 11, 2025 2 files

1.5.4

Jul 2, 2025 2 files

1.5.3

Jun 25, 2025 2 files

1.5.2

Jun 18, 2025 2 files

1.5.1

Jun 11, 2025 2 files

1.5.0

Jun 5, 2025 2 files

Yanked

1.4.7

Jun 5, 2025 2 files

Yanked reason: The version accidently was set to 1.4.7 instead of 1.5.0

1.4.6

May 28, 2025 2 files

1.4.5

May 22, 2025 2 files

1.4.4

May 14, 2025 2 files

1.4.3

May 7, 2025 2 files

1.4.2

Apr 24, 2025 2 files

1.4.1

Apr 22, 2025 2 files

1.4.0

Apr 15, 2025 2 files

1.3.14

Apr 9, 2025 2 files

1.3.13

Apr 3, 2025 2 files

1.3.12

Mar 26, 2025 2 files

1.3.11

Mar 19, 2025 2 files

1.3.10

Mar 12, 2025 2 files

1.3.9

Mar 5, 2025 2 files

1.3.8

Feb 26, 2025 2 files

1.3.7

Feb 19, 2025 2 files

1.3.6

Feb 14, 2025 2 files

1.3.5

Feb 3, 2025 2 files

1.3.4

Jan 29, 2025 2 files

1.3.3

Jan 22, 2025 2 files

1.3.2

Jan 17, 2025 2 files

1.3.1

Jan 8, 2025 2 files

1.3.0

Dec 19, 2024 2 files

1.2.6

Dec 11, 2024 2 files

1.2.5

Dec 4, 2024 2 files

1.2.4

Nov 20, 2024 2 files

1.2.3

Nov 14, 2024 2 files

1.2.2

Nov 7, 2024 2 files

1.2.1

Oct 31, 2024 2 files

1.2.0

Oct 24, 2024 2 files

1.1.3

Oct 15, 2024 2 files

1.1.2

Oct 10, 2024 2 files

1.1.1

Oct 8, 2024 2 files

1.1.0

Oct 3, 2024 2 files

1.0.6

Oct 1, 2024 2 files

1.0.5

Sep 19, 2024 2 files

1.0.4

Sep 16, 2024 2 files

Yanked

1.0.3

Sep 12, 2024 2 files

1.0.2

Sep 5, 2024 2 files

1.0.1

Aug 29, 2024 2 files

1.0.0

Aug 22, 2024 2 files

Pre-release

1.0.0a6

Aug 21, 2024 2 files

Pre-release

1.0.0a5

Aug 6, 2024 2 files

Pre-release

1.0.0a4

May 16, 2024 2 files

Pre-release

1.0.0a3

Apr 29, 2024 2 files

Pre-release

1.0.0a2

Apr 15, 2024 2 files

Pre-release

1.0.0a1

Feb 15, 2024 2 files

0.18.22

Oct 25, 2024 2 files

0.18.21

Sep 18, 2024 2 files

0.18.20

Sep 10, 2024 2 files

0.18.19

Jul 15, 2024 2 files

0.18.18

Jul 3, 2024 2 files

0.18.17

Jun 28, 2024 2 files

0.18.16

Jun 18, 2024 2 files

0.18.15

May 28, 2024 2 files

0.18.14

May 22, 2024 2 files

0.18.13

Apr 29, 2024 2 files

0.18.12

Mar 20, 2024 2 files

0.18.11

Mar 14, 2024 2 files

0.18.10

Feb 26, 2024 2 files

0.18.9

Feb 16, 2024 2 files

0.18.8

Jan 11, 2024 2 files

0.18.7

Dec 22, 2023 2 files

Yanked

0.18.6

Dec 20, 2023 2 files

0.18.5

Dec 14, 2023 2 files

0.18.4

Dec 8, 2023 2 files

0.18.3

Nov 16, 2023 2 files

0.18.2

Nov 9, 2023 2 files

0.18.1

Nov 2, 2023 2 files

0.18.0

Oct 30, 2023 2 files

0.17.23

Oct 20, 2023 2 files

0.17.22

Oct 12, 2023 2 files

0.17.21

Oct 6, 2023 2 files

Yanked

0.17.20

Sep 28, 2023 2 files

Yanked reason: Breaks GX Agent

0.17.19

Sep 21, 2023 2 files

0.17.18

Sep 20, 2023 2 files

0.17.17

Sep 18, 2023 2 files

0.17.16

Sep 15, 2023 2 files

0.17.15

Sep 7, 2023 2 files

0.17.14

Sep 1, 2023 2 files

Yanked

0.17.13

Aug 31, 2023 2 files

0.17.12

Aug 24, 2023 2 files

0.17.11

Aug 17, 2023 2 files

0.17.9

Aug 10, 2023 2 files

0.17.8

Aug 4, 2023 2 files

0.17.7

Jul 27, 2023 2 files

0.17.6

Jul 21, 2023 2 files

0.17.5

Jul 13, 2023 2 files

0.17.4

Jul 10, 2023 2 files

0.17.3

Jul 7, 2023 2 files

0.17.2

Jun 29, 2023 2 files

0.17.1

Jun 22, 2023 2 files

0.17.0

Jun 15, 2023 2 files

0.16.16

Jun 8, 2023 2 files

0.16.15

Jun 1, 2023 2 files

0.16.14

May 26, 2023 2 files

0.16.13

May 18, 2023 2 files

0.16.12

May 11, 2023 2 files

0.16.11

May 4, 2023 2 files

0.16.10

Apr 28, 2023 2 files

Yanked

0.16.9

Apr 28, 2023 2 files

0.16.8

Apr 20, 2023 2 files

0.16.7

Apr 13, 2023 2 files

0.16.6

Apr 6, 2023 2 files

0.16.5

Apr 2, 2023 2 files

Yanked

0.16.4

Mar 31, 2023 2 files

0.16.3

Mar 24, 2023 2 files

Yanked

0.16.2

Mar 23, 2023 2 files

0.16.1

Mar 16, 2023 2 files

0.16.0

Mar 10, 2023 2 files

0.15.50

Feb 23, 2023 2 files

0.15.49

Feb 17, 2023 2 files

0.15.48

Feb 9, 2023 2 files

0.15.47

Feb 2, 2023 2 files

0.15.46

Jan 26, 2023 2 files

0.15.45

Jan 26, 2023 2 files

0.15.44

Jan 19, 2023 2 files

0.15.43

Jan 12, 2023 2 files

0.15.42

Jan 5, 2023 2 files

0.15.41

Dec 15, 2022 2 files

0.15.40

Dec 13, 2022 2 files

Yanked

0.15.39

Dec 10, 2022 2 files

Yanked

0.15.38

Dec 9, 2022 2 files

Yanked

0.15.37

Dec 8, 2022 2 files

0.15.36

Dec 1, 2022 2 files

0.15.35

Dec 1, 2022 2 files

0.15.34

Nov 18, 2022 2 files

0.15.33

Nov 17, 2022 2 files

0.15.32

Nov 10, 2022 2 files

0.15.31

Nov 4, 2022 2 files

0.15.30

Nov 3, 2022 2 files

0.15.29

Oct 28, 2022 2 files

0.15.28

Oct 20, 2022 2 files

0.15.27

Oct 13, 2022 2 files

0.15.26

Sep 29, 2022 2 files

0.15.25

Sep 23, 2022 2 files

0.15.24

Sep 19, 2022 2 files

0.15.23

Sep 16, 2022 2 files

0.15.22

Sep 8, 2022 2 files

0.15.21

Sep 1, 2022 2 files

0.15.20

Aug 25, 2022 2 files

0.15.19

Aug 18, 2022 2 files

0.15.18

Aug 11, 2022 2 files

0.15.17

Aug 4, 2022 2 files

0.15.16

Jul 29, 2022 2 files

0.15.15

Jul 21, 2022 2 files

0.15.14

Jul 14, 2022 2 files

0.15.13

Jul 7, 2022 2 files

0.15.12

Jun 30, 2022 2 files

0.15.11

Jun 22, 2022 2 files

0.15.10

Jun 15, 2022 2 files

0.15.9

Jun 9, 2022 2 files

0.15.8

Jun 2, 2022 2 files

0.15.7

May 26, 2022 2 files

0.15.6

May 19, 2022 2 files

0.15.5

May 12, 2022 2 files

0.15.4

May 5, 2022 2 files

0.15.3

Apr 28, 2022 2 files

0.15.2

Apr 21, 2022 2 files

0.15.1

Apr 14, 2022 2 files

0.15.0

Apr 8, 2022 2 files

0.14.13

Mar 31, 2022 2 files

0.14.12

Mar 24, 2022 2 files

0.14.11

Mar 17, 2022 2 files

0.14.10

Mar 10, 2022 2 files

0.14.9

Mar 4, 2022 2 files

0.14.8

Feb 24, 2022 2 files

0.14.7

Feb 17, 2022 2 files

0.14.6

Feb 10, 2022 2 files

0.14.5

Feb 3, 2022 2 files

0.14.4

Jan 28, 2022 2 files

0.14.3

Jan 27, 2022 2 files

0.14.2

Jan 20, 2022 2 files

0.14.1

Jan 13, 2022 2 files

0.14.0

Jan 6, 2022 2 files

0.13.49

Dec 24, 2021 2 files

0.13.48

Dec 23, 2021 2 files

0.13.47

Dec 18, 2021 2 files

0.13.46

Dec 9, 2021 2 files

0.13.45

Dec 2, 2021 2 files

0.13.44

Nov 24, 2021 2 files

0.13.43

Nov 18, 2021 2 files

0.13.42

Nov 12, 2021 2 files

0.13.41

Nov 4, 2021 2 files

0.13.40

Oct 27, 2021 2 files

0.13.39

Oct 21, 2021 2 files

0.13.38

Oct 14, 2021 2 files

0.13.37

Oct 7, 2021 2 files

0.13.36

Sep 30, 2021 2 files

0.13.35

Sep 23, 2021 2 files

0.13.34

Sep 16, 2021 2 files

0.13.33

Sep 9, 2021 2 files

0.13.32

Sep 2, 2021 2 files

0.13.31

Aug 26, 2021 2 files

0.13.30

Aug 23, 2021 2 files

0.13.29

Aug 19, 2021 2 files

0.13.28

Aug 13, 2021 2 files

0.13.27

Aug 12, 2021 2 files

0.13.26

Aug 5, 2021 2 files

0.13.25

Jul 30, 2021 2 files

0.13.24

Jul 22, 2021 2 files

0.13.23

Jul 15, 2021 2 files

0.13.22

Jul 9, 2021 2 files

0.13.21

Jun 30, 2021 2 files

0.13.20

Jun 23, 2021 2 files

0.13.19

Apr 23, 2021 2 files

0.13.18

Apr 22, 2021 2 files

0.13.17

Apr 2, 2021 2 files

0.13.16

Apr 1, 2021 2 files

0.13.15

Mar 26, 2021 2 files

0.13.14

Mar 17, 2021 2 files

0.13.13

Mar 12, 2021 2 files

0.13.12

Mar 5, 2021 2 files

0.13.11

Feb 25, 2021 2 files

0.13.10

Feb 13, 2021 2 files

0.13.9

Feb 8, 2021 2 files

0.13.8

Jan 28, 2021 2 files

0.13.7

Jan 23, 2021 2 files

0.13.6

Jan 21, 2021 2 files

0.13.5

Jan 19, 2021 2 files

0.13.4

Dec 23, 2020 2 files

0.13.3

Dec 15, 2020 2 files

0.13.2

Dec 8, 2020 2 files

0.13.1

Dec 3, 2020 2 files

0.13.0

Dec 1, 2020 2 files

0.12.10

Nov 25, 2020 1 file

0.12.9

Nov 17, 2020 2 files

0.12.8

Nov 16, 2020 2 files

0.12.7

Oct 29, 2020 2 files

0.12.6

Oct 20, 2020 2 files

0.12.5

Oct 19, 2020 2 files

0.12.4

Oct 7, 2020 2 files

0.12.3

Sep 28, 2020 2 files

0.12.2

Sep 22, 2020 2 files

0.12.1

Sep 2, 2020 2 files

0.12.0

Aug 13, 2020 2 files

0.11.9

Jul 30, 2020 2 files

0.11.8

Jul 16, 2020 2 files

0.11.7

Jul 2, 2020 2 files

0.11.6

Jun 27, 2020 2 files

0.11.5

Jun 19, 2020 2 files

0.11.4

Jun 13, 2020 2 files

0.11.3

Jun 13, 2020 2 files

0.11.2

Jun 5, 2020 2 files

0.11.1

May 29, 2020 2 files

0.11.0

May 23, 2020 2 files

Pre-release

0.11.0b0

May 10, 2020 2 files

0.10.12

May 20, 2020 2 files

0.10.11

May 15, 2020 2 files

0.10.10

May 14, 2020 2 files

0.10.9

May 8, 2020 2 files

0.10.8

May 4, 2020 2 files

0.10.7

May 1, 2020 2 files

0.10.6

May 1, 2020 2 files

0.10.5

Apr 29, 2020 2 files

0.10.4

Apr 24, 2020 2 files

0.10.3

Apr 22, 2020 2 files

0.10.2

Apr 21, 2020 2 files

0.10.1

Apr 16, 2020 2 files

0.10.0

Apr 15, 2020 2 files

0.9.11

Apr 10, 2020 2 files

0.9.10

Apr 8, 2020 2 files

0.9.9

Apr 7, 2020 2 files

0.9.8

Apr 3, 2020 2 files

0.9.7

Mar 19, 2020 2 files

0.9.6

Mar 18, 2020 2 files

0.9.5

Mar 13, 2020 2 files

0.9.4

Mar 11, 2020 2 files

0.9.3

Mar 6, 2020 2 files

0.9.2

Feb 22, 2020 2 files

0.9.1

Feb 21, 2020 2 files

0.9.0

Feb 19, 2020 2 files

Pre-release

0.9.0b2

Feb 11, 2020 2 files

Pre-release

0.9.0b1

Dec 12, 2019 2 files

Pre-release

0.9.0b0

Nov 26, 2019 2 files

0.8.8

Feb 7, 2020 2 files

0.8.7

Jan 15, 2020 2 files

0.8.6

Dec 4, 2019 2 files

0.8.5

Nov 19, 2019 2 files

0.8.4.post0

Nov 7, 2019 2 files

0.8.4

Nov 6, 2019 2 files

0.8.3

Oct 29, 2019 2 files

0.8.2.post0

Oct 24, 2019 2 files

0.8.2

Oct 23, 2019 2 files

0.8.1

Oct 16, 2019 2 files

0.8.0

Oct 16, 2019 2 files

Pre-release

0.8.0a4

Oct 11, 2019 2 files

Pre-release

0.8.0a3

Oct 7, 2019 2 files

Pre-release

0.8.0a2

Oct 3, 2019 2 files

Pre-release

0.8.0a1

Sep 30, 2019 2 files

0.7.11

Oct 4, 2019 2 files

0.7.10

Sep 19, 2019 2 files

0.7.9

Sep 18, 2019 2 files

0.7.8

Sep 4, 2019 2 files

0.7.7

Aug 19, 2019 2 files

0.7.6

Aug 12, 2019 2 files

0.7.5

Aug 3, 2019 2 files

0.7.4

Aug 3, 2019 2 files

0.7.3

Jul 29, 2019 2 files

0.7.2

Jul 22, 2019 2 files

0.7.1

Jul 13, 2019 2 files

0.7.0

Jul 4, 2019 2 files

0.6.1

Jun 3, 2019 1 file

0.6.0

May 24, 2019 2 files

0.5.1

Apr 30, 2019 2 files

0.5.0

Apr 25, 2019 2 files

0.4.5

Dec 19, 2018 2 files

0.4.4

Aug 29, 2018 2 files

0.4.3

Jul 12, 2018 2 files

0.4.2

May 17, 2018 2 files

0.4.1

Mar 24, 2018 2 files

0.4.0

Mar 23, 2018 2 files

0.3.2

Feb 8, 2018 2 files

0.3.1

Feb 8, 2018 2 files

0.3.0

Dec 22, 2017 2 files

Pre-release

0.0.1111.post0.dev17

Aug 4, 2023 2 files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page

"PyPI", "Python Package Index", and the blocks logos are registered trademarks of the Python Software Foundation .
© 2026 Python Software Foundation
Site map
Deployed from 490a846

Additional notes

Compatibility: new extra oracle

Highlights

great_expectations[oracle] is a supported install — Oracle is now a published install path: installing the oracle extra brings in the Oracle driver and floors SQLAlchemy at 2.0, so the oracle+oracledb dialect the connection string needs is always available. The SQL dialect installation-commands table documents the new row. (#12091)
```
pip install 'great_expectations[oracle]'
```
Daily and monthly Batch Definitions work on Oracle query assets — Adding a daily or monthly Batch Definition to a query asset on Oracle previously failed with ORA-00907: missing right parenthesis, reported misleadingly as the partition column not being verifiable as a date or datetime. A query asset's SQL is now wrapped whole, so the Batch Definition can be created. As a side effect, a query whose SQL ends in a trailing line comment no longer breaks Batch Definition creation on any backend. (#12162)
```
asset = datasource.add_query_asset(name="orders", query="SELECT id, created_at FROM my_table")
asset.add_batch_definition_daily(name="daily", column="created_at")
```
Consistent verdict for z-score checks on zero or undefined variance — ExpectColumnValueZScoresToBeLessThan used to disagree by backend on a constant column: pandas flagged every row as an outlier, PostgreSQL and SQL Server surfaced a division-by-zero error, and SQLite and MySQL quietly succeeded. All engines now agree that a column with zero or undefined standard deviation succeeds with no unexpected values. Anyone who relied on this Expectation to catch a stuck or constant column should use ExpectColumnStdevToBeBetween with a non-zero min_value instead. (#12145)
```
gxe.ExpectColumnValueZScoresToBeLessThan(column="constant", threshold=1.96, double_sided=True)
# success=True, unexpected_count=0 on every backend
```
Mixed-case column names no longer break uniqueness checks on SQL backends — ExpectColumnValuesToBeUnique raised KeyError: '<column>' on case-insensitive SQL dialects (Databricks, PostgreSQL, Snowflake, SQL Server, Trino) whenever the column name was not already lower case and the result format asked for rows or unexpected indices. It now evaluates normally, so users who pinned to 1.19.1 for this reason can unpin. (#12180)
```
gxe.ExpectColumnValuesToBeUnique(column="CustomerID")  # result_format="COMPLETE"
```
SQL execution engines are reused instead of rebuilt on every validation — Every validation against a SQL datasource used to build a new execution engine, with its own SQLAlchemy engine and connection pool, leaking an idle pooled connection per validation and re-running dialect setup each time. The engine is now cached as intended and rebuilt only when the datasource's connection configuration changes; a validation that follows a schema change still reflects the table afresh. (#12148)
```
datasource.get_execution_engine() is datasource.get_execution_engine()  # now True
```

Changes

Bug fixes

ExpectColumnStdevToBeBetween on SQLite now reports an undefined standard deviation as an observed_value of None — matching every other backend — instead of returning a result with no observed_value and an opaque "user-defined function raised exception" error, for columns with fewer than two non-null values and for empty tables. (#12168)
ExpectColumnValuesToBeUnique no longer raises KeyError on SQL backends when a column name is not lower case and the result format requests rows or unexpected indices. (#12180)
ExpectColumnValueZScoresToBeLessThan now succeeds with no unexpected values on columns whose standard deviation is zero or undefined, on every backend, instead of failing on pandas or raising a division-by-zero error on PostgreSQL and SQL Server. (#12145)
Reading validation results and project YAML no longer fails with a UnicodeDecodeError under a non-UTF-8 system locale: filesystem store reads and project-configuration reads and writes are now pinned to UTF-8. (#12125)
Daily and monthly Batch Definitions can now be added to Oracle query assets, which previously failed with ORA-00907: missing right parenthesis; a query asset whose SQL ends in a line comment also works on every backend now. (#12162)
A query asset whose SQL ends in a -- comment no longer fails validation: the raw SQL is normalized before it is wrapped, so appended text cannot land inside a trailing comment. (#12124)
SQL datasources now reuse their cached execution engine across calls instead of rebuilding it — and leaking a pooled connection — on every validation, while a validation following a schema change still reflects the table afresh. (#12148)

Docs

Updated repository links in the development, docs-contribution, support and compatibility-reference pages to point at the fivetran/great_expectations repository. (#12126)

<details> <summary>Maintenance</summary>

pip install 'great_expectations[oracle]' is now a documented, supported install path, with SQLAlchemy floored at 2.0 so the Oracle dialect is available, and an install row added to the SQL dialect installation-commands table. (#12091)
Updated the docs site dependency joi from 17.13.4 to 17.13.7. (#12176)
Updated the docs site dependency svgo from 3.3.4 to 3.3.5, picking up security hardening. (#12175)
Removed five unreferenced checkpoint test fixtures and an entirely dead test conftest.py. (#12174)
Brought the previously excluded modules under tests/integration/ into the type check, correcting their annotations and removing the covering exclude patterns. (#12173)
Brought the seven previously excluded modules under tests/data_context/ into the type check and removed their exclude patterns. (#12171)
Updated the docs site dependency colord from 2.9.3 to 2.10.0. (#12170)
Brought the partition-and-sample execution engine test modules into the type check, fixing their annotations and removing the covering exclude pattern. (#12169)
Removed the unreachable match_on value key from the not-match-like-pattern-list metric, so the metric declares only the options it actually reads; ExpectColumnValuesToNotMatchLikePatternList never accepted match_on and still rejects it. (#12157)
Brought all modules under tests/core/ into the type check and removed their exclude patterns. (#12140)
Brought the modules under tests/render/ into the type check with annotation-only changes, leaving rendering behavior unchanged. (#12152)
Brought the validator metric-calculator and validation-graph test modules into the type check and removed their exclude patterns, with no production behavior change. (#12149)
Updated the docs site dependency fast-uri from 3.1.5 to 3.1.7, picking up security fixes. (#12151)
Updated the docs site dependency browserslist from 4.28.1 to 4.28.8. (#12155)
Contributors opening pull requests from forks once again receive the welcome comment: the broken Slack notification steps and unreachable workflow conditions were removed, and the pyspark 4 and marshmallow 4 lanes now gate required CI and publishing. (#12159)
Brought tests/test_utils.py, tests/actions/ and tests/checkpoint/test_checkpoint.py into the type check, including fixes so a failed database connection surfaces its original error instead of an AttributeError from the cleanup path. (#12146)
The integration test harness now maps fixture float and datetime columns to portable SQL types, so fixture values are stored as declared instead of being silently rounded or failing table creation on some backends, with new tests pinning the per-backend renderings. (#12147)
Added a contract suite covering create, update and create-or-update for every registered fluent datasource type, and completed the type stubs so nineteen previously untyped factory methods now expose real signatures and return types to callers. (#12141)
Generalized the test harness's data source declaration record so non-SQL data sources can declare support tiers, and derived the lists that gate CI from those declarations rather than maintaining them by hand. (#12110)
Tightened the declared types for resolved metric values, render-content payloads and metric cache keys so callers type-checking against these APIs see fewer false diagnostics, with no runtime or signature changes. (#12116)

</details>

Contributors

Thanks to @siddharthgaur1 (first contribution), @Star-cloud626 (first contribution), @nanjeshramesh, @adimalkar (first contribution), @AnandkumarMall (first contribution), @Ryota-Di (first contribution), @yigitcan-ozturk (first contribution), @MannXo, @iamfeldman (first contribution).

1.22.031 Aug 2026

[BUGFIX] Give the date-part string cast a length Oracle accepts

[BUGFIX] Give the date-part string cast a length Oracle accepts (#12102)
[BUGFIX] Add an Oracle branch to the dialect-regex helper (#12103)
[BUGFIX] Render the derived-table alias in the form each grammar accepts (#12104)
[BUGFIX] Restore the metrics coverage MySQL, SQL Server and Redshift were silently missing (#12107)
[BUGFIX] Render Data Docs pages for results whose meta has no run_id (#12098) (thanks @MannXo)
[BUGFIX] pass usedforsecurity=False on non-security md5 calls for FIPS hosts (#12099) (thanks @nanjeshramesh)
[BUGFIX] Drop the taxi test table by name instead of scanning the database (#12119)
[BUGFIX] Stop add_store from crashing and truncating great_expectations.yml (#12081)
[DOCS] Teach users to install, use, and remove the agent skills bundled with GX (#12074)
[DOCS] Teach integer batch parameters across the documentation (#12066)
[DOCS] Restructure the agent skills page for progressive disclosure (#12106)
[MAINTENANCE] Add Oracle to the SQL test harness with live curated coverage (#12085)
[MAINTENANCE] Make the mypy configuration mean what it says (#12112)
[MAINTENANCE] Remove two structural blockers to type-checking the test tree (#12113)
[MAINTENANCE] Refuse silent relaxations of the type-check configuration (#12115)
[MAINTENANCE] Restore the S3 docs fixtures (#12111)
[MAINTENANCE] Request the docs-creds-needed backends that are live again (#12114)
[MAINTENANCE] Run the Azure Blob docs tests against live storage again (#12117)
[MAINTENANCE] Bound every apt call in the SQL Server ODBC driver install (#12079)
[MAINTENANCE] Add a marshmallow 4 CI lane (#12118)
[CONTRIB] Mask Azure connection-string AccountKey regardless of field order (#12094) (thanks @dkling-it, @hemalrajput18)
[CONTRIB] Fix pandas column-pair validation with non-default indexes (#12097) (thanks @joebasrawi)
[CONTRIB] add explanatory message to NotImplementedError for unsupported regex dialects (#12109) (thanks @ArjunPakhan)
[CONTRIB] Support Marshmallow 4.x (#12092) (thanks @Dev-iL)

Open source →

Additional notes

Aug 31, 2026 2 files

Open source →

Additional notes

Compatibility: marshmallow minimum 3.7.1 → 3.18.0

Highlights

Marshmallow 4 is now supported — Great Expectations now installs and runs against both Marshmallow 3 and Marshmallow 4, so it can be installed alongside deployments that pin Marshmallow 4 (such as Apache Airflow 3.3). The supported range is now marshmallow>=3.18.0 with no upper bound; the declared 3.7.1 floor was unreachable in practice, so no currently-working environment is excluded. (#12092, #12118)
```
pip install great_expectations marshmallow==4.3.1
```
Oracle is now a live-tested backend, with three Oracle defects fixed — Oracle joins the SQL test harness with live curated coverage, and the gaps that coverage exposed are fixed: regex Expectations now execute on Oracle instead of raising, query-based Expectations such as UnexpectedRowsExpectation now run because the derived-table alias is rendered in the form Oracle's grammar accepts, and daily and monthly batch definitions now work because the date-part string cast carries a length Oracle accepts. No other backend's rendered SQL changes. (#12085, #12103, #12104, #12102)
```
batch_definition = asset.add_batch_definition_daily(
    name="daily", column="event_date"
)
```
add_store no longer crashes and empties great_expectations.yml — Calling context.add_store() with an existing store's name and a config containing a store_backend key crashed and left great_expectations.yml at 0 bytes, making the project unloadable. The config is now serialized before the file is opened, so a serialization failure leaves the existing file byte-for-byte intact, and the context id is written as a string that YAML can represent. (#12081)
```
current = context.config.stores[context.expectations_store_name]
context.add_store(
    name=context.expectations_store_name,
    config={
        "class_name": current["class_name"],
        "store_backend": dict(current["store_backend"]),
    },
)
```
Clearer errors for unsupported regex dialects and masked Azure account keys — Regex Expectations run against a SQL dialect with no regex support now report Regex is not supported for dialect <name> in exception_info instead of an empty message, and Azure connection strings are masked regardless of field order so an account key can no longer appear unmasked in a StoreConfigurationError. (#12109, #12094)
Data Docs pages render for validation results with no run_id — context.build_data_docs() silently dropped a validation result's page when the result's meta had no "run_id" key. Such results now render, with run name and run time shown as __none__. (#12098)
```
context.build_data_docs()  # renders a page for every persisted result
```
Documentation for the bundled agent skills — A new environment-setup page teaches how to install, verify, use, upgrade, and remove the agent skills that ship inside the great_expectations package, including the overwrite contract and the three skills as one path. (#12074, #12106)
```
python -m great_expectations skills install
python -m great_expectations skills list
```

Changes

Features

Marshmallow 4.x is now supported alongside Marshmallow 3 on every supported Python version, with the supported range narrowed to marshmallow>=3.18.0 and the <4.0.0 cap removed; config_version bounds checking behaves identically on both majors. (#12092)

Bug fixes

context.add_store() no longer crashes and truncate great_expectations.yml to 0 bytes when re-supplying a store's own config containing a store_backend key; the project config is now serialized before the file is opened, the context id is persisted as a string, and an absent context id stays empty rather than becoming the string "None". (#12081)
The BigQuery taxi test fixtures now drop their table by name during teardown instead of enumerating every schema on the server, removing the multi-minute stalls that cancelled the docs-snippets CI job. Test infrastructure only; no shipped code path changes. (#12119)
Regex Expectations run against a SQL dialect without regex support (such as SQL Server) now report "Regex is not supported for dialect <name>" in exception_info instead of an empty exception message, with the dialect name rendered cleanly. (#12109)
ExpectColumnPairValuesToBeInSet on pandas now returns a verdict instead of a MetricResolutionError when the evaluated rows do not use a zero-based consecutive index, such as after null filtering or with a custom DataFrame index. (#12097)
Non-security md5 calls now pass usedforsecurity=False, so batch identification, dataframe fingerprinting, and partitioning/sampling work on FIPS-enabled hosts. No computed digests change. (#12099)
Azure Blob Storage connection strings are now masked by parsing key=value pairs rather than matching one fixed field order, so a string with reordered fields or no EndpointSuffix no longer raises a StoreConfigurationError containing the raw URL and account key. (#12094)
context.build_data_docs() now renders a page for a validation result whose meta has no "run_id" key (or whose run_id is None), defaulting run name and run time to __none__ instead of silently dropping the page. (#12098)
MySQL, Microsoft SQL Server, and Redshift now declare a backend tier, so they run the metrics parameterizations they were silently absent from (+56, +44, and +56 tests respectively), and a new comparison keeps the two data-source list definitions from parting again; four regex metric modules exclude SQL Server, which has no regex operator. (#12107)
Query-based Expectations such as UnexpectedRowsExpectation now execute on Oracle: the derived-table alias is rendered through one shared helper that omits AS only for the grammar that rejects it, and the literal-boolean predicate rewrite now also applies to Oracle. No other backend's rendered SQL changes. (#12104)
Regex Expectations now execute on Oracle instead of raising NotImplementedError, via a new Oracle branch in the dialect-regex helper that renders REGEXP_LIKE in both positive and negated forms. No other dialect's rendered SQL changes. (#12103)
add_batch_definition_daily and add_batch_definition_monthly now work on Oracle: the multi-date-part partition query's string cast supplies an explicit length for the dialect that requires one, so batch retrieval no longer fails with ORA-00906. Curated coverage for daily and monthly batch definitions was added for every curated backend. (#12102)

Docs

Restructured the "Install agent skills" page for progressive disclosure — what the skills do now comes before prerequisites and install, a new "Use the skills" section follows install, maintenance detail is grouped under keeping the skills up to date, and the --symlink failure wording matches the installer's actual behavior. (#12106)
Documentation that teaches batch parameters now passes integers for numeric batch parameters uniformly across file, SQL, and directory sources, and the prose saying the accepted type depends on the asset family has been removed. (#12066)
Added a documentation page covering the agent skills bundled with GX: what they are, how to install and verify them with python -m great_expectations skills install and skills list, the overwrite contract, the three skills as one path, and how to upgrade and remove them. (#12074)

<details> <summary>Maintenance</summary>

Added a CI lane that installs the Marshmallow 4.x line, asserts the resolution actually landed on 4.x, and runs the unit suite against it. (#12118)
The SQL Server ODBC driver install script now bounds every apt call with a timeout, restricts its index refresh to the Microsoft repository, retries a failed install once, and the docs-snippets job gained a 30-minute timeout, so the step can no longer hang indefinitely. (#12079)
The Azure Blob Storage docs fixtures run against live storage again, with the account URL and container read from environment variables instead of a retired hardcoded host; the Spark ABS fixtures remain off. (#12117)
The docs-creds-needed CI leg now requests the BigQuery, SQL Server, and Redshift backends it already installs, so seven previously-skipped docs fixtures run; Snowflake and Azure stay unrequested and one Redshift fixture stays gated by name for lack of test data. (#12114)
The five pandas S3 docs fixtures run again, with the bucket read from a repository variable and authentication moved from static access keys to a role assumed through GitHub's OIDC provider; the S3 Spark fixtures remain skipped. (#12111)
Added a guard that compares the mypy configuration's relaxation surface against a committed inventory and fails the CI type-check in both directions, so adding an exclusion or relaxing override must be made visible in review. (#12115)
Removed two structural blockers to type-checking the test tree: deleted an __init__.py under a hyphenated, unimportable directory and gave tests/integration/test_script_runner.py its own shell helper instead of importing one from assets/. (#12113)
Cleaned up the mypy configuration so it reflects the codebase: 41 dead exclude patterns, 3 dead overrides, the SQLAlchemy import suppression and 1.x plugin, and 6 inert third-party suppressions removed; the generated version-file exclusion is anchored and the linter-ignore script's path filter corrected. (#12112)
Oracle joins the SQL data-source test harness as a live-tested backend, with a thin-mode oracledb driver requirement for the test lane, a pinned Oracle 21c container, an oracle pytest marker and CI lane, and curated-tier coverage; no great_expectations[oracle] extra is published yet. (#12085)

</details>

Contributors

Thanks to @Dev-iL (first contribution), @ArjunPakhan (first contribution), @joebasrawi (first contribution), @nanjeshramesh, @dkling-it (first contribution), @hemalrajput18 (first contribution), @MannXo (first contribution).

1.21.019 Aug 2026Deprecated

[MAINTENANCE] Sync SparkDBFSDatasource schema with its deprecation notice

[FEATURE] SQL Harness Backend Framework (#12049)
[FEATURE] Trino SQL backend test harness (#12050)
[FEATURE] ClickHouse SQL backend test harness (#12053)
[FEATURE] Ship agent-skill guidance for configuring data sources and expectations (#12061)
[FEATURE] Harden the data-source skill's driver, cadence, and example guidance (#12062)
[FEATURE] Gate project creation on the user having named the directory (#12063)
[FEATURE] Accept integers for numeric batch parameters on every datasource family (#12065)
[FEATURE] Route validated expectations onward into checkpoint orchestration (#12068)
[FEATURE] Make environment-affecting actions reachable-in-flow and mechanically gated (#12073)
[BUGFIX] Restore the install and test steps to py312-min-versions (#12076)
[MAINTENANCE] Bump dompurify from 3.4.12 to 3.4.13 in /docs/docusaurus (#12048)
[MAINTENANCE] Bump nanoid from 3.3.16 to 3.3.18 in /docs/docusaurus (#12052)
[MAINTENANCE] Bump mermaid from 11.15.0 to 11.16.1 in /docs/docusaurus (#12047)
[MAINTENANCE] Pin SingleStore dev image to 0.2.82 instead of :latest (#12054)
[MAINTENANCE] Read the GCS test bucket from a repository variable (#12056)
[MAINTENANCE] Restore the GCS datasource tests and run them in CI (#12058)
[MAINTENANCE] Restore the GCP credentials setup for the docs-snippets job (#12057)
[MAINTENANCE] Run the GCS docs snippet tests (#12059)
[MAINTENANCE] Surface the RFC threshold in the PR template, AGENTS.md, and a new check (#12043)
[MAINTENANCE] Use the upstream sqlalchemy-redshift dialect instead of the GX fork (#12044)
[MAINTENANCE] Drop the gx-redshift CI marker (#12060)
[MAINTENANCE] Add integration coverage for quoted schema names on PostgreSQL (#12051)
[MAINTENANCE] Create BigQuery test tables in the configured dataset and sweep tables, not datasets (#12024)
[MAINTENANCE] Ship version-matched expectation and datasource catalogs (#12055)
[MAINTENANCE] Sync SparkDBFSDatasource schema with its deprecation notice (#12077)
[MAINTENANCE] Remove the "How to Edit This Suite" button from Data Docs (#12078)
[CONTRIB] Add a great-expectations user-agent suffix to S3 clients (#11937) (thanks @goanpeca)
[CONTRIB] Promote multicolumn values equal expectation (#12018) (thanks @AtomicGlance)
[CONTRIB] Add ExpectColumnValuesToNotBeOutliers across Pandas, SQL, and Spark (#12011) (thanks @chavalasantosh)

Open source →

Additional notes

Aug 19, 2026 2 files

Open source →

Additional notes

Compatibility: new extra gcs; gx-sqlalchemy-redshift removed (extra gx-redshift); sqlalchemy-redshift added (extra gx-redshift); sqlalchemy minimum → 1.4.0 (extra redshift)

Highlights

Integer batch parameters work on every datasource family — Numeric batch parameters such as year and month now accept integers on file, directory, and SQL assets alike, so a single batch_parameters dict drives one checkpoint spanning files and a warehouse. Digit strings still work but now emit a deprecation warning. A SQL request that matches nothing also explains why, distinguishing an empty table or column from candidates that exist but did not match, and naming the offending parameter and value. (#12065)
```
checkpoint.run(batch_parameters={"year": 2020, "month": 4})
```
Agent-skill guidance ships with the package — Great Expectations now bundles version-matched guidance for coding agents covering data source configuration, expectation authoring, and checkpoint orchestration, installable into a project's agent discovery directories. The guidance names the right optional dependency group for a missing driver, offers a batching cadence instead of assuming one, carries reuse-safe worked examples, and will not install packages, edit configuration files, create a project directory, or save files unless the user asked for it. (#12061, #12062, #12063, #12068, #12073)
```
python -m great_expectations skills install --target all
python -m great_expectations skills list
```
Two experimental expectations promoted into the core library — ExpectColumnValuesToNotBeOutliers (IQR and standard-deviation methods) and the multicolumn values-equal expectation are now supported core expectations on Pandas, SQL, and Spark, with null-safe evaluation, Gallery metadata, prescriptive rendering, and public exports. (#12011, #12018)
```
gx.expectations.ExpectColumnValuesToNotBeOutliers(
    column="fare_amount", method="iqr", multiplier=1.5
)
```
Redshift installs the upstream SQLAlchemy dialect — pip install 'great_expectations[redshift]' now resolves the upstream sqlalchemy-redshift 1.0.0 dialect with SQLAlchemy 2, replacing the Great Expectations fork and lifting a sqlalchemy<2.0.0 pin that had been holding the dialect back at a three-year-old release. The gx-redshift extra keeps working as a deprecated alias that resolves identically. (#12044)
```
pip install 'great_expectations[redshift]'
```
S3 requests are attributable to Great Expectations — S3 clients built by Great Expectations now send a great-expectations/<version> user-agent suffix, appended to any user-supplied agent string rather than replacing it, so operators and S3-compatible providers can see which requests originate from Great Expectations. The S3 Data Source docs also clarify that endpoint_url is how you connect to a non-AWS S3-compatible store. (#11937)
Data Docs no longer advertises a removed suite-editing workflow — The "How to Edit This Suite" button and its popup, which pointed at a CLI command and notebook workflow that no longer exist, are gone from expectation suite and validation results pages, and expectation suite, profiling, and site index pages no longer render an empty "Actions" card. Validation results pages keep the Actions card and its Show All / Failed Only filter. (#12078)

Deprecations

Digit strings for numeric batch parameters (for example \{"year": "2024", "month": "02"}) are deprecated; pass integers instead (\{"year": 2024, "month": 2}). Removal in 2.0.0. (#12065)
The gx-redshift install extra is deprecated and is now an alias that resolves identically to redshift; use great_expectations[redshift]. Removal in 2.0.0. (#12044)

Changes

Features

S3 clients now carry a great-expectations/<version> user-agent suffix, appended to any user-supplied agent string, and the S3 Data Source docs clarify that endpoint_url connects to an S3-compatible object store. (#11937)
ExpectColumnValuesToNotBeOutliers is now a supported core expectation on Pandas, SQL, and Spark, with IQR and standard-deviation detection, consistent null handling, inclusive threshold boundaries, and a clear error for unsupported methods. (#12011)
The bundled agent skills now state up front, and again at each point they could act, that they will not install dependencies, edit configuration files, or save unrequested files without the user asking; the data source skill's batch-parameter examples were also corrected to use integers. (#12073)
The multicolumn values-equal expectation is promoted into the core library with null-safe equality on Pandas, SQLAlchemy, and Spark, plus Gallery metadata, prescriptive rendering, and public exports. (#12018)
A third bundled agent skill covers checkpoint orchestration — binding assets and suites into validation definitions, grouping them into a named checkpoint with post-run actions, and verifying with one run — and the expectations skill now routes onward into it; writing a session out to a project also persists validation definitions and checkpoints and reports when an object already existed. (#12068)
Numeric batch parameters accept integers on file, directory, and SQL assets, so one batch_parameters dict drives a checkpoint spanning files and SQL; digit strings still work but warn, and a SQL request matching nothing now explains whether the data is absent or the parameter did not match. (#12065)
The bundled agent skills will not create a project directory unless the user has agreed to it and named the path, and the write-out offer is now the end of the flow rather than something done in the same breath as reporting results. (#12063)
The data source skill now names the correct optional dependency group for a missing driver (read from the installed distribution), keeps the batching question open until the verification probe reports the available columns, and carries the reuse guard inside its worked examples so copying one cannot silently replace an existing data source. (#12062)
Great Expectations ships agent-skill guidance for configuring data sources and expectations, installable with python -m great_expectations skills install and listable with skills list; the installer leaves already-correct destinations alone, refuses directories it did not write, and requires --force to replace user-edited copies. (#12061)
ClickHouse is onboarded as a first-class backend in the SQL integration test harness, which along the way fixed a quantile-metric helper that called a nonexistent execution-engine method and two shared test-harness defects. (#12053)
Trino is onboarded onto the SQL backend integration-test harness with a pinned container, a declared backend record, corrected double-quote identifier quoting for the dialect, and full curated-tier coverage. (#12050)
The SQL integration test harness gains a declarative backend framework: a new SQL backend is onboarded by declaring one frozen backend record (schema support, column type overrides, transaction mode, insert parameter limits, table schema items, tiers, and CI wiring) instead of adding dialect-specific branches, with onboarding documentation and a wiring drift check. (#12049)

Bug fixes

The Python 3.12 minimum-version test job installs dependencies and runs tests again, restoring minimum-version coverage and unblocking the required CI gate. (#12076)

<details> <summary>Maintenance</summary>

The "How to Edit This Suite" button and its popup no longer appear in Data Docs, and expectation suite, profiling, and site index pages no longer render an empty "Actions" card; validation results pages keep the Actions card and its validation filter, and the show_how_to_buttons site config option still loads but gates nothing. (#12078)
The shipped SparkDBFSDatasource JSON schema description now includes the deprecation notice the Python API has carried since 1.16.0, so schema-driven consumers see it too. (#12077)
The distribution now ships version-matched expectation and datasource schema catalogs with index files mapping each datasource schema to its add_or_update_* factory method and each expectation to its schema, description, data quality issues, and supported data sources. (#12055)
BigQuery test tables are created in the configured dataset rather than in a per-config dataset, and the cleanup job sweeps stale tables within that dataset instead of querying project-level metadata it lacks permission to read. (#12024)
Added PostgreSQL integration coverage documenting that a quoted schema name on an asset is not honored as quoting today, with the passing bare-name control beside it so the behavior cannot be corrected without the test being updated. (#12051)
Removed the gx-redshift CI launch key now that the Redshift lane selects the canonical redshift marker; the deprecated gx-redshift install extra is unaffected. (#12060)
The redshift extra now installs upstream sqlalchemy-redshift 1.0.0 with sqlalchemy>=1.4.0 instead of the Great Expectations fork, fixing a pin that had been silently installing a three-year-old dialect; gx-redshift remains as a deprecated alias resolving identically. (#12044)
The pull request template and AGENTS.md now state the RFC threshold for new backend support, a new advisory check asks contributors to answer that question when a change looks like new backend support, and the superseded Markdown issue templates that bypassed triage labeling are removed. (#12043)
The GCS documentation snippets run in CI again behind a dedicated --gcs flag, and the Spark-on-GCS guide is migrated to the current asset API so it no longer documents a call that raises. (#12059)
Restored GCP credentials for the docs snippets CI job and made the credentials path absolute so it resolves after the snippet runner changes directories. (#12057)
Corrected the stale patch paths in the pandas and Spark GCS datasource tests and added a gcs_deps marker and requirements file so they actually run in CI. (#12058)
GCS test and docs fixtures read the bucket name from a GX_GCS_TEST_BUCKET environment variable and raise clearly when it is unset; the published GCS guides now show a my_bucket placeholder instead of the real CI bucket. (#12056)
The SingleStore development container image used by CI is pinned to 0.2.82 instead of tracking :latest. (#12054)
Bumped mermaid from 11.15.0 to 11.16.1 in the documentation site dependencies. (#12047)
Bumped nanoid from 3.3.16 to 3.3.18 in the documentation site dependencies. (#12052)
Bumped dompurify from 3.4.12 to 3.4.13 in the documentation site dependencies. (#12048)

</details>

Contributors

Thanks to @goanpeca (first contribution), @chavalasantosh (first contribution), @AtomicGlance (first contribution).

1.20.07 Aug 2026

[BUGFIX] Support standing up a FileDataContext on a read-only filesystem

[BUGFIX] Support standing up a FileDataContext on a read-only filesystem (#12000)
[BUGFIX] Narrow single-pass column_values.unique on SQLAlchemy (Redshift WLM) (#11863) (thanks @leodrivera)
[BUGFIX] deduplicate sql metric aliases to prevent view schema collisions (#10926) (#11905) (thanks @TemidayoA)
[BUGFIX] Report an unmet expectation when a column has no quantiles (#12026)
[BUGFIX] Exclude nulls and fix the rank offset in the SQLite quantile metric (#12008) (thanks @SreeramaYeshwanthGowd)
[MAINTENANCE] Explain why a not-ready issue can't be self-claimed (#11999)
[MAINTENANCE] Bump postcss from 8.5.12 to 8.5.25 in /docs/docusaurus (#12013)
[MAINTENANCE] Bump brace-expansion from 1.1.16 to 1.1.18 in /docs/docusaurus (#12014)
[MAINTENANCE] Namespace ephemeral SQL test schemas under gx_ci_test_ and fix cleanup regex charsets (#12015)
[MAINTENANCE] Bump fast-uri from 3.1.4 to 3.1.5 in /docs/docusaurus (#12019)
[MAINTENANCE] Defer the schema listing in TableAsset.test_connection to the failure path (#12020)
[MAINTENANCE] BigQuery CI (#12016)
[CONTRIB] Promote ExpectColumnValuesToMatchStrftimeFormat to supported-core (#12009) (thanks @nanjeshramesh)

Open source →

Additional notes

Aug 7, 2026 2 files

Open source →

Additional notes

Highlights

Quantile expectations are correct on SQLite and no longer error on all-null columns — ExpectColumnQuantileValuesToBeBetween now selects the right rank on SQLite and ignores null values when computing quantiles, so observed quantiles match the other backends. A column with no non-null values now reports an unmet expectation — success: false with null observed values and per-quantile success details — on every backend instead of raising a TypeError on SQL backends or an IndexError on Spark. (#12008, #12026)
```
import great_expectations.expectations as gxe

suite.add_expectation(
    gxe.ExpectColumnQuantileValuesToBeBetween(
        column="passenger_count",
        quantile_ranges={"quantiles": [0.25, 0.5], "value_ranges": [[1, 2], [1, 3]]},
    )
)
```
Faster expect_column_values_to_be_unique on wide SQL tables — The SQLAlchemy implementation of column_values.unique now scans the source table once through a narrow window over only the target column, and only retrieves full rows (via a narrow duplicate-key join) when SUMMARY or COMPLETE result formats are requested. Wide column-store tables — where the previous query was cancelled by Redshift's workload-management timeouts — now validate reliably. (#11863)
```
import great_expectations.expectations as gxe

gxe.ExpectColumnValuesToBeUnique(column="id")
```
Validating multiple expectations on the same metric no longer fails on strict SQL backends — When several expectations in a suite depend on the same underlying metric, the generated SQL now gives each metric a unique alias, so backends such as Postgres no longer reject the query with Duplicated field name in view schema. (#11905)
File-backed Data Contexts work on read-only, version-controlled projects — gx.get_context(mode="file") now recognizes a project as already set up based on a committed great_expectations.yml alone, instead of requiring the gitignored uncommitted/ runtime directories. A clean checkout on a read-only filesystem is no longer mistaken for an unscaffolded project and no longer crashes during initialization. (#12000)
```
import great_expectations as gx

context = gx.get_context(mode="file", project_root_dir="/path/to/checkout")
```
ExpectColumnValuesToMatchStrftimeFormat is now a supported core Expectation — The Expectation now carries full support metadata and a generated schema, appears in the Expectation Gallery with a properly rendered docstring and examples, and declares a backend matrix of Pandas and Spark (SQL is out of scope). (#12009)
```
import great_expectations.expectations as gxe

gxe.ExpectColumnValuesToMatchStrftimeFormat(
    column="event_date",
    strftime_format="%Y-%m-%d",
    mostly=0.95,
)
```
Adding a table asset is much faster on projects with many schemas — TableAsset.test_connection() now probes the table first and only lists server schemas if that probe fails, purely to refine the error message. On backends where schema listing is a server-wide metadata operation — for example a BigQuery project with thousands of datasets — adding a table asset no longer pays that cost. Table configurations whose schema name did not match the normalized schema listing but were otherwise accessible now succeed. (#12020)
```
asset = datasource.add_table_asset(name="my_asset", table_name="my_table", schema_name="my_schema")
```

Changes

Features

ExpectColumnValuesToMatchStrftimeFormat is promoted to a supported core Expectation, with a corrected and Gallery-formatted docstring, support metadata, a generated JSON schema, a declared Pandas and Spark backend matrix, and expanded test coverage including mostly thresholds. (#12009)

Bug fixes

ExpectColumnQuantileValuesToBeBetween no longer reports a quantile one rank too low on SQLite and no longer raises on columns containing null values; quantile ranks are computed from non-null counts with exact fractional arithmetic, and the MySQL query applies the same null filter. (#12008)
ExpectColumnQuantileValuesToBeBetween now reports an unmet expectation, with null observed values and per-quantile success details, for a column that has no non-null values, instead of raising on SQL backends and Spark; the Spark metric returns one null per requested quantile so column.quantile_values has the same shape on every backend. (#12026)
Validating multiple expectations that share an underlying metric against a SQL backend no longer fails with a duplicated-field-name view schema error, because each bundled metric is now given a unique SQL alias. (#11905)
expect_column_values_to_be_unique on SQLAlchemy backends now runs a single narrow pass over the target column, and only joins back to the source for full-row details under SUMMARY/COMPLETE result formats, eliminating the Redshift workload-management timeouts seen on very wide tables. (#11863)
A file-backed Data Context can now be created against a fully-scaffolded, version-controlled project on a read-only filesystem: an already-set-up project is recognized from its committed great_expectations.yml rather than from gitignored uncommitted/ directories, so a clean checkout is no longer destructively re-scaffolded. (#12000)

<details> <summary>Maintenance</summary>

BigQuery tests run in CI again — the temporary unconditional skip for BigQuery-marked tests was removed — and the external-warehouse CI jobs now fail after 30 minutes instead of hanging for hours. (#12016)
Testing the connection for a table asset now probes the table first and only lists schemas on the failure path to refine the error message, so the operation no longer pays a server-wide metadata scan; error messages are unchanged. (#12020)
Updated the documentation site's fast-uri dependency from 3.1.4 to 3.1.5, which includes a security fix. (#12019)
Ephemeral schemas created by the SQL integration test suite are now namespaced under a gx_ci_test_ prefix, and the stale-schema cleanup patterns were corrected to match hex suffixes so stale schemas are actually swept. No library behavior changes. (#12015)
Updated the documentation site's brace-expansion dependency from 1.1.16 to 1.1.18. (#12014)
Updated the documentation site's postcss dependency from 8.5.12 to 8.5.25. (#12013)
Commenting /assign-me on an issue that is not yet labeled ready for work now gets a posted explanation of why the claim was declined and where to find issues open for claiming, instead of silently doing nothing. (#11999)

</details>

Contributors

Thanks to @SreeramaYeshwanthGowd (first contribution), @TemidayoA (first contribution), @leodrivera, @nanjeshramesh (first contribution).

1.19.124 Jul 2026Deprecated

[MAINTENANCE] Ignore pyOpenSSL X509.get_subject deprecation warning for snowflake

[FEATURE] Self-hosted CLA status check (#11985)
[BUGFIX] Pin actions/checkout to v4.3.1 across workflows (#11988)
[BUGFIX] Sync CLA labels directly from the CLA check, not via the status webhook (#11992)
[BUGFIX] Render id/pk-only unexpected indices instead of raising (fixes #11933) (#11935) (thanks @anxkhn)
[MAINTENANCE] Update CLA links (#11974)
[MAINTENANCE] Ignore pyOpenSSL X509.get_subject deprecation warning for snowflake (#11979)
[MAINTENANCE] CLA Enforcement (#11980)
[MAINTENANCE] Update cla links (#11982)
[MAINTENANCE] Report CLA status on merge-queue commits (#11983)
[MAINTENANCE] enable redshift ci (#11984)
[MAINTENANCE] Bump websocket-driver from 0.7.4 to 0.7.5 in /docs/docusaurus (#11977)
[MAINTENANCE] Bump webpack-dev-server from 5.2.5 to 5.2.6 in /docs/docusaurus (#11990)
[MAINTENANCE] pre-commit autoupdate (ruff 0.15.12 -> 0.15.15) (#11895)
[MAINTENANCE] Assert store access in DatasourceDict just-in-time tests (#11949) (thanks @anxkhn)
[MAINTENANCE] Bump brace-expansion from 1.1.13 to 1.1.16 in /docs/docusaurus (#11991)
[MAINTENANCE] Bump body-parser from 1.20.4 to 1.20.6 in /docs/docusaurus (#11989)
[MAINTENANCE] Bump svgo from 3.3.3 to 3.3.4 in /docs/docusaurus (#11994)
[MAINTENANCE] Bump immutable from 4.3.8 to 4.3.9 in /docs/docusaurus (#11996)
[MAINTENANCE] Bump fast-uri from 3.1.2 to 3.1.4 in /docs/docusaurus (#11995)
[MAINTENANCE] Bump dompurify from 3.4.11 to 3.4.12 in /docs/docusaurus (#11997)
[MAINTENANCE] Add per-metric override hooks for SqlAlchemy row-retrieval providers (#11998)
[CONTRIB] Update distinct values set Expectations to document observed_value contract (#11934) (thanks @EshwarCVS)

Open source →

Additional notes

Jul 24, 2026 2 files

Open source →

Additional notes

Highlights

Data Docs no longer errors when unexpected indices contain only id/pk columns — Validation results that report unexpected indices made up solely of the configured id/pk columns — including Spark and SQL runs and any run with unexpected values excluded — now render a count and index table in Data Docs instead of failing the result page with "No group keys passed!". (#11935)
Contributor License Agreement checks are now run by the project itself — The verification/cla-signed check is posted by the repository's own workflows rather than a third-party hosted app: it is reported on pull request heads and on merge-queue commits, fails closed when contributor status cannot be confirmed, leaves a single guiding comment naming any unsigned or unidentified committer, and keeps the cla-signed / cla-not-signed labels in sync with the check result. CLA signing links now point at the current forms. (#11985, #11983, #11980, #11992, #11982, #11974)
Distinct-values set Expectations document their observed_value contract — Documentation and JSON schemas for the distinct values set Expectations now state that observed_value is always None, and their code examples show unexpected_count and partial_unexpected_list (plus the missing-value variants) instead. (#11934)

Changes

Features

The verification/cla-signed check is now posted by this repository's own workflows instead of a third-party app: it enumerates a pull request's committers, fails closed when it cannot confirm them, supports re-running via an @cla-bot check comment, and posts a single guiding comment naming any unsigned or unidentified committer. (#11985)

Bug fixes

Data Docs now renders the unexpected count and index table when unexpected-index records contain only the id/pk columns, instead of failing the result page with "No group keys passed!". (#11935)
CLA labels on a pull request are now updated in the same run that posts the CLA status, so a pull request no longer keeps a stale cla-not-signed label after signing. (#11992)
Pinned the checkout action to v4.3.1 across workflows, restoring CI runs for contributor pull requests from forks. (#11988)

Docs

Documented that observed_value is always None for the distinct values set Expectations, refreshed their code examples to use unexpected_count and partial_unexpected_list, and synced the published JSON schemas to match. (#11934)

<details> <summary>Maintenance</summary>

SqlAlchemy row-retrieval providers for map metrics (unexpected rows, unexpected index list, and unexpected index query) can now be overridden individually by a subclass without double-registering the metric; registration behavior for all existing map metrics is unchanged. (#11998)
Updated the documentation site's dompurify dependency from 3.4.11 to 3.4.12. (#11997)
Updated the documentation site's fast-uri dependency from 3.1.2 to 3.1.4, picking up security fixes. (#11995)
Updated the documentation site's immutable dependency from 4.3.8 to 4.3.9, picking up security fixes. (#11996)
Updated the documentation site's svgo dependency from 3.3.3 to 3.3.4, picking up a security fix. (#11994)
Updated the documentation site's body-parser dependency from 1.20.4 to 1.20.6, picking up a security fix. (#11989)
Updated the documentation site's brace-expansion dependency from 1.1.13 to 1.1.16, picking up a security fix. (#11991)
Strengthened the tests that verify datasource lookups read their store just in time, so they now genuinely guard that behavior. (#11949)
Updated the pre-commit ruff hook from 0.15.12 to 0.15.15. (#11895)
Updated the documentation site's webpack-dev-server dependency from 5.2.5 to 5.2.6, picking up security fixes. (#11990)
Updated the documentation site's websocket-driver dependency from 0.7.4 to 0.7.5. (#11977)
Re-enabled Redshift tests in CI. (#11984)
The verification/cla-signed status is now reported on merge-queue commits, so the required CLA check can be satisfied in the merge queue instead of hanging pending. (#11983)
Updated the Contributor License Agreement links in CLA.md and the CLA bot message to the current signing forms. (#11982)
Pull requests are now blocked until every committer has a valid Contributor License Agreement signature, with cla-signed / cla-not-signed labels kept in sync with the CLA status. (#11980)
Snowflake connection tests no longer fail on a new pyOpenSSL deprecation warning raised during the TLS handshake. (#11979)
Corrected the Contributor License Agreement form links in CLA.md. (#11974)

</details>

Contributors

Thanks to @anxkhn, @EshwarCVS.

1.19.013 Jul 2026

[BUGFIX] Preserve date-like strings in SQL distinct value sets ( #11947 ) (thanks @yuricavalcanti06 )

[FEATURE] Spark 4 support (#11969)
[BUGFIX] Preserve date-like strings in SQL distinct value sets (#11947) (thanks @yuricavalcanti06)
[BUGFIX] Reject empty regex_list in ExpectColumnValuesToMatchRegexList (#11958) (thanks @anxkhn)
[BUGFIX] Fix broken link in issue assign welcome message (#11961)
[DOCS] Enumerate Ephemeral Data Context use cases in type overview (#11931) (thanks @zozo123)
[MAINTENANCE] remove broken redirect entries (#11936)
[MAINTENANCE] Replace broken Wistia embed with YouTube on GX Core intro page (#11932)
[MAINTENANCE] Skip broken GCP SDK setup in docs-snippets during CI transition (#11959)
[MAINTENANCE] Self-provision data in Snowflake connection tests (#11945)
[MAINTENANCE] Contrib Docs (#11950)
[MAINTENANCE] Add bot-enforced issue claiming and scope stale bot to needs-info issues (#11957)
[MAINTENANCE] Bump launch-editor from 2.12.0 to 2.14.1 in /docs/docusaurus (#11918)
[MAINTENANCE] clean up unused code (#11717)
[MAINTENANCE] Extract type comparison logic into dedicated module (#11798)
[MAINTENANCE] Bump dompurify from 3.4.3 to 3.4.11 in /docs/docusaurus (#11917)
[MAINTENANCE] Make pyarrow compatibility type-ignore environment-independent (#11966)
[MAINTENANCE] Bump joi from 17.13.3 to 17.13.4 in /docs/docusaurus (#11914)
[MAINTENANCE] Remove dead expectation helpers and types/attributes (#11765)
[MAINTENANCE] Databricks CI (#11968)
[MAINTENANCE] Authenticate Databricks CI via OAuth M2M service principal (#11970)
[MAINTENANCE] Remove Codecov from CI and documentation (#11971)

Open source →

Additional notes

Jul 13, 2026 2 files

Open source →

Additional notes

Compatibility: zstandard added (extra spark-connect)

Highlights

Spark 4 and ANSI mode support — Great Expectations now works with Spark 4, including ANSI mode, so you can validate Spark DataFrames on the latest Spark release without pinning to Spark 3. The spark-connect extra now also installs zstandard. (#11969)

import great_expectations as gx

context = gx.get_context()
data_source = context.data_sources.add_spark(name="my_spark")
asset = data_source.add_dataframe_asset(name="my_df")
batch = asset.add_batch_definition_whole_dataframe("batch").get_batch(
    batch_parameters={"dataframe": spark_df}
)

Date-like strings stay strings in SQL value sets — Distinct-value expectations against SQL data sources no longer turn non-ISO, date-like text such as "10-20" into a date, so text bins in a value_set compare correctly against text columns. Only strict YYYY-MM-DD strings are converted to dates. (#11947)
```
import great_expectations.expectations as gxe

gxe.ExpectColumnDistinctValuesToBeInSet(
    column="bin",
    value_set=["10-20", "20-30"],
)
```
Clearer failure for an empty regex_list — ExpectColumnValuesToMatchRegexList now rejects an empty regex_list at construction time with the message "regex_list must not be empty", instead of failing later during validation with an opaque "No objects to concatenate" error. This matches the behavior of ExpectColumnValuesToNotMatchRegexList. (#11958)
```
import great_expectations.expectations as gxe

gxe.ExpectColumnValuesToMatchRegexList(column="my_col", regex_list=[])
# pydantic.ValidationError: ... regex_list must not be empty
```

Changes

Features

Added support for Spark 4, including ANSI mode. (#11969)

Bug fixes

Fixed the broken contributing-guide link in the welcome message posted when an issue is assigned. (#11961)
ExpectColumnValuesToMatchRegexList now fails at construction with "regex_list must not be empty" when given an empty list, instead of raising an opaque error at validation time. (#11958)
Non-ISO date-like strings such as "10-20" in a value_set are no longer parsed into dates for SQL distinct-value expectations; only strict YYYY-MM-DD strings are converted. (#11947)

Docs

Expanded the Ephemeral Data Context description on the Create a Data Context page to name CI pipelines and disposable or read-only compute environments as use cases. (#11931)

<details> <summary>Maintenance</summary>

Removed the Codecov integration from continuous integration and from the project README; no library behavior changes. (#11971)
Databricks test runs now authenticate with a service principal using short-lived OAuth machine-to-machine tokens instead of a stored personal access token. (#11970)
Re-enabled the Databricks test suite and made the target catalog configurable rather than hard-coded. (#11968)
Removed unreachable expectation helper modules and unused internal types that were not part of the public API. (#11765)
Updated the documentation site dependency joi from 17.13.3 to 17.13.4. (#11914)
Made the pyarrow compatibility type-ignore valid whether or not pyarrow is installed, fixing static-analysis failures; no runtime behavior change. (#11966)
Updated the documentation site dependency dompurify from 3.4.3 to 3.4.11. (#11917)
Moved dialect-aware column type comparison used by the type and type-list expectations into a dedicated internal module with expanded unit-test coverage; behavior is unchanged. (#11798)
Removed unused code from the codebase. (#11717)
Updated the documentation site dependency launch-editor from 2.12.0 to 2.14.1. (#11918)
Added a comment-driven issue-claiming workflow (/assign-me, /unassign-me) with automatic release of idle claims, and narrowed issue staleness to issues labeled as needing more information. (#11957)
Rewrote the contributor documentation and added structured issue, bug-report, and request-for-comment templates. (#11950)
Snowflake connection tests now provision their own schema and table, so they no longer depend on pre-existing warehouse state or grants. (#11945)
Skipped the broken Google Cloud SDK setup and one credential-dependent docs snippet test so the documentation test job runs again during the continuous-integration transition. (#11959)
Replaced the broken video embed on the GX Core introduction page with a working YouTube embed of the same demo. (#11932)
Removed broken and malformed entries from the documentation site's redirect list. (#11936)

</details>

Contributors

Thanks to @anxkhn (first contribution), @yuricavalcanti06 (first contribution).

1.18.02 Jun 2026

[MINORBUMP] GX Cloud shutdown: raise on CloudDataContext construction and remove cloud test suites

[MINORBUMP] GX Cloud shutdown: raise on CloudDataContext construction and remove cloud test suites (#11894)
[MAINTENANCE] Remove dead CodeSee architecture diagram workflow (#11886)
[MAINTENANCE] Accept Snowflake parameterized BINARY observed type in type-list expectation test (#11892)
[MAINTENANCE] Skip Microsoft Teams webhook integration tests during CI transition (#11893)
[MAINTENANCE] Pull CI Docker images directly from Docker Hub (#11898)
[MAINTENANCE] Skip external warehouse backend tests during CI transition (#11896)
[MAINTENANCE] Temporarily skip cloud object-store docs tests during CI transition (#11897)

Open source →

Additional notes

Jun 2, 2026 2 files

Open source →

Additional notes

Highlights

GX Cloud paths now fail immediately with a clear explanation — GX Cloud has been shut down. Constructing a CloudDataContext directly, or asking get_context(...) for a cloud context (via mode="cloud", cloud_mode=True, a complete set of cloud_* arguments, or GX_CLOUD_* environment configuration), now raises a GreatExpectationsError right away instead of failing later with an opaque connection error. The message states that GX Cloud has been shut down and that these entry points will be removed in great_expectations 2.0. Non-cloud usage is unchanged, and the cloud classes and parameters remain importable with unchanged signatures through the 1.x line. (#11894)
```
import great_expectations as gx

# Raises GreatExpectationsError:
# "GX Cloud has been shut down, so this no longer functions and will be
#  removed in great_expectations 2.0."
context = gx.get_context(mode="cloud")

# Non-cloud contexts still work as before
context = gx.get_context(mode="file")
```

Deprecations

CloudDataContext and the GX Cloud branch of get_context(...) (the cloud_* parameters, mode="cloud", cloud_mode=True, and GX_CLOUD_* environment configuration) are deprecated and now raise an error; the cloud-only exception, store, config, and identifier symbols remain importable only as shells. Use a non-cloud context such as gx.get_context(mode="file") or gx.get_context(mode="ephemeral"). Removal in 2.0.0. (#11894)

Changes

Features

GX Cloud has been shut down: constructing a CloudDataContext or requesting a cloud context from get_context(...) now raises a GreatExpectationsError explaining the shutdown instead of failing with an opaque connection error. Cloud classes and parameters stay importable with unchanged signatures until they are removed in great_expectations 2.0, and non-cloud usage is unaffected. (#11894)

<details> <summary>Maintenance</summary>

Temporarily skip the cloud object-store documentation examples (S3, GCS, and Azure Blob, plus Athena and AWS Glue) and the BigQuery, Redshift, and Snowflake documentation tests so the documentation-snippet CI job can run while that backend infrastructure is unavailable; Trino documentation tests still run. (#11897)
Tests marked for the Snowflake, BigQuery, Redshift, Databricks, and Athena backends are now skipped with an explicit reason while that test infrastructure is unavailable. (#11896)
CI service container images (Spark, Postgres, MySQL, Trino, and others) are now pulled directly from Docker Hub, and the retired ECR pull-through cache and its login steps have been removed from the workflows. (#11898)
The Microsoft Teams notification integration tests are skipped because the webhook endpoint they posted to has been decommissioned; the mocked unit tests for that action are unchanged. (#11893)
The Snowflake type-list expectation test now accepts the length-parameterized BINARY(8388608) observed type that the Snowflake connector reports for VARBINARY columns. (#11892)
Removed the unused CodeSee architecture diagram workflow and its documentation entry. (#11886)

</details>

Open source →

1.17.214 May 2026

[BUGFIX] Preserve boolean values passed to add_csv_asset (fixes #11206 ) ( #11867 ) (thanks @EshwarCVS )

[BUGFIX] Preserve boolean values passed to add_csv_asset (fixes #11206) (#11867) (thanks @EshwarCVS)
[BUGFIX] Restore SQLAlchemy 1.4 compatibility in column_values_unique (fixes #11875) (#11876) (thanks @ranophoenix)
[MAINTENANCE] [pre-commit.ci] pre-commit autoupdate (#11864)
[MAINTENANCE] Bump fast-uri from 3.1.0 to 3.1.2 in /docs/docusaurus (#11872)
[MAINTENANCE] Bump @babel/plugin-transform-modules-systemjs from 7.28.5 to 7.29.4 in /docs/docusaurus (#11873)
[MAINTENANCE] Bump mermaid from 11.12.2 to 11.15.0 in /docs/docusaurus (#11874)

Open source →

Additional notes

May 14, 2026 2 files

Open source →

Additional notes

Highlights

SQLAlchemy 1.4 users can run uniqueness expectations again — Expectations that resolve the column_values.unique.condition metric no longer fail on SQLAlchemy 1.4 with AttributeError: module 'sqlalchemy' has no attribute 'Select', restoring compatibility for dialects still pinned to SQLAlchemy 1.x (such as ClickHouse, Redshift, and Teradata). (#11876)
Boolean options passed to pandas assets are preserved — Boolean values such as index_col=False handed to add_csv_asset are kept as booleans instead of being silently converted to strings, so pandas interprets them as flags rather than column names. This applies to boolean options across the pandas asset types. (#11867)
```
data_source.add_csv_asset(name="my_asset", path="data.csv", index_col=False)
```

Changes

Bug fixes

Fixed an AttributeError on SQLAlchemy 1.4 when an expectation resolved the column_values.unique.condition metric, restoring SQLAlchemy 1.4 compatibility for unique-value expectations. (#11876)
Boolean arguments passed to add_csv_asset and other pandas assets, such as index_col=False, are no longer coerced to strings and are now applied as the boolean flags pandas expects. (#11867)

<details> <summary>Maintenance</summary>

Updated the documentation site's mermaid dependency from 11.12.2 to 11.15.0. (#11874)
Updated the documentation site's @babel/plugin-transform-modules-systemjs dependency from 7.28.5 to 7.29.4. (#11873)
Updated the documentation site's fast-uri dependency from 3.1.0 to 3.1.2, picking up upstream security fixes. (#11872)
Updated pre-commit hooks, moving ruff from v0.15.9 to v0.15.12. (#11864)

</details>

Contributors

Thanks to @ranophoenix (first contribution), @EshwarCVS (first contribution).

1.17.15 May 2026Security

[BUGFIX] Data Docs uses vulnerable jQuery 3.4.1

[BUGFIX] Spark nested columns break unexpected_index_column_names (GX-3253) (#11835)
[BUGFIX] Spark column names with dots not recognized in BatchData (GX-3274) (#11851)
[BUGFIX] Fix docs-snippets CI broken by sqlalchemy-redshift 1.0.0 (#11857)
[BUGFIX] Pydantic Field alias not respected during expectation validation (#11854)
[BUGFIX] Avoid .toPandas() in Spark multicolumn unexpected values (#11633) (#11861) (thanks @smcl)
[BUGFIX] Use uuid4 for test datasource names to avoid global-RNG collisions (#11862)
[BUGFIX] Data Docs uses vulnerable jQuery 3.4.1 (#11856)
[DOCS] Backfill 1.17.0 changelog entry (#11865)
[MAINTENANCE] Remove deprecated _atomic_prescriptive_template (v0.15.43) (#11847)
[MAINTENANCE] Temporarily run CI on maint/shard-marker-tests branch pushes (#11858)
[MAINTENANCE] Shard snowflake marker-tests + xdist for bigquery/databricks (#11850)
[MAINTENANCE] Bump postcss from 8.5.6 to 8.5.12 in /docs/docusaurus (#11859)

Open source →

Additional notes

May 5, 2026 2 files

Open source →

Additional notes

Compatibility: pytest-split added (extra test)

Highlights

Data Docs now load a patched jQuery — Data Docs pages generated by Great Expectations now reference jQuery 3.7.1 instead of the vulnerable 3.4.1 (CVE-2020-11022, CVE-2020-11023). The Data Docs UI is unchanged and security scanners no longer flag GX-generated pages for this issue. (#11856)
Spark column names containing dots now work — Spark-backed data assets whose column names contain dots (for example Data.Entrega) can now be used in expectations without the spurious "The column X in BatchData does not exist" error. (#11851)
```
batch.validate(gxe.ExpectColumnValuesToNotBeNull(column="Data.Entrega"))
```
Nested Spark struct paths supported in unexpected_index_column_names — Referencing a nested Spark struct path such as Data.evt.id in unexpected_index_column_names no longer raises InvalidMetricAccessorDomainKwargsKeyError; unexpected rows are surfaced keyed by the full dotted path. (#11835)
```
batch.validate(
    gxe.ExpectColumnValuesToBeInSet(column="value", value_set=[1, 2]),
    result_format={
        "result_format": "COMPLETE",
        "unexpected_index_column_names": ["Data.evt.id"],
    },
)
```
Compound uniqueness expectations work on Spark timestamps with Pandas 2.x — expect_compound_columns_to_be_unique and expect_select_column_values_to_be_unique_within_record no longer fail with ValueError: Passing in 'datetime64' dtype with no precision is not allowed. on Spark DataFrames that contain timestamp columns. Entries in partial_unexpected_list are now native Python values (for example datetime.datetime and None) rather than pandas/numpy equivalents. (#11861)
Expectation subclasses can use aliased Pydantic fields — Subclassing a built-in Expectation and declaring a field with Field(alias=...) no longer causes ValidationError: extra fields not permitted when validating a batch. (#11854)
```
class MyExpectation(gxe.ExpectColumnValuesToStartWith):
    regex: str = pydantic.Field(alias="pattern")

batch.validate(MyExpectation(column="name", pattern="^a"))
```

Changes

Bug fixes

Data Docs pages now load jQuery 3.7.1 instead of the vulnerable jQuery 3.4.1, addressing CVE-2020-11022 and CVE-2020-11023, with no visible change to the Data Docs UI. (#11856)
Test datasource names are now generated from UUIDs, removing a source of intermittent name-collision failures in the test suite; no library behavior changes. (#11862)
Compound and within-record uniqueness expectations no longer fail on Spark DataFrames containing timestamp columns under Pandas 2.x, and unexpected-value lists now contain native Python values. (#11861)
Expectation subclasses that declare a field with a Pydantic alias can now be validated without a spurious "extra fields not permitted" error. (#11854)
Restored the documentation-snippet integration tests that broke with sqlalchemy-redshift 1.0.0: the Redshift deployment snippet no longer references removed S3 store backends, and Snowflake key-pair authentication is used where configured. (#11857)
Expectations targeting Spark columns whose names contain dots now resolve correctly instead of reporting that the column does not exist in the batch. (#11851)
Nested Spark struct column paths such as Data.evt.id can now be used for expectation columns and unexpected index columns without raising an invalid-domain-kwargs error, and results are keyed by the full dotted path. (#11835)

Docs

The published changelog now includes the 1.17.0 release section alongside surrounding versions. (#11865)
Integration setup pages now link to the corresponding usage pages, making it easier to move from configuring an integration to using it. (#11848)

<details> <summary>Maintenance</summary>

Updated the documentation site's postcss dependency from 8.5.6 to 8.5.12. (#11859)
Continuous integration now shards and parallelizes the slowest database marker test jobs, cutting overall CI wall-clock time; pytest-split is now part of the development test requirements. (#11850)
Added a temporary continuous-integration trigger for pushes to a maintenance branch so workflow changes could be exercised before merge; no user-facing effect. (#11858)
Removed the long-deprecated Expectation._atomic_prescriptive_template method and its add_values_with_json_schema_from_list_in_params helper, both deprecated in v0.15.43; use _prescriptive_template instead. (#11847)

</details>

1.17.022 Apr 2026Security

[MAINTENANCE] bump jest-environment-jsdom 30.2.0 → 30.3.0 ( CVE-2026-33671 )

[FEATURE] Pass batch_definition_id to GET /expectation-parameters (#11831)
[BUGFIX] Fix pact-broker command not found in record-release CI step (#11819)
[BUGFIX] Fix SingleStoreDB expectations (#11828)
[BUGFIX] Fix Spark metric strftime validation failing with timezone-aware formats (#11817)
[BUGFIX] SQLAlchemy ignores strict_min/strict_max in column value lengths (GX-3252) (#11836)
[BUGFIX] Spark ignores strict_min/strict_max in column value lengths (GX-3252) (#11834)
[DOCS] expect column proportion of non-null values to be between supports forecasted range (#11821)
[DOCS] trino and bigquery (#11747)
[MAINTENANCE] Bump dompurify from 3.3.2 to 3.4.0 in /docs/docusaurus (#11820)
[MAINTENANCE] Add can-i-deploy check to cloud-tests CI (#11801)
[MAINTENANCE] bump jest-environment-jsdom 30.2.0 → 30.3.0 (CVE-2026-33671) (#11823)
[MAINTENANCE] fix minimatch 3.1.2→3.1.5 and lodash-es 4.17.x→4.18.1 in docs (#11827)
[MAINTENANCE] fix path-to-regexp 0.1.12→0.1.13 and picomatch 2.3.1→2.3.2 in docs (CVE-2026-4867, CVE-2026-33671) (#11824)
[MAINTENANCE] allow Can I deploy? step to continue on failure (#11829)
[MAINTENANCE] Move pact contract check into dedicated parallel job (#11822)
[MAINTENANCE] Remove deprecated str support for Validator.validate run_id (#11826)
[MAINTENANCE] Silence mypy assignment error after pyarrow 24.0.0 (#11838)
[MAINTENANCE] Remove DeprecatedMetaMetricProvider and ColumnMetricProvider (#11832)
[MAINTENANCE] Add singlestore documentation (#11837)
[MAINTENANCE] Backfill test on custom sql expectations (#11844)
[MAINTENANCE] Remove deprecated Batch args (data_context, datasource_name, batch_parameters, batch_kwargs) (#11843)
[MAINTENANCE] SingleStore quoted identifier support (#11839)
[MAINTENANCE] Init singlestore db when container is created (#11842)
[MAINTENANCE] Extend pact can-i-deploy retry window to 20 minutes (#11846)

Open source →

Additional notes

Apr 22, 2026 2 files

Open source →

Additional notes

Compatibility: new extra singlestore

Highlights

SingleStore support — Great Expectations now works against SingleStore databases: SingleStoreDB is recognized as its own SQL dialect, regex and uniqueness expectations produce correct results, quoted identifiers are handled, and setup is covered in the documentation. Install with the new singlestore extra. (#11828, #11839, #11837, #11842)
```
pip install 'great_expectations[singlestore]'

import great_expectations as gx

context = gx.get_context()
data_source = context.data_sources.add_sql(
    name="my_singlestore",
    connection_string="singlestoredb://user:password@host:3306/my_db",
)
```
strict_min and strict_max now respected in expect_column_value_lengths_to_be_between on Spark and SQL — Passing strict_min=True or strict_max=True to expect_column_value_lengths_to_be_between previously produced inclusive-bound results on the Spark and SQL backends. Both backends now apply strictly exclusive bounds, matching the documented semantics and the Pandas backend. Non-strict usage is unchanged. (#11834, #11836)
```
import great_expectations.expectations as gxe

suite.add_expectation(
    gxe.ExpectColumnValueLengthsToBeBetween(
        column="name", min_value=2, max_value=4, strict_min=True, strict_max=True
    )
)
```
Timezone-aware strftime formats accepted — ExpectColumnValuesToMatchStrftimeFormat no longer raises a validation error when the format contains %z, and the Spark implementation of the underlying metric now validates timezone-aware formats correctly. (#11812, #11817)
```
import great_expectations.expectations as gxe

gxe.ExpectColumnValuesToMatchStrftimeFormat(
    column="ts", strftime_format="%Y-%m-%d %H:%M:%S%z"
)
```
Forecast store bounds used for windowed expectations — Windowed expectations now send each expectation's batch definition to the expectation-parameters endpoint, so users on the asynchronous forecast store path receive stored forecast bounds instead of falling back to inline training. Checkpoints spanning several batch definitions fetch and merge parameters for each one. (#11831)

Changes

Features

Windowed expectations now pass their batch definition when fetching expectation parameters, so forecast store bounds are used instead of inline forecast training; checkpoints with multiple batch definitions fetch and merge parameters per definition. (#11831)

Bug fixes

expect_column_value_lengths_to_be_between on Spark now honors strict_min and strict_max, excluding boundary lengths as documented. (#11834)
expect_column_value_lengths_to_be_between on SQL data sources now honors strict_min and strict_max, applying strictly exclusive bounds as documented. (#11836)
Fixed the Spark implementation of the strftime-format metric so timezone directives such as %z validate correctly. (#11817)
Fixed regex and uniqueness expectations against SingleStoreDB by recognizing it as its own SQL dialect, and enabled SingleStore tests in continuous integration. (#11828)
Fixed the release pipeline step that recorded contract releases, which failed because the pact command was unavailable; PyPI releases are unblocked with no end-user behavior change. (#11819)
ExpectColumnValuesToMatchStrftimeFormat no longer raises a validation error when the format string includes the %z timezone directive. (#11812)

Docs

Updated the Trino and BigQuery documentation, including coverage of query assets. (#11747)
Documented that expect_column_proportion_of_non_null_values_to_be_between supports a forecasted range. (#11821)

<details> <summary>Maintenance</summary>

Extended the retry window on the continuous-integration contract deployment check to 20 minutes so it can outwait slow provider verification. (#11846)
SingleStore test database is now initialized as a container service rather than through a test fixture. (#11842)
Added support for quoted identifiers when working with SingleStore data sources. (#11839)
Removed the long-deprecated data_context, datasource_name, batch_parameters, and batch_kwargs arguments (and their read-only properties) from the internal Batch class. (#11843)
Added test coverage for custom SQL expectations. (#11844)
Added documentation for connecting to SingleStore. (#11837)
Removed the ColumnMetricProvider alias and its deprecation-warning metaclass, which were deprecated in favor of ColumnAggregateMetricProvider. (#11832)
Restored clean static type checking after the pyarrow 24.0.0 release began shipping type information; no runtime behavior changed. (#11838)
Removed support for passing a plain string as run_id to Validator.validate(); a run identifier or dict is required. (#11826)
Contract verification now runs as its own independently retryable continuous-integration job that waits for provider verification to complete. (#11822)
The contract deployment check no longer fails its continuous-integration job, keeping the signal visible without blocking unrelated work. (#11829)
Resolved two denial-of-service advisories in documentation build dependencies by pinning path-to-regexp to 0.1.13 and picking up picomatch 2.3.2. (#11824)
Resolved documentation tooling advisories by upgrading minimatch to 3.1.5 and lodash-es to 4.18.1. (#11827)
Bumped jest-environment-jsdom to 30.3.0 in the documentation site to pick up a fixed picomatch dependency. (#11823)
Added a contract-compatibility deployment check to the cloud test job in continuous integration and removed a dead log-collection step. (#11801)
Bumped dompurify from 3.3.2 to 3.4.0 in the documentation site. (#11820)

</details>

1.16.09 Apr 2026Breaking

[BUGFIX] Pin invoke==3.0.0 to avoid breaking change in 3.0.2

[MINORBUMP] Deprecate DBFS datasources (GX-2543) (#11759)
[FEATURE] Remove deprecated legacy Pact contract tests (GX-3023) (#11768)
[FEATURE] Migrate pact-python v2 → v3 (GX-3024) (#11769)
[FEATURE] Add client-driven Pact contracts for validation definition and checkpoint CRUD (GX-2730) (#11757)
[FEATURE] Add client-driven Pact contracts for expectation suite CRUD (GX-2729) (#11756)
[FEATURE] Add client-driven Pact contracts for datasource CRUD (GX-2727) (#11754)
[FEATURE] Add multi-step workflow contract test (GX-2731) (#11783)
[FEATURE] Support column.unique_proportion in metric list runs (#11786)
[BUGFIX] Fix suite freshness check after add in ephemeral context (GX-2891) (#11758)
[BUGFIX] Fix unclosed SQLite connections causing ResourceWarning on Python 3.13 (#11766)
[BUGFIX] Pin invoke==3.0.0 to avoid breaking change in 3.0.2 (#11781)
[BUGFIX] Fix default exact_match value in ExpectTableColumnsToMatchSet renderers (#11785)
[DOCS] Jira integration (#11741)
[DOCS] teams integration (#11761)
[DOCS] Hide magic-comment lines when line numbers are enabled (#11731)
[MAINTENANCE] Remove dead code: profile module (#11763)
[MAINTENANCE] Bump lodash from 4.17.23 to 4.18.1 in /docs/docusaurus (#11776)
[MAINTENANCE] Add pact-broker publish step to CI (GX-3025) (#11775)
[MAINTENANCE] Fix pact-broker publish step in CI (#11787)
[MAINTENANCE] Fix pact publish to use PR head SHA (#11790)
[MAINTENANCE] Bump brace-expansion from 1.1.12 to 1.1.13 in /docs/docusaurus (#11777)
[MAINTENANCE] Skip code CI jobs for docs-only PRs (#11792)
[MAINTENANCE] Increase SQL test connection pool size (#11793)
[MAINTENANCE] Use pact regex matcher for Gx-Version header in contract tests (#11791)
[MAINTENANCE] Skip pact publish on release tag CI runs (#11796)

Open source →

Additional notes

Apr 9, 2026 2 files

Open source →

Additional notes

Compatibility: pact-python added (extra cloud); invoke minimum 2.0.0 removed (extra test); pact-python minimum 2.0.1 → 3.1.0 (extra test)

Highlights

column.unique_proportion available in metric list runs — Metric list runs can now compute column.unique_proportion alongside the existing column metrics, so you can retrieve the proportion of unique values per column without a separate run. (#11786)
```
from great_expectations.experimental.metric_repository.metrics import MetricTypes

metrics = [MetricTypes.COLUMN_UNIQUE_PROPORTION]
```
Suites added to an ephemeral context now pass freshness checks — context.suites.add(suite) now returns a suite that matches what was stored, so passing that suite straight into context.validation_definitions.add() in an ephemeral context no longer raises a freshness error. (#11758)
```
suite = context.suites.add(suite)
context.validation_definitions.add(
    gx.ValidationDefinition(name="vd", data=batch_definition, suite=suite)
)
```
Correct exact_match default in ExpectTableColumnsToMatchSet output — Rendered descriptions for ExpectTableColumnsToMatchSet now reflect the expectation's real default for exact_match, so the rendered text no longer contradicts how the expectation actually validates. (#11785)
Microsoft Teams and Jira integration documentation — The documentation now covers setting up the Microsoft Teams and Jira integrations for GX notifications and issue tracking. (#11761, #11741)
No more unclosed-SQLite resource warnings on Python 3.13 — SQLAlchemy execution engines now dispose their connection pool when they are garbage collected, eliminating ResourceWarning: unclosed database noise and the spurious failures it caused on Python 3.13. (#11766)

Deprecations

PandasDBFSDatasource and SparkDBFSDatasource is deprecated; use datasources backed by Unity Catalog volumes, external locations, or workspace files. Removal in 2.0.0. (#11759)

Changes

Features

Metric list runs now support the column.unique_proportion metric, computed alongside other column-level metrics. (#11786)
Added an end-to-end contract test covering the full GX Cloud resource creation flow: datasource, expectation suite, validation definition, and checkpoint. (#11783)
Added client-driven contract tests covering datasource create, read, update, and delete through the Python client. (#11754)
Added client-driven contract tests covering expectation suite add, get, add-or-update, and delete through the Python client. (#11756)
Added client-driven contract tests covering validation definition and checkpoint add, get, and delete through the Python client. (#11757)
Contract testing now runs on pact-python v3, with matchers and interactions updated to the new API and the supported version range moved to 3.x. (#11769)
Removed the legacy hand-crafted HTTP contract tests and their supporting fixtures, which are superseded by the new client-driven contract tests. (#11768)
PandasDBFSDatasource and SparkDBFSDatasource are now marked as deprecated, following Databricks' deprecation of DBFS; the classes still work and will be removed in a future major release. (#11759)

Bug fixes

ExpectTableColumnsToMatchSet renderers now use the correct default value for exact_match in their rendered output. (#11785)
Pinned the development invoke dependency to 3.0.0 to avoid a breaking change introduced in 3.0.2. (#11781)
SQLAlchemy execution engines now dispose their engine and connection pool when garbage collected, preventing unclosed-SQLite resource warnings on Python 3.13. (#11766)
Adding a suite in an ephemeral context now returns a suite that passes later freshness checks, so using it in a validation definition no longer raises a resource freshness error. (#11758)

Docs

Documentation code blocks now keep lines marked as hidden out of view even when line numbers are shown. (#11731)
Added documentation for the Microsoft Teams integration. (#11761)
Added documentation for the Jira integration. (#11741)

<details> <summary>Maintenance</summary>

Release-tag CI runs no longer attempt to re-publish pact contracts, unblocking tagged releases. (#11796)
Contract tests match the Gx-Version request header with a pattern instead of a literal value, so generated contracts are stable across commits. (#11791)
Increased the SQL test connection pool size to reduce connection contention in Databricks and Snowflake test runs. (#11793)
CI skips code and test jobs for pull requests that only change documentation, and adds a single aggregate status check for branch protection. (#11792)
Bumped brace-expansion from 1.1.12 to 1.1.13 in the documentation site dependencies. (#11777)
CI publishes pact contracts against the pull request's head commit instead of the base branch commit, so provider verification runs against the actual changes. (#11790)
Fixed the CI pact-broker publish step so contract publishing actually succeeds and fails loudly when it cannot. (#11787)
CI now publishes generated pact contract files to PactFlow after cloud tests pass, skipping gracefully when no contracts exist. (#11775)
Bumped lodash from 4.17.23 to 4.18.1 in the documentation site dependencies, picking up prototype-pollution and template code-injection fixes. (#11776)
Removed the unreachable great_expectations.profile module and its tests as dead code. (#11763)

</details>

1.15.21 Apr 2026

[FEATURE] Refactor rest_contracts/conftest.py for client-driven Pact testing (GX-2725)

[FEATURE] Refactor rest_contracts/conftest.py for client-driven Pact testing (GX-2725) (#11753)
[DOCS] Small wording change (#11715)
[DOCS] Add documentation for ValidationDefinition.get_unexpected_rows() (#11712)
[MAINTENANCE] Add generic sql test harness (#11718)
[MAINTENANCE] update schema names to be unique to prevent collision between CI runs (#11733)
[MAINTENANCE] Pin localstack to 4.14.0 (#11740)
[MAINTENANCE] Bump flatted from 3.3.3 to 3.4.2 in /docs/docusaurus (#11737)
[MAINTENANCE] CI improvements (#11743)
[MAINTENANCE] [pre-commit.ci] pre-commit autoupdate (#11582)
[MAINTENANCE] Bump yaml from 1.10.2 to 1.10.3 in /docs/docusaurus (#11746)
[MAINTENANCE] ci health report script (#11751)
[CONTRIB] add BigQuery datasource methods to sources.pyi stub file (#11736) (thanks @Julian901)

Open source →

Additional notes

Apr 1, 2026 2 files

Open source →

Additional notes

Highlights

BigQuery datasource methods now surface in IDE autocomplete and type checking — The typed stub for context.data_sources now declares add_bigquery, update_bigquery, add_or_update_bigquery, and delete_bigquery, so BigQuery-specific datasource methods are discoverable in editor autocomplete and recognized by type checkers instead of pushing you toward the generic add_sql method. (#11736)
```
datasource = context.data_sources.add_bigquery(
    name="my_bigquery_ds",
    connection_string="bigquery://my-project/my_dataset",
)
```
New how-to guide: retrieve all unexpected rows — The documentation now includes a "Retrieve all unexpected rows" guide under Run Validations, with a runnable example showing how to get the full set of unexpected rows from a validation definition, plus cross-references from the custom SQL Expectation guide and the result format reference table. (#11712)
```
unexpected_rows = validation_definition.get_unexpected_rows(batch_parameters={})
```

Deprecations

The run_rest_api_pact_test REST contract test helper is deprecated; use the client-driven Pact test approach built on the pact_cloud_context fixture. Removal in 2.0.0. (#11753)

Changes

Features

REST contract testing can now be driven from the client side: a pact_cloud_context fixture builds a CloudDataContext against the Pact mock server without real cloud credentials, a shared data-context configuration response and interaction helper are available for reuse, and the older run_rest_api_pact_test helper is marked deprecated. (#11753)
BigQuery datasource methods (add_bigquery, update_bigquery, add_or_update_bigquery, delete_bigquery) and the BigQueryDatasource type are now declared in the datasources type stub, so they appear in IDE autocomplete and type checking instead of requiring the generic add_sql method. (#11736)

Docs

Added a "Retrieve all unexpected rows" how-to guide with a runnable example script, a sidebar and landing-page entry, and cross-references from the custom SQL Expectation guide and the result format reference table. (#11712)
Applied a small documentation wording change based on engineering feedback. (#11715)

<details> <summary>Maintenance</summary>

Added a script that inspects the last 28 days of scheduled CI runs and generates a markdown CI health report. (#11751)
Bumped the docs site yaml dependency from 1.10.2 to 1.10.3. (#11746)
Updated pre-commit hooks, moving ruff-pre-commit from v0.14.9 to v0.15.4. (#11582)
Fixed three CI configuration problems: a too-short timeout for cloud services, a mismatched docs matrix key, and a malformed Spark command. (#11743)
Bumped the docs site flatted dependency from 3.3.3 to 3.4.2. (#11737)
Pinned localstack to 4.14.0 to restore broken CI runs. (#11740)
Made test schema names unique so concurrent Databricks CI runs no longer clean up each other's test setup and fail with table-not-found errors. (#11733)
Added a generic SQL datasource test harness (GenericSQLDatasourceTestConfig) to make it easier to try out new SQL datasources. (#11718)

</details>

Contributors

Thanks to @Julian901 (first contribution).

1.15.011 Mar 2026

[MINORBUMP] SQL Server and Fabric Data Sources

[MINORBUMP] SQL Server and Fabric Data Sources (#11686)
[MINORBUMP] Add ValidationDefinition.get_unexpected_rows() (#11711)
[BUGFIX] retry individual metrics when bulk resolution fails (#11708)
[DOCS] reflect changes to Validate button (#11691)
[DOCS] loose ends for new result format option (#11705)
[DOCS] Set workspace ID for the agent (#11709)
[DOCS] DOC-1168 - How to edit expectations using the API (#11697)
[DOCS] Slack alerts (#11681)
[MAINTENANCE] Bump immutable from 4.3.7 to 4.3.8 in /docs/docusaurus (#11703)
[MAINTENANCE] Bump svgo from 3.3.2 to 3.3.3 in /docs/docusaurus (#11701)
[MAINTENANCE] Bump dompurify from 3.3.1 to 3.3.2 in /docs/docusaurus (#11706)
[MAINTENANCE] Suppress mypy assignment errors in Trino compatibility module (#11707)

Open source →

Additional notes

Mar 11, 2026 2 files

Open source →

Additional notes

Highlights

Fetch all unexpected rows from an UnexpectedRowsExpectation — ValidationDefinition.get_unexpected_rows() returns every failing row for an UnexpectedRowsExpectation, without the 200-row cap applied to validation results. Validation results also gained an ExpectationValidationResult.expectation property and an ExpectationSuiteValidationResult.batch_parameters property, so you can feed a failed result straight back in to retrieve its rows. (#11711)
```
result = validation_definition.run(batch_parameters={"year": 2026, "month": 3})

for evr in result.results:
    if not evr.success:
        rows = validation_definition.get_unexpected_rows(
            evr.expectation,
            batch_parameters=result.batch_parameters,
        )
        if rows:
            write_to_quarantine(rows)
```
Documentation for SQL Server and Fabric data sources — The docs now cover creating and using SQL Server and Microsoft Fabric data sources. (#11686)
A single failing metric no longer fails the whole batch of metrics — When bulk metric resolution hits an error, metrics are now retried individually, so one problematic metric no longer causes every metric in the run to error. (#11708)

Changes

Features

Added ValidationDefinition.get_unexpected_rows() to fetch all failing rows for an UnexpectedRowsExpectation without the 200-row cap, plus an ExpectationValidationResult.expectation property and an ExpectationSuiteValidationResult.batch_parameters property for post-run workflows. (#11711)

Bug fixes

Fixed a regression where an error in a single metric caused every metric to fail; metrics are now retried individually when bulk resolution fails. (#11708)

Docs

Added documentation for setting up Slack alerts. (#11681)
Added a section to the Manage Expectations documentation covering how to edit expectations using the API. (#11697)
Updated the agent deployment documentation to explain how to set a default workspace ID. (#11709)
Documented the new result format option in the Core docs as well as Cloud, including how partial_unexpected_count controls the number of values shown in partial_missing_list. (#11705)
Updated documentation to reflect the current Validate button behavior and removed references to the share button, which no longer exists. (#11691)
Documented support for SQL Server and Fabric data sources. (#11686)

<details> <summary>Maintenance</summary>

Silenced new mypy assignment errors in the Trino compatibility module that appeared after the trino package began shipping type information. (#11707)
Bumped dompurify from 3.3.1 to 3.3.2 in the documentation site dependencies. (#11706)
Bumped svgo from 3.3.2 to 3.3.3 in the documentation site dependencies. (#11701)
Bumped immutable from 4.3.7 to 4.3.8 in the documentation site dependencies. (#11703)

</details>

1.13.14 Mar 2026Deprecated

[MAINTENANCE] Deprecate schema_name on all TableAssets

[FEATURE] Add trust_server_certificate to SQLServerDatasource (#11694)
[DOCS] Add ServiceNow to email alerts Cloud documentation (#11669)
[MAINTENANCE] Bump qs from 6.14.1 to 6.14.2 in /docs/docusaurus (#11660)
[MAINTENANCE] Improve error message on ConfigStr substitution (#11693)
[MAINTENANCE] Add sentry error tracking for the Doc Site (#11695)
[MAINTENANCE] Install ODBC driver in docs-creds-needed (#11698)
[MAINTENANCE] Deprecate schema_name on all TableAssets (#11689)

Open source →

Additional notes

Mar 4, 2026 2 files

Open source →

Additional notes

Highlights

Trust a SQL Server certificate without turning off encryption — SQL Server and Fabric data sources accept a new trust_server_certificate option, so you can connect to a server presenting a self-signed or otherwise untrusted certificate while keeping encryption enabled instead of weakening encrypt to "Optional". The option works with both SQL Server authentication and Entra ID. (#11694)
```
import great_expectations as gx

context = gx.get_context()
datasource = context.data_sources.add_sql_server(
    name="my_sql_server",
    host="my-host",
    database="my_database",
    username="my_user",
    password="my_password",
    trust_server_certificate=True,
)
```
Config variable substitution errors no longer echo secret text — When a password or secret contains a literal $, Great Expectations no longer includes the text following the $ in the resulting missing-config-variable error message, so part of the secret is not leaked in logs. The error guidance also no longer points at the retired $MY_CONFIG_VAR substitution syntax. (#11693)

Deprecations

The schema_name parameter on TableAsset and add_table_asset is deprecated; use the schema configured on the SQL data source's connection string. Removal in 2.0.0. (#11689)

Changes

Features

SQL Server and Fabric data sources accept a new trust_server_certificate option, letting you trust a self-signed or untrusted server certificate while keeping the connection encrypted, with both SQL Server authentication and Entra ID. (#11694)

Docs

The Cloud email alert documentation now lists ServiceNow as a supported third-party service and includes *.service-now.com in the default allowed email domains. (#11669)

<details> <summary>Maintenance</summary>

The schema_name parameter on TableAsset and add_table_asset is deprecated; table assets now resolve their schema from the SQL data source they belong to, so specify the schema in the data source's connection configuration instead. (#11689)
Installed the SQL Server ODBC driver in the credentials-backed documentation test step so SQL Server examples in the docs are exercised in CI. (#11698)
Added Sentry error tracking to the documentation site, initialized early enough to capture errors that occur before the page finishes loading; the DSN comes from a SENTRY_DSN environment variable and no performance data is collected. (#11695)
Config variable substitution errors no longer include text that follows a literal $ in a password or secret, avoiding partial secret leakage, and their guidance no longer references the removed $MY_CONFIG_VAR syntax. (#11693)
Updated the documentation site's qs dependency from 6.14.1 to 6.14.2. (#11660)

</details>

Open source →

1.13.026 Feb 2026Deprecated

[MAINTENANCE] remove deprecated store backends

[MINORBUMP] ExpectColumnDistinctValuesToBeInSet with database-pushed comparison (#11614)
[MINORBUMP] ExpectColumnDistinctValuesToContainSet with database-pushed comparison (#11615)
[MINORBUMP] ExpectColumnDistinctValuesToEqualSet with database-pushed comparison (#11616)
[FEATURE] Auto-strip ORDER BY for MSSQL COUNT(*) subqueries (#11670)
[FEATURE] Add FabricDatasource (#11685)
[BUGFIX] Fix Databricks identifier quoting in batch query compilation. (#11671)
[BUGFIX] Roll back connection after failed retry in _execute_query_with_recovery to prevent SQL Server teardown hang (#11680)
[DOCS] ExpectAI for the agent (#11644)
[DOCS] fix ExpectAI for the agent prereqs (#11678)
[MAINTENANCE] Add azure ad service principal auth connection details (#11653)
[MAINTENANCE] Add SQLServerDatasource schema (#11662)
[MAINTENANCE] Remove unsupported Entra ID Password authentication (#11665)
[MAINTENANCE] Human-readable MSSQL test connection exceptions (#11661)
[MAINTENANCE] Only run python 3.13 marker tests on PRs (#11666)
[MAINTENANCE] Rename MSSQL references to SQL Server for brand consistency (#11674)
[MAINTENANCE] Avoid unnecessary _get_default_value() calls for non-field keys (#11626) (thanks @jni-bot)
[MAINTENANCE] Remove problematic runtime context fixture (#11683)
[MAINTENANCE] Remove Pandas Upper Pin (#11677)
[MAINTENANCE] Normalize SQL Server column type metrics (#11684)
[MAINTENANCE] remove deprecated store backends (#11675)

Open source →

Additional notes

Feb 26, 2026 2 files

Open source →

Additional notes

Compatibility: altair minimum 4.2.1 → 5.0.0; new extra fabric; removed extra mssql; new extra sql-server

Highlights

Microsoft Fabric datasource — You can now connect to Microsoft Fabric with the new Fabric datasource, which authenticates with an Entra ID service principal. Install it with the new fabric extra. (#11685, #11662)

import great_expectations as gx

context = gx.get_context()
datasource = context.data_sources.add_fabric(
    name="my_fabric",
    host="my-workspace.datawarehouse.fabric.microsoft.com",
    database="my_warehouse",
    client_id="<client-id>",
    client_secret="<client-secret>",
)

Distinct-value set expectations now compare in the database — ExpectColumnDistinctValuesToBeInSet, ExpectColumnDistinctValuesToContainSet, and ExpectColumnDistinctValuesToEqualSet now push set comparison into the database instead of pulling every distinct value into memory, so they stay fast and produce small results on high-cardinality columns. Results no longer include the full list of distinct values as observed_value; instead they report unexpected_count/partial_unexpected_list and/or missing_count/partial_missing_list, each capped at 20 values. (#11614, #11615, #11616)
```
import great_expectations.expectations as gxe

suite.add_expectation(
    gxe.ExpectColumnDistinctValuesToBeInSet(
        column="my_col",
        value_set=["a", "b", "c"],
    )
)
```
"SQL Server" naming throughout, including the pip extra — User-facing references to MSSQL are now written as SQL Server. Install SQL Server support with the renamed extra. (#11674)
```
pip install 'great_expectations[sql-server]'
```
pandas 3 support — The upper pin on pandas has been removed, so Great Expectations can be installed alongside pandas 3. BigQuery reads fall back to pandas_gbq.read_gbq, and chart rendering works with pandas 3's new string dtype default (requires altair 5). (#11677)
ExpectAI documentation for the agent — The documentation now covers ExpectAI for the agent, including its prerequisites. (#11644, #11678)

Changes

Features

ExpectColumnDistinctValuesToEqualSet now compares the value set inside the database rather than loading all distinct values into memory. Results return observed_value: None along with unexpected_count, partial_unexpected_list, missing_count, and partial_missing_list (each capped at 20 values), and the rendered output marks unexpected and missing values accordingly. (#11616)
ExpectColumnDistinctValuesToBeInSet now compares the value set inside the database rather than loading all distinct values into memory. Results return observed_value: None along with unexpected_count and partial_unexpected_list (capped at 20 values), and the descriptive value-counts bar chart is no longer produced. (#11614)
ExpectColumnDistinctValuesToContainSet now compares the value set inside the database rather than loading all distinct values into memory. Results return observed_value: None along with missing_count and partial_missing_list (capped at 20 values), and the rendered output marks missing values. (#11615)
Added a Microsoft Fabric datasource with add_fabric(), update_fabric(), and delete_fabric() APIs, authenticated with an Entra ID service principal. (#11685)
A top-level ORDER BY in a user-supplied query is now stripped automatically when that query is wrapped in a row count, so SQL Server no longer rejects it. ORDER BY inside window functions or nested subqueries, and queries using OFFSET, are left untouched. (#11670)

Bug fixes

When a metric computation fails twice on a SQL Server connection, the connection is now explicitly rolled back before the error is raised, so closing the connection no longer hangs. (#11680)
SQL expectations using the {batch} placeholder on Databricks no longer fail with a cast error, because batch queries are now compiled with the datasource's own dialect so identifiers are quoted correctly. (#11671)

Docs

Refined the prerequisites documentation for using ExpectAI with the agent. (#11678)
Added documentation for using ExpectAI with the agent. (#11644)

<details> <summary>Maintenance</summary>

Store backend implementations that were deprecated in the v1 release, along with the documentation examples that referenced them, have been removed. (#11675)
SQL Server column types are now reported consistently between the metric repository and ExpectColumnValuesToBeOfType/ExpectColumnValuesToBeInTypeList, compared case-insensitively and without COLLATE clauses in the type string. (#11684)
The upper pin on pandas has been removed so Great Expectations works with pandas 3, with BigQuery reads falling back to pandas_gbq.read_gbq and chart rendering updated for the new string dtype default. (#11677)
Test helpers now build an ephemeral context via gx.get_context(mode="ephemeral") instead of the removed build_in_memory_runtime_context helper, eliminating a source of flaky tests. (#11683)
Removed the repeated noisy _get_default_value called with key ... but it is not a known field INFO log messages emitted during checkpoint validation. (#11626)
MSSQL references are now named SQL Server throughout: the pip extra is sql-server instead of mssql, and related enum members, helper names, pytest markers, and the test CLI flag were renamed to match. The SQLAlchemy dialect value mssql is unchanged. (#11674)
Datasource marker tests now run against a single Python version on pull requests, with the full version matrix reserved for releases. (#11666)
Failed SQL Server connection tests now report human-readable error messages. (#11661)
Removed Entra ID Password authentication, which Microsoft's mandatory MFA enforcement makes unusable. (#11665)
Added a published JSON schema for the SQL Server datasource, documenting its connection-detail options. (#11662)
Added Azure AD service principal authentication details for SQL Server connections and corrected the casing of the authentication query parameter used for Azure AD password authentication. (#11653)

</details>

1.12.313 Feb 2026Deprecated

[DOCS] remove info about deprecated DBFS

[MINORBUMP] Add support for pd.Timestamp for datetime comparison operations (#11637) (thanks @subediparas5)
[FEATURE] Add sql server types stubs and use api in integration tests (#11643)
[FEATURE] MSSQL support for UnexpectedRowsExpectation (#11646)
[FEATURE] SQL Server Azure AD password authentication (#11645)
[FEATURE] MSSQL schema support (#11649)
[FEATURE] Support asymmetric quoted identifiers in dialect quoting (#11652)
[BUGFIX] Remove table domain key from ExpectColumnToExist (#11630)
[BUGFIX] Fix missing unexpected_index_query for ExpectCompoundColumnsToBeUnique on SQL (#11639)
[DOCS] Agent request next steps (#11619)
[DOCS] define asterisks in core result format docs (#11631)
[DOCS] Fix link to expect_table_row_count_to_equal_other_table in documentation (#11634) (thanks @teixeirazeus)
[DOCS] Email alerts (#11628)
[DOCS] Revise contribution guidelines and readiness criteria (#11638) (thanks @adeola-ak)
[DOCS] Add an Expectation using the GX Cloud API (#11567)
[DOCS] remove info about deprecated DBFS (#11648)
[MAINTENANCE] Remove duplicate flaky test (#11623)
[MAINTENANCE] Bump diff from 3.5.0 to 3.5.1 in /docs/docusaurus (#11627)
[MAINTENANCE] Add database-pushdown metrics for distinct values set comparisons (#11629)
[MAINTENANCE] Bump webpack from 5.94.0 to 5.104.1 in /docs/docusaurus (#11636)
[MAINTENANCE] SQLServerDatasource with SQL Server authentication (#11640)
[MAINTENANCE] Add start_period to mercury healthcheck to prevent flaky CI failures (#11655)
[MAINTENANCE] Bump docker compose timeout to 3 min (#11659)
[MAINTENANCE] Dispose of mssql connections in integration tests (#11663)

Open source →

Additional notes

Feb 13, 2026 2 files

Open source →

Additional notes

Highlights

SQL Server datasources with Azure AD password authentication — You can now connect to SQL Server with a flat set of connection keyword arguments, including Azure Active Directory password authentication, without hand-building a connection string. (#11645, #11640, #11643)
```
context.data_sources.add_sql_server(
    name="my_sql_server",
    host="my-server.database.windows.net",
    database="my_db",
    username="user@example.com",
    password="${MY_PASSWORD}",
)
```
Broader Microsoft SQL Server support — SQL Server now works with schemas, with bracket-quoted identifiers such as [my column], and with UnexpectedRowsExpectation queries. (#11649, #11652, #11646)
unexpected_index_query is returned for ExpectCompoundColumnsToBeUnique on SQL — ExpectCompoundColumnsToBeUnique run against SQL data sources now returns unexpected_index_query when you request it with return_unexpected_index_query=True or use the COMPLETE result format, so you can retrieve every failing row beyond the 200-row unexpected_list limit. COMPLETE also now honors return_unexpected_index_query=False when you set it explicitly. (#11639)
```
result = batch.validate(
    ExpectCompoundColumnsToBeUnique(column_list=["a", "b"]),
    result_format={"result_format": "COMPLETE"},
)
print(result.result["unexpected_index_query"])
```
pandas Timestamp values accepted in datetime comparisons — Datetime comparison expectations now handle pandas.Timestamp values correctly, checking the most specific type first so Timestamps are no longer mis-handled as plain dates. (#11637)

Changes

Features

Dialect quoting now supports asymmetric identifier quote characters, so SQL Server bracket-quoted identifiers such as [my column] are handled correctly (double-quoted identifiers are also accepted for SQL Server). (#11652)
SQL Server data sources now support schemas, including schema names that contain upper-case characters. (#11649)
Added SQL Server Azure AD password authentication and a flat keyword-argument style for add_sql_server, update_sql_server, and add_or_update_sql_server, with type stubs for both calling styles. (#11645)
UnexpectedRowsExpectation now works against SQL Server, including queries that previously relied on unsupported SQL constructs. (#11646)
Added SQL Server type stubs and switched the integration tests to the public SQL Server datasource API. (#11643)

Bug fixes

Datetime comparison operations now accept pandas.Timestamp values, checking Timestamp before datetime and date so timestamps are compared correctly. (#11637)
ExpectCompoundColumnsToBeUnique on SQL data sources now returns unexpected_index_query when requested or when using the COMPLETE result format, and COMPLETE respects return_unexpected_index_query=False when explicitly set. (#11639)
Removed the vestigial, unsupported table domain key from ExpectColumnToExist. (#11630)

Docs

Removed documentation about the deprecated DBFS support. (#11648)
Added documentation for adding an Expectation using the GX Cloud API. (#11567)
Revised the contribution guidelines to encourage pull requests for new features and to clarify the acceptance criteria for contributions. (#11638)
Added documentation for email alerts. (#11628)
Fixed a broken documentation link to expect_table_row_count_to_equal_other_table. (#11634)
The core result format documentation now defines the meaning of the asterisks used in its tables. (#11631)
Added documentation describing the next steps after an agent request. (#11619)

<details> <summary>Maintenance</summary>

Integration tests now dispose of SQL Server connections when they finish. (#11663)
Raised the container startup timeout used by the test suites to three minutes to reduce flaky CI failures. (#11659)
Added a healthcheck start period and RabbitMQ readiness check to the local Mercury docker-compose stack to prevent flaky CI failures. (#11655)
Introduced a SQLServerDatasource with structured SQL Server authentication connection details, validating that connection URLs use the mssql+pyodbc scheme and supporting config-substituted passwords. (#11640)
Bumped webpack from 5.94.0 to 5.104.1 in the documentation site. (#11636)
Added database-pushdown metrics for distinct-value set comparisons (column.distinct_values.not_in_set, column.distinct_values.not_in_set.count, column.distinct_values.missing_from_column, and column.distinct_values.missing_from_column.count) that evaluate set comparisons in the database instead of fetching all distinct values into memory, with type coercion for date strings. (#11629)
Bumped diff from 3.5.0 to 3.5.1 in the documentation site. (#11627)
Removed a duplicated flaky pandas result-format test from the test suite. (#11623)

</details>

Contributors

Thanks to @subediparas5, @teixeirazeus (first contribution).

1.11.012 Jan 2026Deprecated

[MAINTENANCE] Ignore DeprecationWarning emitted by deps

[MINORBUMP] Format unexpected_rows as dicts in map expectation validation results (#11591)
[BUGFIX] batch.columns() returns empty list for Redshift batches (#11534) (thanks @leodrivera)
[DOCS] reframe and clarify Data Source limitations (#11570)
[DOCS] Prevent copying hidden lines from code blocks when pressing copy button (#11571)
[DOCS] remove temporary row conditions notes (#11581)
[DOCS] remove beta from ExpectAI (#11590)
[MAINTENANCE] Improve local type checking developer experience (#11574)
[MAINTENANCE] Add --pty and --no-pty flags to invoke deps (#11586)
[MAINTENANCE] Serialize unexpected_rows for all Map expectations when opt-in flag is provided (#11583)
[MAINTENANCE] Ignore DeprecationWarning emitted by deps (#11587)
[MAINTENANCE] Include SUMMARY in result_format query support (#11594)

Open source →

Additional notes

Jan 12, 2026 2 files

Open source →

Additional notes

Highlights

Unexpected rows are returned as dictionaries for Map expectations — Map expectation validation results now report unexpected_rows as dictionaries keyed by column name instead of database-specific row objects rendered as tuples, so results are easier to parse and no longer depend on an opt-in flag. (#11591, #11583)
```
result = batch.validate(expectation)
# result["result"]["unexpected_rows"]
# [{"col_a": 1.0, "col_b": 1.0, "col_c": 2.0}]
```
Column-based validations work on Redshift batches — batch.columns() no longer returns an empty list for Redshift batches on clusters with restricted information_schema access, and table names given as "schema.table" are resolved correctly, so column-based expectations run instead of failing with a metric domain error. (#11534)
```
batch = batch_definition.get_batch()
print(batch.columns())
```

Unexpected index query available with SUMMARY result format — return_unexpected_index_query is now supported with the SUMMARY result format, matching what BASIC already offered. (#11594)

result = batch.validate(
    expectation,
    result_format={
        "result_format": "SUMMARY",
        "unexpected_index_column_names": ["pk"],
        "return_unexpected_index_query": True,
    },
)

Changes

Features

Map expectation validation results now serialize unexpected_rows as dictionaries by default, and the map_expectation_unexpected_rows_as_dict opt-in flag is no longer needed. (#11591)

Bug fixes

Fixed batch.columns() returning an empty list for Redshift batches, which caused column-based expectations to fail; column names are now retrieved via a fallback query when information_schema is inaccessible, and table_name values of the form "schema.table" are parsed correctly. (#11534)

Docs

Documentation no longer labels ExpectAI as beta. (#11590)
Removed temporary notes about row conditions from the documentation. (#11581)
The copy button on documentation code blocks no longer copies hidden lines. (#11571)
Reframed and clarified the documented Data Source limitations. (#11570)

<details> <summary>Maintenance</summary>

return_unexpected_index_query is now supported with the SUMMARY result format, so SUMMARY is no longer more limited than BASIC. (#11594)
Suppressed DeprecationWarnings emitted by dependencies so local test runs are not failed by them. (#11587)
Added an opt-in map_expectation_unexpected_rows_as_dict Checkpoint setting that serializes unexpected_rows as dictionaries for all Map expectations on SQLAlchemy and Spark, with the default output unchanged. (#11583)
invoke deps accepts --pty and --no-pty flags so automated environments can control pseudo-terminal usage. (#11586)
Improved the local type-checking developer experience: fixed type errors, silenced pyparsing deprecation warnings via the compatibility layer, and documented the type-checking workflow so local runs match CI. (#11574)

</details>

Contributors

Thanks to @leodrivera (first contribution).

1.9.311 Dec 2025

[FEATURE] infer primary keys during column_types metric fetch

[FEATURE] infer primary keys during column_types metric fetch (#11554)
[BUGFIX] Prevent FROM DUAL from being added to a properly formatted ORACLE SQL query when using SQLAlchemy (#11538) (thanks @konnor-b)
[DOCS] asset history (#11543)
[DOCS] typos on manage_expectations.md (#11547)
[DOCS] combined compatibility reference updates (#11555)
[DOCS] unexpected_rows (#11553)
[MAINTENANCE] pre-commit autoupdate (#11539)
[MAINTENANCE] Fix build_docs invocation of invoke (#11545)
[MAINTENANCE] Bump ruff to 0.14.8 (#11550)
[MAINTENANCE] restore null columns to rendered_content tables on multi-source expectations (#11548)
[MAINTENANCE] ensure sqlite tests cleanup connections (#11552)
[MAINTENANCE] Bump mypy to 0.19.0 (#11551)
[MAINTENANCE] Update ruff pre-commit and remove TCH001 from tests ruff ignore list (#11557)

Open source →

Additional notes

Dec 11, 2025 2 files

Open source →

Additional notes

Highlights

Primary key information in column type metrics — The table.column_types metric now reports whether each column is part of the table's primary key when using a SQL (SQLAlchemy) execution engine. Single-column, composite, and quoted primary keys are all detected, and columns in tables without a primary key are reported as not primary keys. (#11554)
```
# Each entry in the metric value now includes a `primary_key` flag:
# [
#     {"name": "id", "type": "UUID", "primary_key": True},
#     {"name": "created_at", "type": "TIMESTAMP WITH TIME ZONE", "primary_key": False},
# ]
```
Oracle query assets no longer get an unwanted FROM DUAL clause — Querying Oracle data sources through SQLAlchemy no longer appends a spurious FROM DUAL clause to an already well-formed SQL query, so query assets against Oracle run as written. Verified against Oracle 19c and PostgreSQL 10.16. (#11538)

Changes

Features

The table.column_types metric now includes a primary_key flag for each column when read through a SQL execution engine, covering single-column, composite, and quoted primary keys. (#11554)

Bug fixes

Queries against Oracle data sources are no longer rewritten with an extra FROM DUAL clause when the query is already properly formatted. (#11538)

Docs

Updated the documentation covering unexpected rows. (#11553)
Updated the compatibility reference page with combined reference updates. (#11555)
Fixed typos in the manage expectations documentation page. (#11547)
Added documentation for asset history. (#11543)

<details> <summary>Maintenance</summary>

Aligned the ruff pre-commit version with the pinned requirements version and re-enabled the TC001 lint rule for tests. (#11557)
Upgraded the mypy version used for type checking and updated type annotations across the codebase to match. (#11551)
SQLite test fixtures now close their raw database connections, eliminating ResourceWarnings from unclosed connections during test runs. (#11552)
Rendered unexpected-rows tables for multi-source expectations such as expect_query_results_to_match_comparison again include columns whose values are null, so column headers line up with the data. (#11548)
Bumped the ruff linter to 0.14.8. (#11550)
Fixed the documentation build so that the invoke command is found during the docs build step. (#11545)
Ran pre-commit autoupdate, moving the ruff pre-commit hook from v0.14.3 to v0.14.7. (#11539)

</details>

Contributors

Thanks to @konnor-b (first contribution).

1.9.23 Dec 2025

[DOCS] validations with the Cloud API

[DOCS] validations with the Cloud API (#11400)
[DOCS] remove "read-only" deployment pattern as a misnomer (#11533)
[DOCS] DOC-1180 - Add documentation for Custom Actions in GX Cloud (#11521)
[DOCS] add dependencies to compatibility reference (#11530)
[MAINTENANCE] Warn when snowflake pkey exists in kwargs (#11520)
[MAINTENANCE] pin posthog-docusaurus (#11531)
[MAINTENANCE] Bump node-forge from 1.3.1 to 1.3.2 in /docs/docusaurus (#11536)
[MAINTENANCE] Reenable stale bot for PRs (#11544)
[MAINTENANCE] Bump mdast-util-to-hast from 13.2.0 to 13.2.1 in /docs/docusaurus (#11541)
[MAINTENANCE] Bump express from 4.21.2 to 4.22.1 in /docs/docusaurus (#11540)
[MAINTENANCE] pre-commit autoupdate (#11539)

Open source →

Additional notes

Dec 3, 2025 2 files

Open source →

Additional notes

Highlights

Fluent Snowflake datasource update methods — Snowflake datasources can now be updated or upserted through the fluent API with update_snowflake and add_or_update_snowflake, matching the methods already available for other datasource types. (#11520)
```
context.data_sources.add_or_update_snowflake(
    name="my_snowflake_ds",
    connection_string="snowflake://<user>@<account>/<database>/<schema>?warehouse=<wh>&role=<role>",
)
```
Documentation for running validations with the GX Cloud API — The documentation now explains how to run validations using the GX Cloud API, including what is supported and how results are handled. (#11400)
Documentation for Custom Actions in GX Cloud — New documentation describes how to configure and use Custom Actions in GX Cloud. (#11521)
Dependency compatibility reference expanded — The compatibility reference in the docs now lists supported dependencies, so you can check which versions work with your installation before upgrading. (#11530)

Deprecations

Passing private_key inside a Snowflake datasource's kwargs is deprecated; use the datasource's dedicated private_key connection argument. Removal in 2.0.0. (#11520)

Changes

Docs

Added supported dependencies to the compatibility reference documentation. (#11530)
Added documentation for Custom Actions in GX Cloud. (#11521)
Removed the misleading "read-only" deployment pattern from the deployment documentation. (#11533)
Added documentation on running validations with the GX Cloud API. (#11400)

<details> <summary>Maintenance</summary>

Bumped express from 4.21.2 to 4.22.1 in the documentation site dependencies. (#11540)
Re-enabled the stale bot for pull requests on a nightly schedule, covering all pull requests regardless of labels and excluding issues. (#11544)
Bumped mdast-util-to-hast from 13.2.0 to 13.2.1 in the documentation site dependencies. (#11541)
Bumped node-forge from 1.3.1 to 1.3.2 in the documentation site dependencies, picking up its security fixes. (#11536)
Pinned the posthog-docusaurus version used by the documentation site. (#11531)
Creating, updating, or loading a Snowflake datasource that supplies private_key through kwargs now emits a deprecation warning, and the fluent API gained update_snowflake and add_or_update_snowflake. (#11520)

</details>

1.9.07 Nov 2025Deprecated

[MAINTENANCE] Ignore boto warning about deprecating python 3.9 support

[MINORBUMP] row conditions (#11478)
[MINORBUMP] Drop support for Python 3.9 (#11501)
[DOCS] row conditions screenshots (#11509)
[DOCS] Cloud API version for row conditions (#11512)
[DOCS] minimum python version now 3.10 (#11485)
[MAINTENANCE] Fix flaky validation definition test (#11495)
[MAINTENANCE] Make only parameter to conditions Column positional (#11497)
[MAINTENANCE] Add snowflake private key to ci env variables (#11499)
[MAINTENANCE] Narrow iterable member types for ComparisonCondition with in/not in operators (#11494)
[MAINTENANCE] Don't allow bool values in Column.is_in( ) / Column.is_not_in() (#11500)
[MAINTENANCE] Move Snowflake tests to key-pair auth (#11498)
[MAINTENANCE] Ignore boto warning about deprecating python 3.9 support (#11505)
[MAINTENANCE] Move row conditions Column import (#11506)
[MAINTENANCE] Update root validator to always map to a Condition (#11504)
[MAINTENANCE] Render multiple conditions in data docs (#11507)
[MAINTENANCE] Add posthog-docusaurus plugin to enable pageview capture (#11508)
[MAINTENANCE] Add Condition subclasses to public API (#11511)
[MAINTENANCE] ci: add ld-dev-server (#11510)
[MAINTENANCE] Delete CommonRoom script from docs - load Posthog directly (#11514)
[MAINTENANCE] Update exclude list (#11517)
[MAINTENANCE] Deprecate string-style row_conditions (#11515)
[CONTRIB] Include unexpected_index_column_names in ExpectColumnValuesToNotBeNull results (#11513) (thanks @chay0112)

Open source →

Additional notes

Nov 7, 2025 2 files

Open source →

Additional notes

Compatibility: Python <3.14,>=3.9 → <3.14,>=3.10; numpy removed (python_version == "3.9"); pandas removed (python_version == "3.9")

Highlights

Row conditions: documented, importable, and rendered in Data Docs — Row conditions are now a supported way to scope an Expectation to a subset of rows. The condition classes (including Column and the comparison, nullity, and boolean conditions) are part of the public API and are imported from great_expectations.expectations.row_conditions; the previous great_expectations.expectations.conditions import path still works. Conditions are validated more strictly (an in/not-in parameter must be an iterable whose members share a single type, and boolean members are rejected), a bare string condition is always turned into a condition object even when no condition parser is given, and Data Docs now renders every condition when an Expectation carries more than one. New documentation pages, screenshots, and notes on the minimum GX Cloud API and agent versions required for certain row-condition features round this out. (#11478, #11494, #11497, #11500, #11504, #11506, #11507, #11509, #11511, #11512)
```
from great_expectations.expectations.row_conditions import Column

condition = Column("age") > 21
```
Python 3.10 is now the minimum supported version — Great Expectations no longer supports Python 3.9. Install on Python 3.10 or newer; the documentation now states 3.10 as the minimum. (#11501, #11485)
unexpected_index_column_names returned by ExpectColumnValuesToNotBeNull — When you request unexpected index columns in result_format, ExpectColumnValuesToNotBeNull now includes unexpected_index_column_names in its validation result, matching the other column-value Expectations. (#11513)
```
result_format={"result_format": "COMPLETE", "unexpected_index_column_names": ["customer_id"]}
```

Changes

Bug fixes

ExpectColumnValuesToNotBeNull now includes unexpected_index_column_names in its validation results when they are requested through result_format. (#11513)

Docs

Documented the minimum GX Cloud API and agent versions required to use certain row-condition capabilities. (#11512)
Updated the documented minimum supported Python version to 3.10. (#11485)
Added screenshots to the row conditions documentation. (#11509)
Added documentation and code samples for using row conditions to scope Expectations to a subset of rows. (#11478)

<details> <summary>Maintenance</summary>

Removed the discontinued Common Room script from the documentation site and captured documentation page views directly with PostHog. (#11514)
Dropped support for Python 3.9; Great Expectations now requires Python 3.10 or newer. (#11501)
Continuous integration now runs against a mock LaunchDarkly server instead of a live feature-flag service. (#11510)
The row condition subclasses are now part of the documented public API. (#11511)
Restored page-view capture on the documentation site so single-page navigation is tracked again. (#11508)
Data Docs now renders every condition when an Expectation is scoped by more than one row condition. (#11507)
A row condition supplied as a string is now always converted into a condition object, including when no condition parser is specified. (#11504)
The row conditions Column class and its siblings are now imported from great_expectations.expectations.row_conditions; the previous great_expectations.expectations.conditions import path continues to work. (#11506)
Suppressed the boto warning about the deprecation of Python 3.9 support. (#11505)
Snowflake tests now authenticate with key-pair authentication. (#11498)
Boolean values are no longer accepted as members of the parameter passed to Column.is_in() and Column.is_not_in(). (#11500)
Comparison conditions using the in/not-in operators now require an iterable parameter whose members are all of the same type (or all numeric), and report an error otherwise. (#11494)
Added the Snowflake private key to the continuous integration environment variables. (#11499)
The conditions Column class now takes its column name as a positional argument, so it can be constructed as Column("age"). (#11497)
Fixed a flaky validation definition test caused by a race between raised errors when reusing a context from other tests. (#11495)

</details>

Contributors

Thanks to @chay0112 (first contribution).

1.8.130 Oct 2025Deprecated

[DOCS] snowflake password deprecation

[BUGFIX] Fix render_content generation with new row_condition types (#11481)
[BUGFIX] Prevent unmocked HTTP requests in cloud tests (#11492)
[BUGFIX] Update pytest filter to handle google.api_core Python 3.10 end-of-life warning (#11493)
[DOCS] snowflake password deprecation (#11416)
[DOCS] correct Snowflake private key guidance (#11490)
[MAINTENANCE] Stop running python 3.12 marker tests on all PRs (#11457)
[MAINTENANCE] Disable link checker (#11477)
[MAINTENANCE] Transform legacy row_condition string into new Condition object (#11474)
[MAINTENANCE] Keep condition_parser field intact for backwards compatibility (#11484)
[MAINTENANCE] Replace broken bitnami image with apache image for spark (#11444)
[MAINTENANCE] Ignore future warning Python 3.9 EOL (#11486)
[MAINTENANCE] Add passthrough path for pandas and spark Condition parser (#11480)
[MAINTENANCE] Constrain row_condition groups (#11488)
[MAINTENANCE] Only install mssql drivers when needed (#11489)
[MAINTENANCE] Use Suites v2 REST endpoints (#11487)
[MAINTENANCE] Handle None parameter in Conditions (#11491)

Open source →

Additional notes

Oct 30, 2025 2 files

Open source →

Additional notes

Highlights

Null checks in row conditions — Row conditions now express null comparisons explicitly with is_null() and is_not_null() on a column, and passing None as the value of a comparison operator is rejected instead of silently producing an invalid condition. (#11491)
```
import great_expectations.expectations as gxe
from great_expectations.core.expectation_condition import Column

gxe.ExpectColumnValuesToBeBetween(
    column="amount",
    min_value=0,
    row_condition=Column("cancelled_at").is_null(),
)
```
Legacy row condition strings keep working alongside condition objects — Existing string-based row_condition values are accepted and converted into the new condition objects, the condition_parser field is preserved, pandas and Spark conditions have a passthrough path, and rendered expectation content displays the new condition types correctly. (#11474, #11484, #11480, #11481)
Clearer limits on combining row conditions — Nested AndConditions are flattened automatically, while OrConditions nested inside AndConditions or other OrConditions now raise an explicit error, as does supplying more than 100 conditions. (#11488)
Updated Snowflake connection documentation — The Snowflake documentation now covers the deprecation of password authentication and gives corrected guidance for configuring private key authentication. (#11416, #11490)

Changes

Bug fixes

Test runs no longer fail on the google.api_core Python 3.10 end-of-life warning. (#11493)
Cloud-marked tests can no longer reach the live API: unmocked HTTP requests are blocked, preventing accidental production calls during test runs. (#11492)
Rendered expectation content is generated correctly for the new row condition types. (#11481)

Docs

Corrected the Snowflake private key authentication guidance. (#11490)
Documented the deprecation of Snowflake password authentication and the recommended alternatives. (#11416)

<details> <summary>Maintenance</summary>

Passing None as the value of a column comparison in a row condition is now rejected; use is_null() or is_not_null() for null checks, and the comparison parameter is required. (#11491)
Expectation suites stored in GX Cloud are now read and written through the v2 REST endpoints. (#11487)
Microsoft SQL Server drivers are installed in CI only for the jobs that need them, and the installation script reports failures more clearly. (#11489)
Row condition groups are constrained: nested AndConditions are flattened, OrConditions nested inside AndConditions or OrConditions raise an error, and more than 100 conditions raises an error. (#11488)
Added a passthrough path for the pandas and Spark row condition parsers so existing condition expressions are handled directly. (#11480)
Test runs no longer surface the Python 3.9 end-of-life future warning. (#11486)
Spark test environments now use the Apache-published Spark image after the previously used Bitnami image was removed. (#11444)
The condition_parser field is retained for backwards compatibility so single-condition row conditions still convert to string syntax on older API versions. (#11484)
Legacy string row_condition values are transformed into the new condition objects. (#11474)
Disabled the documentation link checker, which was reporting too many false positives. (#11477)
Python 3.12 marker tests no longer run on every pull request; only the minimum and maximum supported Python versions are exercised for those events. (#11457)

</details>

1.8.023 Oct 2025

[FEATURE] Snowflake Key Pair Auth API

[FEATURE] Snowflake Key Pair Auth API (#11395)
[BUGFIX] Row condition for Volume Expectations (#11467)
[DOCS] Data Health - Failed Expectations (#11419)
[DOCS] Data Health failed Expectations screenshot updates (#11458)
[DOCS] GX Cloud Architecture Supporting Content (#11439)
[DOCS] restore link checker (#11449)
[DOCS] schema in connection string for redshift and postgreSQL (#11433)
[MAINTENANCE] Row condition comparison classes (#11450)
[MAINTENANCE] Row condition execution engine scaffolding (#11452)
[MAINTENANCE] Run CI datasource cleanup hourly and clean up more schemas (#11462)
[MAINTENANCE] Filter clauses for pandas conditions (#11455)
[MAINTENANCE] Filter clauses for spark conditions (#11456)
[MAINTENANCE] Ignore pkg_resources warning (#11465)
[MAINTENANCE] Add Condition type to row condition instances (#11464)
[MAINTENANCE] Clean up datasource schemas that are over an hour old (#11463)
[MAINTENANCE] Filter clauses for SQLA conditions (#11459)
[MAINTENANCE] Condition class handling for pandas (#11469)
[MAINTENANCE] Bump lycheeverse/lychee-action from 2.0.1 to 2.0.2 in /.github/workflows (#11466)
[MAINTENANCE] Condition class handling for spark (#11470)
[MAINTENANCE] Condition class handling for SQL (#11473)

Open source →

Additional notes

Oct 23, 2025 2 files

Open source →

Additional notes

Highlights

Snowflake key pair authentication — Snowflake data sources now accept key pair authentication as a first-class part of the connection API, so you can configure a Snowflake data source with a private key instead of a password. (#11395)

context.data_sources.add_snowflake(
    name="my_snowflake",
    connection_details={
        "account": "myOrg-my_account",
        "user": "my_user",
        "database": "my_db",
        "schema": "my_schema",
        "warehouse": "my_wh",
        "role": "my_role",
        "private_key": "<PEM-encoded private key>",
    },
)

Row conditions are honored by Volume Expectations — Volume Expectations now apply the configured row condition, so expected row counts are evaluated against the filtered rows rather than the whole batch. (#11467)
Documented schema handling in Redshift and PostgreSQL connection strings — The connection documentation now explains how to specify a schema in Redshift and PostgreSQL connection strings. (#11433)
GX Cloud Data Health documentation for failed Expectations — New GX Cloud documentation covers the Data Health view for failed Expectations, with screenshots refreshed to match the current interface, alongside new GX Cloud architecture supporting content. (#11419, #11458, #11439)

Changes

Features

Snowflake data sources can now be configured with key pair authentication as a supported set of connection details. (#11395)

Bug fixes

Volume Expectations now respect the configured row condition when counting rows. (#11467)

Docs

Documentation now describes how to include a schema in Redshift and PostgreSQL connection strings. (#11433)
Re-enabled the documentation link checker. (#11449)
Added supporting content to the GX Cloud architecture documentation. (#11439)
Updated the Data Health failed Expectations screenshots to match recent interface changes. (#11458)
Added GX Cloud documentation for the Data Health view of failed Expectations. (#11419)

<details> <summary>Maintenance</summary>

SQL execution now handles the structured row-condition type internally, with no change to how existing string conditions behave. (#11473)
Spark execution now handles the structured row-condition type internally, with no change to how existing string conditions behave. (#11470)
Updated the lychee link-checking GitHub Action used in CI from 2.0.1 to 2.0.2. (#11466)
The pandas execution engine now handles the structured row-condition type for row conditions, with no change to existing behavior. (#11469)
Added internal filter-clause support for SQLAlchemy row conditions. (#11459)
CI cleanup of test data source schemas now removes schemas older than one hour instead of two. (#11463)
Expectation row conditions now accept the new structured condition type in their schemas, though passing a condition object currently raises an error. (#11464)
Suppressed a pkg_resources deprecation warning that was causing spurious test failures. (#11465)
Added internal filter-clause support for Spark row conditions. (#11456)
Added internal filter-clause support for pandas row conditions. (#11455)
CI now cleans up test data sources hourly and covers more leftover schemas. (#11462)
Added internal execution-engine scaffolding for structured row conditions. (#11452)
Added comparison, nullity, and column classes used to build row conditions. (#11450)
The SQLite connection used to register helper functions is now closed, removing a resource warning seen in test runs. (#11451)
Added AND/OR classes for combining multiple row conditions. (#11448)

</details>

1.7.09 Oct 2025

[MINORBUMP] Remove Pandas Upper Bound Constraint

[MINORBUMP] Remove Pandas Upper Bound Constraint (#11423)
[MINORBUMP] Remove Renderer class from public API (#10866)
[MINORBUMP] Add support for Python 3.13 (#11426)
[BUGFIX] ensure snowflake conn str is always transformed to rich type (#11410)
[BUGFIX] Validation authentication for PandasS3Datasource using boto3_options (#11412)
[DOCS] make schedules opt-in (#11408)
[DOCS] Completeness Anomaly Detection is opt-in (#11406)
[DOCS] remove migration guide (#11405)
[MAINTENANCE] Bump SQLAlchemy version on doc snippet tests (#11411)
[MAINTENANCE] Fix test assertions for Numpy 2 compatibility (#11415)
[MAINTENANCE] Pin pact-python due to install error on 3.12 of 3.0. (#11418)
[MAINTENANCE] Skip tests for SQLA < 2 and Pandas >= 2.2 (#11417)
[MAINTENANCE] Upgrade ruff (#11421)
[MAINTENANCE] Upgrade mypy (#11422)
[MAINTENANCE] Remove analytics (#11420)
[MAINTENANCE] Emit a warning when workspace id is unset. (#11425)
[MAINTENANCE] Reenable publishing pact tests (#11427)
[MAINTENANCE] Add schema to Redshift ConnectionDetails (#11431)

Open source →

Additional notes

Oct 9, 2025 2 files

Open source →

Additional notes

Compatibility: Python <3.13,>=3.9 → <3.14,>=3.9; numpy added (python_version >= "3.13"); pandas added (python_version >= "3.13"); posthog removed; pandas removed (extra snowflake) (python_version >= "3.9")

Highlights

Python 3.13 support — Great Expectations now installs and runs on Python 3.13, in addition to the previously supported 3.9 through 3.12. (#11426)
Works with pandas 2.2 and newer — The <2.2 upper bound on pandas has been removed, so you can install Great Expectations alongside pandas 2.2.0 and later and pick up the newest pandas features and fixes. (#11423)
```
pip install great_expectations "pandas>=2.2"
```
Usage analytics removed — Great Expectations no longer collects or sends usage analytics, and the posthog dependency is no longer installed with the library. (#11420)
Reassigning a Snowflake connection string now works as expected — Setting a new connection string on an existing SQL data source — including Snowflake — is now converted to the proper connection type, so the data source stays usable after the reassignment. (#11410)
```
datasource.connection_string = "snowflake://user:password@account/db/schema?warehouse=wh&role=role"
```

Changes

Features

Added support for running Great Expectations on Python 3.13. (#11426)
The Renderer class is no longer part of the public API. (#10866)
Removed the <2.2 upper bound on pandas so Great Expectations can be used with pandas 2.2.0 and above. (#11423)

Bug fixes

Fixed AWS authentication errors at validation time when credentials were supplied to PandasS3Datasource through boto3_options rather than environment variables. (#11412)
Fixed an issue where assigning a new connection string to a SQL data source after creation — most visibly with Snowflake — left the value in an unusable form. (#11410)

Docs

Removed the migration guide from the documentation. (#11405)
Documentation now states that Completeness Anomaly Detection is opt-in. (#11406)
Documentation now states that schedules are opt-in. (#11408)

<details> <summary>Maintenance</summary>

Redshift connection details now accept a discrete schema field, so a schema can be supplied separately when configuring a Redshift connection. (#11431)
Re-enabled publishing of pact contract tests. (#11427)
A warning is now emitted when the workspace ID is not set. (#11425)
Removed usage analytics collection from the library, along with its posthog dependency. (#11420)
Upgraded the mypy version used for type checking. (#11422)
Upgraded the ruff version used for linting and formatting. (#11421)
Skipped tests that fail on the combination of SQLAlchemy below 2.0 and pandas 2.2 or newer to keep CI stable. (#11417)
Pinned pact-python to avoid an installation error on Python 3.12. (#11418)
Updated test assertions to use truthiness checks instead of identity comparisons for NumPy 2.x compatibility. (#11415)
Bumped the SQLAlchemy version used when testing documentation snippets. (#11411)

</details>

1.6.012 Sept 2025

[FEATURE] Make GX Context workspace aware

[FEATURE] Make GX Context workspace aware (#11369)
[BUGFIX] Fix handling of mixed case table names when computing column metadata (#11365)
[BUGFIX] Make handling of quoted schema and table names more robust (#11367)
[BUGFIX] Enable reading multiple pages of S3 directory by popping ContinuationToken from query_options once it is used in a S3 request (#11361) (thanks @pawel99k)
[DOCS] Update docs to reflect that custom docker images should use stable agent (#11353)
[MAINTENANCE] Test unexpected_rows (#11368)
[MAINTENANCE] Plumb in GX_CLOUD_WORKSPACE_ID into cloud-test ci step. (#11373)
[MAINTENANCE] Add workspace_id to store_backend dict (#11371)

Open source →

Additional notes

Sep 12, 2025 2 files

Open source →

Additional notes

Highlights

GX Cloud workspace awareness — Data Contexts are now workspace aware, laying the groundwork for GX Cloud's multi-workspace support. A workspace id supplied to a Cloud context is carried through to Cloud requests and to the credentials used by its stores. (#11369, #11371, #11373)
```
# GX_CLOUD_WORKSPACE_ID is read alongside your Cloud access token and organization id
import great_expectations as gx

context = gx.get_context(mode="cloud")
```
S3 data assets read past the first page of results — Listing files in an S3 directory or bucket with more results than fit in a single response now returns all of them instead of raising an error part-way through. (#11361)
More robust handling of quoted and mixed-case SQL identifiers — Schema and table names that are quoted, use data-source-specific quote characters, or use mixed case are now handled correctly when building queries and when collecting column metadata. (#11367, #11365)

Changes

Features

Data Contexts are now workspace aware, adding initial support for GX Cloud's upcoming multi-workspace feature. (#11369)

Bug fixes

Reading an S3 directory that spans multiple pages of results no longer fails; the continuation token is no longer reused in subsequent requests. (#11361)
Quoted schema and table names are handled more reliably: the quote characters used by all supported SQL data sources are now recognized, and identifiers are quoted correctly when serialized. (#11367)
Column metadata is now computed correctly for tables whose names use mixed case. (#11365)

Docs

Documentation for building custom agent Docker images now recommends the stable base image instead of latest. (#11353)

<details> <summary>Maintenance</summary>

A Cloud context now passes its workspace id along with the access token and organization id in the credentials used by its stores, and requests for the data context configuration are scoped to the workspace. (#11371)
The Cloud test job in continuous integration now supplies a workspace id, so workspace-aware behavior is exercised in CI. (#11373)
Added test coverage confirming that the unexpected_rows result format option behaves consistently across the canonical column map expectations for Pandas and SQL data sources. (#11368)

</details>

Contributors

Thanks to @pawel99k (first contribution).

1.5.218 Jun 2025

[FEATURE] Add SuiteParameterDict to all Expectation Kwargs (#11222) (thanks @Pascal06S)

[FEATURE] Add SuiteParameterDict to all Expectation Kwargs (#11222) (thanks @Pascal06S)
[BUGFIX] fix ValidationError in ExpectColumnPairValuesToHaveDifferenceOfCustomPercentage (#11209) (thanks @sariaslaso)
[BUGFIX] add_batch_definition_whole_directory only loading 1 file (#11254)
[BUGFIX] Remove datetime types from proportion min_value and max_value parameters (#11259)
[DOCS] temporarily remove link checker (#11250)
[DOCS] Cloud API data source tip (#11248)
[DOCS] ExpectColumnProportionOfNonNullValuesToBeBetween (#11245) (#11257)
[MAINTENANCE] Update "fraction" to "proportion" on NonNullProportion renderer (#11253)
[MAINTENANCE] Bump brace-expansion from 1.1.11 to 1.1.12 in /docs/docusaurus (#11251)

Open source →

Additional notes

Jun 18, 2025 2 files

Open source →

Additional notes

Highlights

Suite parameters accepted in every expectation argument — All expectation arguments now accept suite parameters, so any keyword argument of an expectation can be supplied at validation time instead of being hard-coded when the expectation is defined. (#11222)

import great_expectations as gx

expectation = gx.expectations.ExpectColumnValuesToBeBetween(
    column="passenger_count",
    min_value={"$PARAMETER": "min_passengers"},
    max_value={"$PARAMETER": "max_passengers"},
)

results = batch.validate(
    expectation,
    expectation_parameters={"min_passengers": 1, "max_passengers": 6},
)

Whole-directory batch definitions read every file again — Batch definitions created with add_batch_definition_whole_directory on S3, Azure Blob Storage, and Google Cloud Storage data assets now read all files in the directory instead of only one. (#11254)
```
asset = data_source.add_directory_csv_asset(name="my_asset", s3_prefix="data/")
batch_definition = asset.add_batch_definition_whole_directory("all_files")
batch = batch_definition.get_batch()
```

Changes

Features

Every expectation argument now accepts a suite parameter, so all expectation keyword arguments can be parameterized and supplied at validation time. (#11222)

Bug fixes

The min_value and max_value parameters of ExpectColumnProportionOfNonNullValuesToBeBetween and ExpectColumnProportionOfUniqueValuesToBeBetween no longer accept date or datetime values; they now accept only numbers (or a suite parameter), and their published schemas reflect this. (#11259)
Fixed add_batch_definition_whole_directory reading only a single file for S3, Azure Blob Storage, and Google Cloud Storage data assets; the whole directory is now read as one batch. (#11254)
Fixed a ValidationError raised when using ExpectColumnPairValuesToHaveDifferenceOfCustomPercentage; the expectation now declares its required percentage argument. (#11209)

Docs

Added a tip about Cloud API data sources to the documentation. (#11248)
Temporarily disabled the documentation link checker while changed page paths are sorted out. (#11250)

<details> <summary>Maintenance</summary>

Bumped the brace-expansion documentation-site dependency from 1.1.11 to 1.1.12. (#11251)
Rendered output for the non-null proportion expectation now says "proportion" instead of "fraction". (#11253)

</details>

Contributors

Thanks to @Pascal06S (first contribution), @sariaslaso (first contribution).

1.5.05 Jun 2025

[MINORBUMP] docs for Multi-source Expectations

[MINORBUMP] docs for Multi-source Expectations (#11165)
[DOCS] revise use of the term "Cloud" in the Core docs (#11207)
[DOCS] restore link checker (#11212)
[MAINTENANCE] add warning filter (#11217)
[MAINTENANCE] Update renderer labels (#11216)
[MAINTENANCE] Improve ExpectQueryResultsToMatchComparison docstring (#11221)
[MAINTENANCE] pre-commit autoupdate (#11218)
[MAINTENANCE] Remove pkg_resources dependency (#11213)
[MAINTENANCE] Example suite parameter tests (#11225)
[CONTRIB] Add type geometry and super into redshift_execution_engine.py (#11194) (thanks @VolkovGeoPhy)

Open source →

Additional notes

Jun 5, 2025 2 files

Yanked

Open source →

Additional notes

Highlights

Multi-source Expectations documentation — The documentation now covers Multi-source Expectations, explaining how to compare data across two different data sources. (#11165)
Redshift geometry and super column types supported — Redshift data sources now recognize the GEOMETRY and SUPER column types, so assets containing these columns can be introspected and validated. (#11194)
No more pkg_resources dependency — GX Core no longer depends on the deprecated pkg_resources package, removing its import-time deprecation warnings on modern Python installs. (#11213)

Changes

Features

Added support for the Redshift GEOMETRY and SUPER column types. (#11194)

Docs

Restored the documentation link checker that had been temporarily disabled. (#11212)
Added documentation for Multi-source Expectations. (#11165)
Revised how the term "Cloud" is used throughout the GX Core documentation for consistency. (#11207)

<details> <summary>Maintenance</summary>

Added test coverage for suite parameters used as min_value and max_value in ExpectColumnMaxToBeBetween. (#11225)
Removed the pkg_resources dependency, replacing requirements parsing with a pip compatibility module and a self-contained parser in setup.py. (#11213)
Reverted the session-scoped SQL engine pool in the expectation test suite because it broke test schema cleanup. (#11224)
Updated pre-commit hooks, bumping ruff-pre-commit to v0.11.12. (#11218)
Added a session-scoped SQL engine pool to the expectation tests (subsequently reverted in this release). (#11219)
Improved the ExpectQueryResultsToMatchComparison docstring so parameter names and descriptions read consistently. (#11221)
Updated the diagnostic renderer labels shown for ExpectQueryResultsToMatchComparison. (#11216)
Added a warning filter to the test configuration to quiet expected warnings. (#11217)

</details>

Contributors

Thanks to @VolkovGeoPhy.

1.4.628 May 2025Breaking

[MAINTENANCE] limit pyspark to <4.0 due to breaking changes in types

[BUGFIX] Handle unhashable data types for ExpectQueryResultsToMatchComparison (#11193)
[BUGFIX] Make type expectation case-insensitive for Databricks, Snowflake an Postgres (#11192)
[BUGFIX] Handle set rendering cases for ExpectQueryResultsToMatchComparison (#11203)
[BUGFIX] Handle UUID in rendered content (#11204)
[DOCS] Anomaly Detection (#11172)
[DOCS] Refine use of the term "platform" (#11187)
[DOCS] Update docs to reflect that views are supported (#11198)
[DOCS] Revise use of the term "Core" in relation to Cloud (#11205)
[MAINTENANCE] Add CloudAMQP connection string env variable to ci (#11190)
[MAINTENANCE] Show gx cloud logs on cloud test failures. (#11188)
[MAINTENANCE] limit pyspark to <4.0 due to breaking changes in types (#11197)
[MAINTENANCE] Ignore warnings around pkg_resources (#11201)
[MAINTENANCE] Update descriptions for multi-source parameter (#11202)
[MAINTENANCE] Snowflake tests only run with snowflake flag (#10605)

Open source →

Additional notes

May 28, 2025 2 files

Open source →

Additional notes

Highlights

Clearer errors for unhashable column types in ExpectQueryResultsToMatchComparison — When a query returns unhashable data types such as JSONB, ExpectQueryResultsToMatchComparison now raises a helpful error that names the first column containing unhashable data instead of failing with an unclear message. (#11193)
Case-insensitive column type checks on Databricks, Snowflake, and Postgres — expect_column_values_to_be_of_type now treats unquoted identifiers in column_name and column_type as case-insensitive on Databricks, Postgres, and Snowflake, so type expectations pass regardless of the casing you write. (#11192)
```
suite.add_expectation(
    gxe.ExpectColumnValuesToBeOfType(column="my_column", type_="varchar")
)
```
Documentation for anomaly detection — The documentation now covers anomaly detection, alongside refreshed wording for the terms "Core" and "platform" and updated guidance noting that both tables and views are supported as data assets. (#11172, #11187, #11198, #11205)

Changes

Bug fixes

UUID values are now handled correctly when rendering expectation content. (#11204)
Rendering of ExpectQueryResultsToMatchComparison now handles cases where results contain sets. (#11203)
expect_column_values_to_be_of_type now treats unquoted column names and column types as case-insensitive on Databricks, Postgres, and Snowflake. (#11192)
ExpectQueryResultsToMatchComparison now raises a clear error naming the first column with unhashable data (for example JSONB) instead of failing unhelpfully. (#11193)

Docs

Revised how the term "Core" is used in relation to GX Cloud throughout the documentation. (#11205)
Data Source connection and Expectation docs now state that views are supported in addition to tables. (#11198)
Refined how the term "platform" is used across the documentation. (#11187)
Added documentation covering anomaly detection. (#11172)

<details> <summary>Maintenance</summary>

Snowflake tests now run only when the Snowflake flag is enabled, so local test runs no longer hit external backends by default. (#10605)
Updated the parameter descriptions for the multi-source comparison parameter. (#11202)
Suppressed a newly surfaced pkg_resources deprecation warning. (#11201)
Restricted the supported pyspark range to >=2.3.2,<4.0, since pyspark 4.0 introduces incompatible type changes. (#11197)
GX Cloud logs are now surfaced when cloud tests fail. (#11188)
Added a CloudAMQP connection string environment variable to the CI configuration. (#11190)

</details>

1.4.414 May 2025Deprecated

[MAINTENANCE] Resolve datetime deprecation warnings (#11134) (thanks @emmanuel-ferdman)

[FEATURE] Compute missing/unexpected rows for ExpectQueryResultsToMatchSource (#11161)
[FEATURE] ExpectQueryResultsToMatchSource table diagnostic renderer (#11168)
[FEATURE] ExpectQueryResultsToMatchSource renderer for single-column … (#11173)
[DOCS] remove link checker (#11162)
[MAINTENANCE] Fix dying mssql compatibility tests. (#11153)
[MAINTENANCE] Bump prismjs from 1.29.0 to 1.30.0 in /docs/docusaurus (#11126)
[MAINTENANCE] Breakout redshift tests into own job. (#11167)
[MAINTENANCE] Bring back redshift creds for docs tests (#11169)
[MAINTENANCE] Improve ExpectQueryResultsToMatchSource docstring (#11158)
[MAINTENANCE] Nightly redshift cleanup (#11166)
[MAINTENANCE] Bump image-size from 1.2.0 to 1.2.1 in /docs/docusaurus (#11125)
[MAINTENANCE] Add prescriptive atomic renderer to ExpectQueryResultsToMatchSource (#11160)
[MAINTENANCE] Resolve datetime deprecation warnings (#11134) (thanks @emmanuel-ferdman)
[MAINTENANCE] Bump http-proxy-middleware from 2.0.7 to 2.0.9 in /docs/docusaurus (#11120)
[MAINTENANCE] Bump @babel/runtime from 7.26.9 to 7.27.0 in /docs/docusaurus (#11124)
[MAINTENANCE] Bump estree-util-value-to-estree from 3.3.2 to 3.3.3 in /docs/docusaurus (#11121)
[MAINTENANCE] ensure latest docker compose (#11156)
[MAINTENANCE] Fix flakey test and simplify another assertion (#11163)
[MAINTENANCE] Update ExpectQueryResultsToMatchSource DQI (#11164)
[MAINTENANCE] Bump @babel/helpers from 7.26.9 to 7.27.0 in /docs/docusaurus (#11123)
[MAINTENANCE] Change DQ issue for ExpectQueryResultsToMatchSource (#11174)
[CONTRIB] Update ruff badge (#10905) (thanks @esadek)

Open source →

Additional notes

May 14, 2025 2 files

Open source →

Additional notes

Highlights

New Expectation: ExpectQueryResultsToMatchSource — You can now compare the results of a SQL query run against your Data Source with the results of a query run against another Data Source, and require that at least a mostly fraction of records match. Supported on PostgreSQL, Snowflake, Databricks (SQL), Redshift, and SQLite. (#11144)
```
import great_expectations as gx

expectation = gx.expectations.ExpectQueryResultsToMatchSource(
    target_query="SELECT id, amount FROM orders",
    source_data_source_name="my_source_data_source",
    source_query="SELECT id, amount FROM orders",
    mostly=0.95,
)
```
Richer results and rendering for ExpectQueryResultsToMatchSource — Validation results for ExpectQueryResultsToMatchSource now report the specific rows missing from or unexpected in the target query results, and those differences are presented as a diagnostic table — with a simplified presentation when the source and target queries each return a single column. The Expectation also renders a readable summary showing the target query and the source Data Source it is compared against. (#11161, #11168, #11173, #11160)

Changes

Features

ExpectQueryResultsToMatchSource results now use a simplified diagnostic rendering when the target and source queries each return a single column. (#11173)
ExpectQueryResultsToMatchSource validation results are now displayed as a diagnostic table of differences between the target and source query results. (#11168)
ExpectQueryResultsToMatchSource now computes and reports the rows missing from and unexpected in the target query results. (#11161)
Added the ExpectQueryResultsToMatchSource Expectation, which compares the results of a SQL query against the results of a query on another Data Source and passes when at least a mostly fraction of records match. (#11144)

Docs

Updated the Ruff badge in the README to point at the current astral-sh/ruff endpoint. (#10905)
Temporarily removed the documentation link checker as a workaround for a known issue. (#11162)
Updated the ExpectAI documentation to reflect current email alert behavior. (#11154)

<details> <summary>Maintenance</summary>

Updated the data quality issue category reported for ExpectQueryResultsToMatchSource. (#11174)
Bumped @babel/helpers from 7.26.9 to 7.27.0 in the documentation site. (#11123)
Updated the data quality issue metadata for ExpectQueryResultsToMatchSource. (#11164)
Fixed a flaky test caused by floating-point comparison and simplified a related assertion to use pytest.approx. (#11163)
CI now ensures a recent Docker Compose version to avoid a race condition when pulling many images concurrently. (#11156)
Bumped @babel/runtime from 7.26.9 to 7.27.0 in the documentation site. (#11124)
Bumped estree-util-value-to-estree from 3.3.2 to 3.3.3 in the documentation site. (#11121)
Bumped http-proxy-middleware from 2.0.7 to 2.0.9 in the documentation site. (#11120)
ExpectQueryResultsToMatchSource now renders a prescriptive summary showing the target SQL query and the source Data Source it is compared against. (#11160)
Removed deprecated datetime.utcnow() and utcfromtimestamp() usage so no datetime deprecation warnings are emitted on Python 3.12+. (#11134)
Bumped image-size from 1.2.0 to 1.2.1 in the documentation site, picking up a denial-of-service fix. (#11125)
Cleaned up the nightly Redshift test setup. (#11166)
Clarified the ExpectQueryResultsToMatchSource documentation to note that column names do not matter but column order does. (#11158)
Restored Redshift credentials needed by the documentation tests in CI. (#11169)
Split the Redshift tests into their own CI job. (#11167)
Bumped prismjs from 1.29.0 to 1.30.0 in the documentation site. (#11126)
Repaired the failing MSSQL compatibility test runs. (#11153)
Bumped @babel/runtime-corejs3 from 7.26.9 to 7.27.0 in the documentation site. (#11122)

</details>

Contributors

Thanks to @esadek (first contribution), @emmanuel-ferdman (first contribution).

1.4.122 Apr 2025

[FEATURE] Allow user to provide connection details to connect to Redshift

[FEATURE] Allow user to provide connection details to connect to Redshift (#11105)
[FEATURE] Add ColumnDescriptiveStats metric (#11108)
[BUGFIX] MicrosoftTeamsNotificationAction failing with 400 Bad Request (#11106) (thanks @jwalant-dattani)
[BUGFIX] expect_column_values_to_be_between allows both min/max values to be None or empty strings (#11102)
[BUGFIX] redshift: prevent runtime TypeError (#11112)
[MAINTENANCE] Cleanup metrics package (#11109)

Open source →

Additional notes

Apr 22, 2025 2 files

Open source →

Additional notes

Highlights

New ColumnDescriptiveStats metric — You can now compute a column's minimum, maximum, mean, and standard deviation in a single metric with ColumnDescriptiveStats, available on the pandas, SQL, and Spark backends. (#11108, #11109)

from great_expectations.metrics import ColumnDescriptiveStats

result = batch.compute_metrics(ColumnDescriptiveStats(column="passenger_count"))
print(result.value.min, result.value.max, result.value.mean, result.value.standard_deviation)

New ColumnValuesNotMatchRegexCount metric — You can now count the values in a column that do not match a regular expression with ColumnValuesNotMatchRegexCount, available on the pandas, SQL, and Spark backends. (#11103)
```
from great_expectations.metrics import ColumnValuesNotMatchRegexCount

result = batch.compute_metrics(
    ColumnValuesNotMatchRegexCount(column="vendor_id", regex="^(a|d).+")
)
print(result.value)
```
Connect to Redshift with connection details — A Redshift data source can now be configured by supplying individual connection details instead of a full connection string. (#11105)
Redshift schema introspection no longer raises a TypeError — Using the gx-redshift extra to introspect schema information, such as computing column descriptive metrics, no longer fails with a runtime TypeError. (#11112)

Changes

Features

Added the ColumnDescriptiveStats metric, which returns a column's minimum, maximum, mean, and standard deviation on pandas, SQL, and Spark backends. (#11108)
Redshift data sources can now be configured with individual connection details in addition to a connection_string. (#11105)
Added the ColumnValuesNotMatchRegexCount metric, which counts column values that do not match a given regular expression on pandas, SQL, and Spark backends. (#11103)

Bug fixes

Fixed a runtime TypeError when using the gx-redshift extra to perform schema introspection, such as computing column descriptive metrics. (#11112)
ExpectColumnValuesToBeBetween now correctly rejects configurations where both min_value and max_value are omitted, None, or empty strings. (#11102)
Fixed MicrosoftTeamsNotificationAction failing with a 400 Bad Request when sending notifications. (#11106)

Docs

Temporarily disabled documentation link checking while an upstream issue is resolved. (#11099)

<details> <summary>Maintenance</summary>

Reorganized the public metrics package so column metrics are importable from great_expectations.metrics under consistent module names. (#11109)
Fixed pageview analytics tracking and related console errors on the documentation site, and documented local environment setup for it. (#11093)

</details>

Contributors

Thanks to @jwalant-dattani (first contribution).

Open source →

great-expectations

Release timeline

Releases

Highlights

Changes

Bug fixes

Docs

Contributors

Source distribution (sdist)

Built distribution (wheel)

Highlights

Changes

Features

Bug fixes

Docs

Contributors

Highlights

Changes

Features

Bug fixes

Docs

Contributors

Project links

Key dates

2 maintainers

Credits

License

Requires

Provides Extra

Tags

Classifiers

Download files

Source Distribution

Built Distribution

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

Release history Release notifications | RSS feed

Highlights

Changes

Bug fixes

Docs

Contributors

Highlights

Changes

Features

Bug fixes

Docs

Contributors

Highlights

Deprecations

Changes

Features

Bug fixes

Contributors

Highlights

Changes

Features

Bug fixes

Contributors

Highlights

Changes

Features

Bug fixes

Docs

Contributors

Highlights

Changes

Features

Bug fixes

Docs

Contributors

Highlights

Changes

Bug fixes