@prisma/client

Today, we are issuing the 4.6.1 patch release.

Fixes in Prisma Client

• Prisma Client regression bug after upgrading to 4.6.0: findMany errors with PANIC: index out of bounds: the len is 0 but the index is 0
• upsert() with nested selection errors with called Option::unwrap()on aNone value in 4.6.0

Fix in Prisma Migrate

• Prisma 4.6.0 drops and recreates enum field when running db push even if the field has not changed

🌟 Help us spread the word about Prisma by starring the repo or tweeting about the release. 🌟

Highlights

Interactive Transactions for Prisma Data Proxy (Preview)

In 3.8.0, we disabled the interactiveTransactions Preview feature when using the Prisma Data Proxy. This was because the API was not yet supported.

In this release, we're removing the limitation. You can now try the Preview version of interactive transactions with the Prisma Data Proxy. Re-generate Prisma Client using prisma generate --data-proxy after enabling the Preview feature.

Note: The interactiveTransactions Preview feature flag is still needed. We will remove this in a future version when the feature is stable.

Try it out and let us know your thoughts in our interactive transactions feedback GitHub issue.

Native database level upserts for PostgreSQL, SQLite, and CockroachDB

Prisma’s upsert is one of its most powerful and most convenient APIs. In this release, Prisma will now default to the native database upsert for PostgreSQL, SQLite, and CockroachDB whenever possible.

Prisma will use the native database upsert if:

• There are no nested queries in the upsert's create and update options
• The query modifies only one model
• There is only one unique field in the upsert's where option
• The unique field in the where option and the unique field in the create option have the same value

Prisma Client's upsert operation was implemented on a Prisma-level and did not use the native database implementations like, e.g., INSERT .. ON CONFLICT .. UPDATE SET. This allowed Prisma to also upsert nested queries.

The Prisma-implementation came at a cost. In some scenarios, it was more likely for a transaction to roll back because of a conflict when multiple upsert operations were being executed in parallel, and the multiple queries often took longer than the native database query would have taken.

Try it out and let us know what you think. If you run into any issues, don't hesitate to create a GitHub issue.

Relation Mode improvements (Preview)

In 4.5.0, we renamed the "Referential Integrity" Preview feature to "Relation Mode". We also changed the datasource property name of the feature to relationMode.

In this release, we fixed all remaining known bugs of relationMode = "prisma" and cleaned up our documentation. You can now read about Relation mode on its own documentation page which is up to date with the implementation.

If you encounter any problems please comment on our feedback issue. We plan to make this Generally Available soon.

extendedWhereUnique improvements (Preview)

In 4.5.0, we introduced the extendedWhereUnique Preview feature to allow filtering for non-unique properties in unique where queries. In this release, we're adding new rules to decide when concurrent findUnique queries get batched into a findMany query.

Unfortunately, we forgot to adapt our findUnique query batch optimization, which turns multiple concurrent findUnique queries into a single findMany query when possible — GitHub issue.

Therefore, findUnique queries will get batched into a findMany query if:

• All criteria of the filter must be on scalar fields (unique or non-unique) of the same model you're querying
• All criteria must use the equal's filter, whether that's via the shorthand or explicit syntax (where: { field: <val>, field1: { equals: <val> } })

Conversely, suppose the filter object contains any boolean operators, relation filters, or scalar filters that are not using equals. Prisma will fall back to executing the findUnique queries independently.

Let us know your thoughts and share your feedback on the Preview feature in this GitHub issue.

Fixes and improvements

Prisma Client

• upsert across HTTP requests has a race condition
• Client tests for Vitess, with relationMode=prisma
• Concurrent relation upserts conflict
• upsert() should do ON CONFLICT DO UPDATE/NOTHING in postgresql
• Cannot insert Floats (≥9.223372037e18 and <1e21) or (>-1e21 and ≤-9.223372037e18)
• Always serialize float fields in exponent notation
• Postgres integer column silently overflows to bad value
• Add duration to mongo logging
• upsert(): Unique constraint on DateTime sqlite
• "Unique constraint failed on the fields" when using upsert
• using connect in interactive transaction fails
• MongoDB does not log queries with binary Engine
• [MongoDb] Query logger produces incorrect output
• Float values around 2^63 get interpreted as invalid intergers and throw a graphql error
• referentialIntegrity/relationMode preview feature (referentialIntegrity = "prisma"): Make NoAction a synonym/alias of Restrict for the emulation, for all databases except PostgreSQL & SQLite.
• Client: In 1:1,1:n, m:n relations using OnDelete: SetNull with referentialIntegrity = "prisma" user.delete() should fail, but succeeds.
• calling findUnique concurrently on a model with a compound unique constraint causes it to return null
• Support interactive transactions in the Data Proxy

Prisma

• Re-Introspection: referentialIntegrity = prisma not respected when using @@map()
• Validation fails to detect invalid SetNull referential action referencing non-optional fields
• Schema: add validation for datasource property referentialIntegrity & relationMode, so only one can be set.
• Create TypeScript tests for #15655
• referentialIntegrity policy is lost during re-introspection
• Relation fields are removed after npx prisma db pull
• Re-introspection should not remove relations when relationMode="prisma"

Language tools (e.g. VS Code)

• Schema autocompletion intellisense don't match custom datasource name

Design Partner Program

Are you building data-intensive applications in serverless environments using Prisma? If so, you should join our Design Partner Program to help us build the tools that best fit your workflows!

The Design Partner Program aims to help development teams solve operational, data-related challenges in serverless environments. Specifically, we’re looking to build tools that help with the following problems:

• Solutions to listen and react to database changes in real time are either brittle or too complex to build and operate.
• Coordinating workflows executed via a set of isolated functions or services spreads that coordination logic across these services instead of keeping it centralized and maintainable. This adds unnecessary overhead and clutter to your business logic.
• Optimizing the data access layer for scaling performance often involves projecting data into denormalized views, or caching. These methods come with complex logic to figure out strategies for cache invalidation or preventing to use stale data.
• Building web applications on modern Serverless platforms such as Vercel or Netlify often breaks down as soon as you need to execute on any of the topics listed above. This pushes to re-platform on a traditional infrastructure, delaying projects, and losing productivity benefits offered by Vercel or Netlify.

Submit an application through our application form to join the Prisma Design Partner Program to take advantage of new features that you won't have to build or maintain yourselves.

Prisma Data Platform

We're working on the Prisma Data Platform — a collaborative environment for connecting apps to databases. It includes the following:

• Data Browser for navigating, editing, and querying data
• Data Proxy for your database's persistent, reliable, and scalable connection pooling.
• Query Console for experimenting with queries

Try it out. Let us know what you think!

📺 Join us for another "What's new in Prisma" live stream

Learn about the latest release and other news from the Prisma community by joining us for another "What's new in Prisma" live stream.

The stream takes place on YouTube on Thursday, November 10 at 5 pm Berlin | 8 am San Francisco.

Credits

Huge thanks to @cmd-johnson for helping!