Merged pull requests by @AryanRG in external public repositories, grouped by area. Each entry links to the upstream PR and its review history.
Merged PRs: 86 · Projects: 82
- chardet/chardet: Fix UTF-32 BOM detection when max_bytes cuts a code unit (#405)
- docling-project/docling: Preserve table footnotes in JATS documents (#4818)
- jaraco/inflect: Handle signed numeric literals in inflect interpolation (#281)
- jawah/charset_normalizer: Fix exit status for module CLI entry points (#817)
- mammothb/symspellpy: Handle deleting the final dictionary entry (#209)
- mindee/doctr: Handle zero-width horizontal crops (#2173)
- RapidAI/RapidOCR: Preserve resize ratios across both size limits (#761)
- rapidfuzz/RapidFuzz: Fix zero deletion cost in Python Levenshtein distance (#502)
- stanfordnlp/stanza: Fix preserved manual MWT word offsets after automatic expansion (#1706)
- tesseract-ocr/tesseract: Propagate min_characters_to_try to automatic OSD engine (#4651)
- tesseract-ocr/tesseract: Escape input names in ALTO and PAGE XML (#4652)
- executablebooks/markdown-it-py: Preserve task markers in indented list code (#451)
- executablebooks/mdit-py-plugins: Avoid gfm_autolink hangs in link labels (#161)
- executablebooks/MyST-Parser: Preserve literal alt text in HTML images (#1214)
- lepture/mistune: Fix indented definition-list markers (#506)
- csinva/imodels: Fix DistilledRegressor recognition in sklearn ensembles (#323)
- CyberAgentAILab/cmaes: Copy the initial mean in SepCMA (#231)
- david-cortes/contextualbandits: Preserve reward estimates in doubly robust evaluation (#84)
- Farama-Foundation/PettingZoo: Allow non-string agent IDs in aec_to_parallel (#1528)
- KevinMusgrave/pytorch-metric-learning: Preserve Tuplet endpoint values and zero-margin gradients (#765)
- MaxHalford/prince: Preserve fitted column names in PCA inverse_transform (#261)
- neuropsychology/NeuroKit: Normalize signal_merge inputs before merging (#1249)
- optuna/optunahub: Invalidate stale submodules when switching local registry roots (#154)
- rodrigo-arenas/Sklearn-genetic-opt: Refresh optimization direction after set_params (#391)
- sbi-dev/sbi: Update ensemble potential when weights change (#2094)
- shankarpandala/lazypredict: Preserve test-row indexes in prediction results (#575)
- SimplexLab/TorchJD: Use squared distances for Krum scores (#835)
- sknetwork-team/scikit-network: Preserve exact integer node identifiers when reindexing edge lists (#611)
- skops-dev/skops: Support relative output paths in skops update (#559)
- skrub-data/skrub: Fix TableReport counts for unused pandas categories (#2348)
- statsmodels/statsmodels: Preserve Series prediction index with extra parameters (#10590)
- tslearn-team/tslearn: Exclude self from radius queries at large and infinite radii (#786)
- TylerYep/torchinfo: Fix recurrent mult-add counts for cells and unbatched inputs (#429)
- unit8co/darts: Fix MIDAS padding for short multivariate and sampled series (#3304)
- zillow/quantile-forest: Fix sample weight indexing when collecting leaf samples (#168)
- data-apis/array-api-compat: Fix NumPy clip with read-only integer bounds (#486)
- dgasmith/opt_einsum: Track peak size in the BranchBound memory fallback (#266)
- patrick-kidger/lineax: Preserve complex NumPy scalars in operator structure checks (#267)
- pydata/xarray: Fix unstacking after selecting stacked features (#11808)
- tensorly/tensorly: Fix complex CP weights in alternate MTTKRP implementations (#691)
- datachain-ai/datachain: Align read_values rows with explicit output order (#2111)
- dpkp/kafka-python: Raise timeout for unfinished producer futures at deadline (#3169)
- feast-dev/feast: Preserve colliding source timestamps in Dask retrieval (#7071)
- fsspec/filesystem_spec: Create parent directories before FSMap batch writes (#2287)
- kayak/pypika: Fix negative subtraction generating SQL comments (#3548)
- langfuse/langfuse-python: Preserve explicit parent for child events (#2045)
- Mirascope/mirascope: Preserve Annotated Field constraints in tool schemas (#2905)
- psycopg/psycopg: Preserve JSONB OID in custom adapter fallback (#1447)
- weaviate/weaviate-python-client: Parse structured error messages (#2209)
- agronholm/anyio: Fix short writes in UNIXSocketStream.send_fds() (#1422)
- agronholm/exceptiongroup: Honor traceback limits for exceptions nested in groups (#159)
- graphql-python/graphql-core: Close the consumed subscription source iterator (#286)
- openlawlibrary/pygls: Handle offsets in empty documents (#639)
- openlawlibrary/pygls: Apply empty notebook metadata updates (#641)
- openlawlibrary/pygls: Avoid request-future mutation during shutdown (#642)
- rthalley/dnspython: Restore negative integer indexing of RRsets and resolver answers (#1361)
- websocket-client/websocket-client: Bracket IPv6 hosts in HTTP proxy CONNECT authority (#1076)
- cdgriffith/Box: Accept positional source and keyword overrides in Box.update (#342)
- msgspec/msgspec: Release encode_lines iterators (#1284)
- nhairs/python-json-logger: Honor exc_info_as_array for cached exception text (#95)
- pypa/distlib: Preserve regular-file symlinks in manifest discovery (#277)
- pypa/flit: Preserve URL and marker brackets when normalizing dependency extras (#834)
- pypa/packaging: Fix inherited metadata class field validation and serialization (#1454)
- pypa/pipx: Allow selecting dependencies that only ship shell completions (#2157)
- pypa/pyproject-metadata: Normalize extra names and reject collisions (#342)
- python-cmd2/cmd2: Fix history ranges starting past the end (#1778)
- python-cmd2/cmd2: Route annotated command help through the application output stream (#1779)
- python-poetry/poetry-core: Fix empty static project classifiers (#986)
- tox-dev/filelock: Isolate subclass singleton registries (#855)
- tox-dev/platformdirs: Keep root directory accessors consistent during isolation (#625)
- tox-dev/pyproject-api: Contain response serialization failures (#322)
- apache/commons-codec: Handle names emptied during Match Rating Approach cleaning (#448)
- apache/commons-io: Avoid sorting mixed OpenOption types in byte-array channel builder (#887)
- apache/commons-lang: Fix resource paths for unnamed packages (#1824)
- apache/commons-numbers: Fix representable minimum-value powers (#223)
- apache/poi: Reject prohibited NUL and ETX characters in sheet names (#1433)
- Kong/unirest-java: Encode URL-encoded form field names (#582)
- deepset-ai/haystack: Reject multi-label cross encoders in SASEvaluator (#13251)
- esa/torchquad: Promote Boolean samples before quadrature (#271)
- Farama-Foundation/Minari: Preserve image encoding when combining datasets (#375)
- kozistr/pytorch_optimizer: Restore cosine restart schedules when resuming (#670)
- lab-cosmo/torch-pme: Fix slab corrections in combined potentials (#226)
- lezcano/geotorch: Fix wide AlmostOrthogonal sampling and registration (#56)
- nltk/nltk: Fix FreqDist cached totals after removing entries (#3995)
- optuna/optuna: Fix pickling of GA samplers (#6933)
- scikit-activeml/scikit-activeml: Preserve array-like sample weights in EMC bootstraps (#607)
This page is generated from GitHub's merged-PR records. Open, draft, and closed-unmerged PRs are not counted.