`sparse` v0.16.0b2

Changes

• Update sparse_finch notebook and upgrade finch-tensor (#820)
• Allow no-copy construction from SciPy COO arrays. (#822)
• MAINT: Re-enable COO tests after fix (#814)
• Attempt to fix Heisenbug by skipping COO. (#813)
• update finch-tensor-python to fix tests (#812)
• skip changelog(deps): bump codecov/codecov-action from 4 to 5 (#811)
• Testing improvements and minor fixes. (#808)
• skip changelog(deps): bump mamba-org/setup-micromamba from 1 to 2 (#790)
• MAINT: Upgrade MLIR version (#785)
• ENH: Implement reshape function (#776)
• ENH: CSC and CSF formats for MLIR backend (#775)
• MNT: Move constructors to a direct style (#773)
• MNT: Refactor MLIR constructors (#771)
• ENH: Simple COO format (#768)
• skip changelog(deps): bump actions/setup-python from 3 to 5 (#735)
• MNT: Add CI mode to examples (#716)
• Add example notebook (#709)
• BUG: Skip test_where test (#707)
• Fix 1D return dtype in argmax/argmin (#704)
• MNT: sparse import refactor (#695)
• Add Array API tests for Numba backend. (#688)
• scipy.sparse fill-value fix. (#685)
• Upgrade finch-tensor (#684)
• Add tests for scipy.sparse.csgraph and scipy.sparse.linalg (#683)
• Update Finch backend (#678)
• Add support for NumPy 2.0. (#681)
• Add SpMV example (#677)
• Add MTTKRP example (#676)
• Add SDDMM example (#674)
• [pre-commit.ci] pre-commit autoupdate (#675)
• [pre-commit.ci] pre-commit autoupdate (#672)
• Do strict checking of Array API in CI job. (#671)
• [pre-commit.ci] pre-commit autoupdate (#670)
• [pre-commit.ci] pre-commit autoupdate (#667)
• Upgrade finch-tensor dep (#666)
• API: Rename PyData backend to Numba (#665)
• [pre-commit.ci] pre-commit autoupdate (#663)
• API: Enable finch backend for scipy.sparse (#664)
• Update finch-tensor version (#662)
• [pre-commit.ci] pre-commit autoupdate (#661)
• BUG: Fix CSR/CSC matmul (#660)
• Finch+PyData benchmarks (#652)
• API: Update asarray function (#658)
• Support Python 3.12 and drop 3.9. (#656)
• Update Finch backend (#657)
• Remove backend kwarg from asarray. (#655)
• [pre-commit.ci] pre-commit autoupdate (#654)
• Enable benchmarks (#651)
• [pre-commit.ci] pre-commit autoupdate (#650)
• API: Initial finch_backend (#649)
• API: Establish pydata_backend (#646)
• [pre-commit.ci] pre-commit autoupdate (#648)
• Modernize CI (#641)
• Improve performance of GCXS dot ndarray (#643)
• [pre-commit.ci] pre-commit autoupdate (#644)
• [pre-commit.ci] pre-commit autoupdate (#642)
• [pre-commit.ci] pre-commit autoupdate (#640)
• Remove dynamic PyTest options. (#637)
• CI: Remove Python 3.8 and update miniconda (#636)
• [pre-commit.ci] pre-commit autoupdate (#635)
• Always use np.intp for indices. (#634)
• Add missing linear algebra functions from array API (#633)
• Make SciPy a soft dependency (#632)
• [pre-commit.ci] pre-commit autoupdate (#630)
• NumPy 2.0 support (#625)
• Add direct imports from NumPy for Array API support (#624)
• Add functions to public API. (#623)
• Add more lints to Ruff for better code quality. (#621)
• API: Add kwargs to sparse.einsum (#620)
• DOC: Add array representations to sparse array construction examples (#606)
• Preserve gcxs compression (#601)
• Introduce pre-commit. (#617)
• Replace setup.py and versioneer with pyproject.toml and setuptools_scm. (#616)
• API: Improve sparse.asarray (#615)
• API: Refactor argmax and argmin (#614)
• API: Add squeeze method to COO (#613)
• API: Array API support - Part 1 (#612)
• Fixing "GCXS matmul => slice leads to incorrect results" (#611)
• Fix issue with .real/.imag with string dtypes (#609)
• Use matrepr for rich __str__ and _repr_html_ (#605)
• Fix upcast on in-place. (#598)
• Add CodeQL workflow for GitHub code scanning (#561)
• Minor bug: import warnings in _sparse_array.SparseArray (#597)
• test on 3.11 (#572)
• Convert two assignments to augmented source code #531 (#586)
• change parsing of the input in einsum (#579)
• Fix misfire of DeprecationWarning. (#581)
• Prepare for 0.14.0. (#578)
• Fix ReadTheDocs build. (#577)
• einsum (#564)
• Add .format support to GCXS and COO classes (#573)
• Improving indexing performance (#570)
• fix errors/crashes in multi-dimensional slicing (#569)
• Relative imports (helps vendoring) (#565)
• added svg logo and logo-generating script (#556)
• Fix nanmean for complex arrays. (#555)
• Allow an ndarray of integers as a shape. (#544)
• fix scipy url in intersphinx mapping (#543)
• Adding Vitter Random Sampling without replacement (#540)
• removed transpose before _dot_ndarray_coo function (#537)
• Solved Issue 529 (#534)
• Add a missing check for one-argument version of np.where. (#527)
• Update CodeCov action. (#525)
• Fix full_like with GCXS input. (#524)
• Prepare for 0.13.0. (#518)
• Change email on CoC. (#517)
• Fix a number of bugs. (#516)
• Fix incorrect sparse scalar array. (#511)
• Fixing NaN check bug (#508)
• Fix format conversion bugs. (#504)
• Fix DOK indexing and CI failures. (#501)
• Skip doctest on 32bits arch (#486)
• Black formatting in benchmark_matmul.py (#484)
• Fix up action. (#489)
• Add GitHub actions. (#488)
• Fix falsely triggered advanced indexing. (#482)
• Improve matmul benchmark (#467)
• Fix concatenate/stack for sparse.DOK. (#476)
• Added implementation of sparse.pad() (#474)
• Pytest failures - fixed with black re-formatting (#471)
• Miscellaneous improvements in docs (#466)
• Added warnings for NaN propagation (#469)
• Add DOK sparse constructor. (#464)
• Maintance fixes. (#462)
• Fixed typo in docstring (#457)
• Attempt to reformat docstring to match Numpydoc standard. (#456)
• Update docs for GCXS and CSR/CSC (#455)
• Fix typo in var function docstring (#454)
• Fix ZeroDivisionError on html_table. (#447)
• Document elementwise functions for DOK arrays (#451)
• Docstrings to CSC, CSR classes (#450)
• Modify GCXS docstring (#448)
• Prepare for 0.12.0 release. (#446)
• Make a number of immutability changes. (#445)
• Initial CSC/ CSR classes (#442)
• allow for use of different dtypes in coords, indices, and indptr (#441)
• Implement equality comparison for DOK arrays (#440)
• Fix dok getitem (#439)
• Document ability to slice and fancy index DOK arrays (#437)
• Add slicing and subset of fancy indexing to DOK array access (#436)
• Add fancy indexing support for DOK item assignment (#435)
• Add shape= kwarg in {zeros, ones, full}_like. (#423)
• use correct pointer type for coo shape (#421)
• add benchmarks (#419)
• Add gcxs to docs (#414)
• Move conj, real, and imag to SparseArray (#417)
• Add elemwise ops for GCXS (#412)
• Add nnz parameter to sparse.random (#410)
• Fix a number of random bugs. (#407)
• Add fill value pruning to GCXS (#409)
• Prepare for 0.11.2. (#405)
• Fix dot for complex dtypes. (#404)
• add reductions (#397)
• Prepare for 0.11.1. (#402)
• Fix dot for extremely small values. (#399)
• Prepare for 0.11.0 release. (#396)
• Fix warnings in the test suite. (#395)
• Improve dot performance (#389)
• Add the casting kwarg to astype. (#392)
• Fix ND being accepted as COO data. (#386)
• GCXS Slicing and reshape/resize/transpose/change_compressed_axes (#382)
• Fix format kwarg in zeros_like. (#381)
• added method moveaxis (#362)
• Fix scalar initialization with list inputs. (#356)
• Update roadmap. (#353)
• Choose sparse/dense type when doing tensordot (#349)
• Improve dot performance. (#350)
• Fix np.outer and np.ufunc.outer. (#347)
• Added multi-axis advanced indexing support (#343)
• added method swapaxes to COO (#344)
• Changelog for 0.10.0 (#342)
• Fix signature for clip. (#339)
• Prepare for Numba 0.49. (#337)
• Remove the HTML coverage from Azure as it generates its own. (#334)
• Add error msg to tensordot operation when ndim==0 (#333)
• Try to simplify the numba extension a little (#324)
• Replace linear_loc with ravel_multi_index, which is 3x faster (#330)
• Respect copy=False in astype (#328)
• Add Code of Conduct (#327)
• Update issue templates (#325)
• Implement asnumpy. (#322)
• Implement deterministic dask tokenization. (#320)
• Fix setting into numpy slice (#321)
• Remove sparse-dense broadcasting restriction. (#318)
• Add flatten method and outer function. (#317)
• Fix incorrect dtype on empty DOK->COO. (#315)
• 0.9.0 changelog (#313)
• Add changelog for 0.9.0. (#312)
• ENH: Add numba extension support for COO objects (#307)
• Fix attribute access for array_function. (#309)
• Add stack and concatenate (#303)
• Avoid making a copy of coords (#296)
• add repr_html (#283)
• Diagonal/diagonalize functions (#289)
• Change code formatter to black. (#284)
• Fix indexing for empty lists. (#282)
• Modify the visibility of a lot of modules that should be private. (#280)
• Fix scalar operations in certain cases. (#278)
• Change to range length and remove dead code. (#277)
• Convert Numba lists. (#264)
• array_function test coverage; numpy 1.17; result_type() (#272)
• Allow mixed ndarray-sparse ops where the shape doesn't change. (#275)
• Route gufuncs through array_function. (#274)
• Remove redundant import. (#273)
• array_function suppresses AttributeError (#267)
• Add CSD (#258)
• Modifications to enable sparse xarrays (#261)
• Fix empty tensordot. (#256)
• Move Azure Pipelines to Jobs (#248)
• Add isposinf and isneginf. (#253)
• Add resize method and change reshape (#251)
• Switch badge to Azure Pipelines. (#247)
• Azure pipelines (#246)
• Set up CI with Azure Pipelines [skip ci] (#245)
• Add std and var. (#244)
• Fix 1 size non-empty broadcast. (#243)
• Fix LGTM warnings and improve reduce performance. (#240)
• [WIP] Add array_function support. (#239)
• Support reduce fill_value in specific cases. (#238)
• Restructure requirements. (#236)
• Clarify error messages (#232)
• Drop support for legacy Python 2.7 (#235)
• Fix ...