Version 2.6.6

New features

• feat: add slicing CUDA kernels by @ManasviGoyal in #3140
• feat: add some misc CUDA kernels by @ManasviGoyal in #3141
• feat: Adding awkward.semipublic submodule by @tcawlfield in #3152
• feat: 3154 from_parquet should be able to read partial columns by @tcawlfield in #3156
• feat: add reduce kernels by @ManasviGoyal in #3136
• feat: copy_behaviors to make sub-classing easy by @Saransh-cpp in #3137

Bug-fixes and performance

• fix: use grid-stride looping for kernels with variable-length loops by @ManasviGoyal in #3130
• fix: don't let CuPy iterate over Index with Python for loops by @jpivarski in #3142
• fix: correct dtypes for numpy v2 by @ianna in #3159
• fix: ak.from_numpy should fail on zero-dimensional arrays. by @tcawlfield in #3161

Other

• test: reducer CUDA kernel tests by @ManasviGoyal in #3162
• ci: Ensure attestations for awkward-cpp sdist and wheels by @matthewfeickert in #3135
• ci: faster wheel builds by @henryiii in #3146
• chore: update pre-commit hooks by @pre-commit-ci in #3139
• chore(deps): bump the actions group across 1 directory with 3 updates by @dependabot in #3153
• chore: trying atomics and tree reduction for CUDA reducer kernels by @ManasviGoyal in #3123
• chore: update pre-commit hooks by @pre-commit-ci in #3145

Full Changelog: v2.6.5...v2.6.6