PyPI · #2719 most downloaded on PyPI

pygeohash

3.5.12.2M downloads/mowdm0006/pygeohash ↗New releases in your daily digest, changelog included.

Python module for interacting with geohashes

Last release 28 days ago
06 Sep 2026
Ships unpredictably
gaps range from 8 days to 8.9 years
Most releases are documented
notes for 16 of 19 stable releases
Nothing withdrawn
no release was ever pulled
11 years old
19 releases · first in 2016

Release timeline

19 releases since 2016

2016201820202022

Releases

3.5.16 Sept 2026
chore: drop Python 3.9 support, bump to 3.5.1 (fixes cp39 wheel build failure) by @wdm0006 in #152
What's Changed

chore: drop Python 3.9 support, bump to 3.5.1 (fixes cp39 wheel build failure) by @wdm0006 in #152

Full Changelog: v3.5.0...v3.5.1
Open source →
3.3.222 Aug 2026
ci: efficiency + hygiene tuning by @wdm0006 in #76
What's Changed

ci: efficiency + hygiene tuning by @wdm0006 in #76

ci: move CodeQL off per-PR to push + weekly schedule by @wdm0006 in #77

Return original geohashes from cardinal selectors by @wdm0006 in #82

Gate docstring examples with doctest by @wdm0006 in #81

fix: geohashes_in_box() drops cells at box corners by @wdm0006 in #83

Use canonical validation for approximate distance by @wdm0006 in #85

fix: compute geohash mean longitude as a circular mean by @wdm0006 in #86

ci: build Linux ARM64 wheels natively by @tamird in #90

perf: keep optional imports lazy by @tamird in #91

test: stop the test suite from rewriting tracked files under examples/images/ by @wdm0006 in #92

Validate BoundingBox ordering instead of silently returning empty results by @wdm0006 in #93

fix: get_adjacent() east/west neighbors across the antimeridian by @wdm0006 in #94

Fix visualization extra imports by @wdm0006 in #101

fix: accept uppercase and mixed-case geohashes in decode()/decode_exactly() by @wdm0006 in #100

fix: reject bare geohash strings in the collection statistics API by @wdm0006 in #103

fix: reject boolean coordinates and precision values by @wdm0006 in #104

Validate BoundingBox coordinates against world bounds by @wdm0006 in #106

Reject over-precision geohashes when decoding by @wdm0006 in #108

fix: clamp generated Folium grid viewport to world bounds by @wdm0006 in #114

Reject non-string values in geohash Series detection by @wdm0006 in #115

Audit installed dependencies in security CI by @wdm0006 in #116

fix: validate empty visualization collections by @wdm0006 in #121

Reject booleans in the direct C encoder API by @wdm0006 in #120

ci: smoke-test built distributions before publishing by @wdm0006 in #117

Make the logging helpers emit records and register undoubled logger names by @wdm0006 in #127

Validate the geohash and direction arguments in get_adjacent() by @wdm0006 in #126

Publish the cross-library benchmark comparison in the docs by @wdm0006 in #125

Fail the randomized accuracy check on codec regressions by @wdm0006 in #130

Reject boolean coordinates in BoundingBox by @wdm0006 in #131

[autopilot] release v3.3.2 by @wdm0006 in #129

Full Changelog: v3.3.1...v3.3.2

One column per quarter.

Additional notes2 sources agree

[bugfix] behavior change: constructing a BoundingBox with min_lat > max_lat or min_lon > max_lon now raises a ValueError naming the offending pair and the interleaved (min_lat, min_lon, max_lat, max_lon) field order. The grouped-argument mistake BoundingBox(min_lat, max_lat, min_lon, max_lon) previously produced an inverted box that silently degraded everywhere - geohashes_in_box returned [], is_point_in_box returned False, and a box did not intersect itself. Boxes written to span the antimeridian, which the library does not support, are rejected by the same check. Degenerate boxes where a minimum equals its maximum remain valid, and no box get_bounding_box produces is affected.
[bugfix] strengthened validation across encoding, decoding, adjacent-cell lookup, bounding boxes (including out-of-world and boolean coordinates), collection statistics, Pandas Series detection, and visualization collection inputs; invalid values now fail consistently instead of producing misleading results or low-level errors.
[bugfix] decoding and geohash_approximate_distance now accept uppercase and mixed-case geohashes, and cardinal selectors preserve the caller's original geohash string.
[bugfix] fixed antimeridian behavior in adjacent-cell traversal, circular longitude means, and Folium grid viewports near world boundaries.
[bugfix] fixed missing cells at bounding-box corners and corrected visualization imports for isolated viz installations.
[bugfix] logging helpers now emit records at their requested level and avoid doubled pygeohash.pygeohash.* logger names.
[perf] optional modules remain lazy until their public attributes are first accessed.
[packaging] publication now smoke-tests built wheels and source distributions, and Linux ARM64 wheels are built natively.
[docs] published the cross-library benchmark comparison as a README section and docs/source/benchmarks.rst, generated by scripts/run_comparison_benchmark.py.
[tests] dependency-free docstring examples are now verified as doctests, and the randomized accuracy check fails on codec regressions instead of passing them over.

Open source →

3.3.117 Jul 2026

Validate precision in C encode functions to prevent stack buffer overflow by @wdm0006 in #65

What's Changed

Validate precision in C encode functions to prevent stack buffer overflow by @wdm0006 in #65
Fix misleading encode_strictly: document it as identical to encode by @wdm0006 in #63
Remove accidentally committed todo.db and gitignore it by @wdm0006 in #64
Reject non-finite coordinates in C encoders by @wdm0006 in #70
Keep bounding-box enumeration within geographic limits by @wdm0006 in #69
Return zero approximate distance for identical geohashes by @wdm0006 in #74
Ship the PEP 561 marker by @tamird in #73
[autopilot] release v3.3.1 by @wdm0006 in #75

New Contributors

@tamird made their first contribution in #73

Full Changelog: v3.3.0...v3.3.1

Open source →

Additional notes2 sources agree

[bugfix] encode/encode_strictly in the C extension now reject a precision outside 1-12 with a ValueError. Calling pygeohash.cgeohash.geohash_module.encode directly with a larger precision previously overran a fixed-size buffer.
[bugfix] the C encoders now reject NaN and infinite coordinates with a ValueError. Passing an infinite longitude previously hung the call.
[bugfix] geohash_approximate_distance now returns 0.0 for two identical geohashes instead of the precision-table value for their shared prefix.
[bugfix] geohashes_in_box now clamps its sampling range to latitude +/-90 and longitude +/-180, so boxes touching the world limits enumerate correctly.
[bugfix] plot_geohashes works again with matplotlib >= 3.9, which removed matplotlib.cm.get_cmap.
[packaging] the py.typed marker is now shipped in the wheel and sdist, so type checkers finally see pygeohash's inline type hints. The "Typing :: Typed" classifier is now accurate.
[docs] encode_strictly no longer claims to perform extra validation. It is identical to encode and is kept only as a back-compatible alias.

Open source →

3.3.025 May 2026

Add QEMU to linux runners when building wheels by @MatiasFrank in #55

What's Changed

Add QEMU to linux runners when building wheels by @MatiasFrank in #55
trying out wabi theme by @wdm0006 in #56
Add cross-library speed benchmark for geohash operations by @wdm0006 in #57
Optimize encode/decode/bbox by trimming Python wrapper overhead by @wdm0006 in #58
Speed up C decode: cache types, skip the ExactLatLong detour by @wdm0006 in #59

Full Changelog: v3.2.2...v3.3.0

Open source →

Additional notes2 sources agree

[perf] reworked the C decode path: cache the LatLong/ExactLatLong types instead of importing them per call, decode straight into a LatLong without building and discarding an intermediate ExactLatLong, and construct the result tuple directly (decode ~3x faster, bounding box ~1.5x faster)
[perf] trimmed the Python wrappers: dropped redundant per-character validation (the C extension already validates) and per-call debug logging from the encode/decode/bounding-box hot paths
[bugfix] fixed a latent out-of-bounds read in the C decoder for input bytes >= 128 (e.g. multibyte UTF-8); these are now cleanly rejected as invalid characters
[tests] added a cross-library benchmark suite comparing pygeohash against other geohash libraries, behind an optional 'benchmark' extra

Open source →

3.2.29 Dec 2025

Build wheels for ARM64 on Linux as well by @MatiasFrank in #53

What's Changed

Build wheels for ARM64 on Linux as well by @MatiasFrank in #53
release prep by @wdm0006 in #54

Full Changelog: v3.2.1...v3.2.2

Open source →

Additional notes2 sources agree

Fresh release for more wheels

Open source →

3.2.125 Nov 2025

Fix invalid escape sequence '*' by @MatiasFrank in #49

What's Changed

Fix invalid escape sequence '*' by @MatiasFrank in #49
release prep by @wdm0006 in #51

New Contributors

@MatiasFrank made their first contribution in #49

Full Changelog: v3.2.0...v3.2.1

Open source →

Additional notes2 sources agree

bugfix

Open source →

3.2.025 Jul 2025

lat long checks by @wdm0006 in #48

What's Changed

Docs fixes by @wdm0006 in #46
lat long checks by @wdm0006 in #48

Full Changelog: v3.1.3...v3.2.0

Open source →

Additional notes2 sources agree

Added input validation for latitude and longitude in encode functions to prevent invalid coordinates from being processed
Latitude values must now be between -90.0 and 90.0 degrees
Longitude values must now be between -180.0 and 180.0 degrees
Both encode() and encode_strictly() functions now raise ValueError with descriptive messages for invalid coordinates

Open source →

3.1.312 May 2025

email update by @wdm0006 in #45

What's Changed

email update by @wdm0006 in #45

Full Changelog: v3.1.2...v3.1.3

Open source →

Additional notes2 sources agree

Updating pypi email to registered domain

Open source →

3.1.28 May 2025

bugfix for optional dependencies in types by @wdm0006 in #44

What's Changed

bugfix for optional dependencies in types by @wdm0006 in #44

Full Changelog: v3.1.1...v3.1.2

Open source →

Additional notes2 sources agree

bugfix inadvertantly requiring numpy and pandas for type checks; they are now required for mypy but not at runtime.

Open source →

3.1.14 May 20252 sources agree

Added input validation to raise on invliad precisions and geohashes before sending to c lib

Added input validation to raise on invliad precisions and geohashes before sending to c lib

Open source →

3.0.114 Mar 2025

Nothing published for this version

3.0.09 Mar 20252 sources agree

[meta] rewrote core logic from scratch in cpython for performance and to remove geohash.py dependency

[meta] rewrote core logic from scratch in cpython for performance and to remove geohash.py dependency
[meta] relicensed to MIT for more permissive use
[meta] removed numba and numpy deps

Open source →

2.1.01 Mar 20252 sources agree

[bounding_box] added bounding box module

[bounding_box] added bounding box module
[viz] added visualization module
[docs] updated docstrings across the library
[meta] added cursor rulesfiles

Open source →

2.0.327 Feb 20252 sources agree

[meta] restructured as pyproject.toml based project

[meta] restructured as pyproject.toml based project
[meta] added ruff and tox configs
[meta] added gh actions workflows
[docs] added proper sphinxdocs and background info

Open source →

2.0.227 Feb 2025

Nothing published for this version

1.2.05 Apr 20162 sources agree

[stats] added standard deviation

[stats] added variance
[stats] added standard deviation

Open source →

1.1.023 Jan 20162 sources agree

* Added stats module * [stats] added mean geohash * [stats] added northernmost * [stats] added southernmost * [stats] added westernmost * [stats] adde

Added stats module
[stats] added mean geohash
[stats] added northernmost
[stats] added southernmost
[stats] added westernmost
[stats] added easternmost

Open source →

1.0.1

Nothing published for this version

1.0.07 Jan 20162 sources agree

New repository created, based off of geohash

New repository created, based off of geohash
Basic approximate distance added

Open source →

Your coding agent can read these notes before it upgrades. Set up the MCP server →