github-actions
released this
08 Oct 15:05
·
20 commits
to main
since this release
Docs
- python: Add example of python package with nanoarrow C extension (#645)
Feat
- Add ArrowArrayView accessors to inspect buffer properties (#638)
- String/Binary View Support (#596)
- add Footer decoding (#598)
- Revendor flatcc (#592)
- Add IPC integration test executable (#585)
- Add
ArrowArrayViewCompare()
to check for array equality (#578) - Add IPC stream writing (#571)
- add ipc RecordBatch encoding (#555)
- add ArrowIpcOutputStream (#570)
- Add IPC schema encoding (#568)
- Add IPC writer scaffolding (#564)
- Add ArrowArrayViewComputeNullCount (#562)
- Add Meson support in nanoarrow_device (#484)
- Meson build system for nanoarrow-ipc extension (#483)
- Add support for run-end encoded array (#507)
- Add float16 support for
ArrowArrayViewGet{Double,Int,UInt}Unsafe()
(#501) - Add support for appending values to half float
ArrowArray
(#499) - extensions/nanoarrow_device: Implement asynchronous buffer copying (#509)
- python: Add StringView and BinaryView IO to Python bindings (#637)
- python: Implement array from buffer for non-CPU arrays (#550)
- python: Implement bindings to IPC writer (#586)
- python: Implement CUDA build in Python bindings (#547)
- r: Add float16 support for R bindings (#650)
- r: Implement string view support in R bindings (#636)
- r: Allow opt-out of warning for unregistered extension types (#632)
- r: Add bindings to IPC writer (#608)
Fix
- Remove unreachable code (#649)
- Properly ingest Binary View types without variadic buffers (#635)
- python schema repr does not truncate output (#628)
- Accommodate IPC messages without continuation bytes (#629)
- Ignore empty (but present) union validity bitmaps from before 1.0 (#630)
- Only validate relevant type_ids for array view slice (#627)
- Improve validation of offset buffers for sliced arrays (#626)
- Ensure CMake linking against built/installed nanoarrow works for all components (#614)
- Ensure footer test passes on big endian (#609)
- ensure 1 is written for boolean True (#601)
- Ensure that schema metadata is always present even if empty (#591)
- Include missing cases in
ArrowArrayInitFromType()
(#588) - Silence warning when compiling nanoarrow.hpp on at least one version of MSVC (#590)
- don't require metadata order in nanoarrow_ipc_integration (#589)
- IPC streams did not include RecordBatch headers (#582)
- Fix Meson build for separated nanoarrow_testing target (#574)
- Ensure
children
is NULL for zero children in ArrayViewAllocateChildren (#556) - CMake deprecation warnings from subprojects (#535)
- Meson install header files and pkgconfig (#542)
- Fix symbol export visibility in c_data_integration_test (#531)
- Fix Meson include directories (#532)
- Ensure we don't call cuMemAlloc with 0 bytesize (#534)
- Ensure ArrowDeviceArray implementation for AppleMetal passes tests on newer MacOS (#527)
- Check for offset + length > int64_max before using the value to calculate buffer sizes (#524)
- check
run_ends_view->length
before accessing its values (#518) - Force static library build on Windows when building with Meson (#496)
- ci: Fix verify, meson-build, and docker-build weekly runs (#581)
- ci: Fix build and test of nanoarrow on centos7 and s390x (#576)
- ci: Pin r-lib actions as a workaround for latest action updates (#572)
- ci: Fix verification workflow (#552)
- ci: Stop building unbuildable image based on centos7 (#553)
- python: Fix detection of cuda library on hosted runner (#554)
- r: Avoid flatcc aligned_alloc() call when compiling R package (#494)
Refactor
- Consolidate per-target actions in CMakeLists.txt (#573)
- Separate implementation from interface for nanoarrow_testing component (#561)
- Separate components into folders under src/nanoarrow (#536)
- Use ArrowStringView C++ literal in tests (#528)
- Move Meson build targets to top level directory (#530)
- Simplify Meson test generation (#525)
- Remove CMake requirement from Meson IPC config (#522)
- Use inttypes.h macros instead of casts to print fixed-width integers (#520)
- Consolidate device extension into main project (#517)
- Consolidate IPC extension into main project (#511)
- extensions/nanoarrow_device: Migrate CUDA device implementation to use the driver API (#488)
- python: Split ArrowArray and ArrowArrayStream modules (#559)
- python: Separate schema cython classes into their own module (#558)
- python: Split buffer Cython internals into a separate module (#549)
- python: Split device functionality into its own module (#548)
- python: Split type identifier utilities into thier own module (#545)
- python: Extract utility functions into _utils.pyx (#529)
- r: Use JSON in experimental R vctrs extension type (#533)