astral-sh
uv
Blog
Docs
Changelog
Blog
Docs
Changelog
Overview
Branches
Benchmarks
Runs
Performance History
Latest Results
parse GraalPy platform information before version (#19786) ## Summary Fix `fetch-download-metadata.py` by having the GraalPy platform information be processed before parsing the Python version information. ## Test Plan * Ran script locally. Without this change it fails.
main
1 hour ago
Bump for Python 3.13.14 and 3.14.6 Includes the python-build-standalone 20260610 release.
sync_py_20260610
3 hours ago
Clean up native auth integration state
zb/native-auth-as-json
5 hours ago
Reset index settings on URL replacement
koriyoshi2041:rios/uv-19759-index-settings
7 hours ago
Avoid normalizing source distribution names twice (#19784) ## Summary Prior to this change, `SourceDistFilename::parse` constructed and normalized a new `PackageName` for every filename before comparing it with the already-normalized expected package name. This PR compares the filename bytes directly on the successful path, handling ASCII case and separator equivalence without allocating. If the names do not match, we still use the existing parser so invalid package names and mismatches retain their existing errors. In an aggregate benchmark parsing six representative source distribution filenames, this reduces parsing time from 172.7 ns to 78.7 ns, roughly 54%.
main
7 hours ago
Avoid normalizing source distribution names twice
charlie/optimize-distribution-filename-parsing
7 hours ago
Avoid panics for trailing extra separators (#19779) ## Summary Prior to this change, requirement strings with an extra ending in a separator caused the parser to panic: ```text foo[bar-] foo[bar_] foo[bar.] ``` The extra-name scanner accepted the trailing separator, then assumed that `ExtraName` validation could not fail. This PR validates that extras end in an alphanumeric character before constructing the `ExtraName`. Invalid extras now return a positioned parse error, with regression coverage for all three separators.
main
8 hours ago
Reject malformed hash options in requirements files (#19783) ## Summary Prior to this change, malformed hash options such as: ```text flask==3.0.0 --hash--hash=sha256:deadbeef ``` were accepted as if they contained a single valid `--hash` option. The parser used `eat_while("--hash")` for the first hash token, which consumed repeated matching characters instead of requiring one exact option. Subsequent hash tokens already used exact matching. This PR uses exact matching for the first hash option too, so malformed spellings return a positioned parser error.
main
8 hours ago
Latest Branches
CodSpeed Performance Gauge
-5%
Sync Python 3.13.14 and 3.14.6
#19787
3 hours ago
ade9078
sync_py_20260610
CodSpeed Performance Gauge
0%
Rewrite the native authentication storage scheme
#18907
5 hours ago
9d7f18b
zb/native-auth-as-json
CodSpeed Performance Gauge
-1%
Reset stale index settings when replacing URLs
#19785
7 hours ago
6cf1b1a
koriyoshi2041:rios/uv-19759-index-settings
© 2026 CodSpeed Technology
Home
Terms
Privacy
Docs