NVIDIA/cudf v23.02.00

on GitHub

🚨 Breaking Changes

• Pin dask and distributed for release (#12695) @galipremsagar
• Change ways to access ptr in Buffer (#12587) @galipremsagar
• Remove column names (#12578) @vuule
• Default cudf::io::read_json to nested JSON parser (#12544) @vuule
• Switch engine=cudf to the new JSON reader (#12509) @galipremsagar
• Add trailing comma support for nested JSON reader (#12448) @karthikeyann
• Upgrade to arrow-10.0.1 (#12327) @galipremsagar
• Fail loudly to avoid data corruption with unsupported input in read_orc (#12325) @vuule
• CSV, JSON reader to infer integer column with nulls as int64 instead of float64 (#12309) @karthikeyann
• Remove deprecated code for 23.02 (#12281) @vyasr
• Null element for parsing error in numeric types in JSON, CSV reader (#12272) @karthikeyann
• Purge non-empty nulls for superimpose_nulls and push_down_nulls (#12239) @ttnghia
• Rename cudf::structs::detail::superimpose_parent_nulls APIs (#12230) @ttnghia
• Remove JIT type names, refactor id_to_type. (#12158) @bdice
• Floor division uses integer division for integral arguments (#12131) @wence-

🐛 Bug Fixes

• Fix a mask data corruption in UDF (#12647) @galipremsagar
• pre-commit: Update isort version to 5.12.0 (#12645) @wence-
• tests: Skip cuInit tests if cuda-gdb is not found or not working (#12644) @wence-
• Revert regex program java APIs and tests (#12639) @cindyyuanjiang
• Fix leaks in ColumnVectorTest (#12625) @jlowe
• Handle when spillable buffers own each other (#12607) @madsbk
• Fix incorrect null counts for sliced columns in JCudfSerialization (#12589) @jlowe
• lists: Transfer dtypes correctly through list.get (#12586) @wence-
• timedelta: Don't go via float intermediates for floordiv (#12585) @wence-
• Fixing BUG, get_next_chunk() should use the blocking function device_read() (#12584) @madsbk
• Make JNI QuoteStyle accessible outside ai.rapids.cudf (#12572) @mythrocks
• partition_by_hash(): support index (#12554) @madsbk
• Mixed Join benchmark bug due to wrong conditional column (#12553) @divyegala
• Update List Lexicographical Comparator (#12538) @divyegala
• Dynamically read PTX version (#12534) @brandon-b-miller
• build.sh switch to use RAPIDS magic value (#12525) @robertmaynard
• Loosen runtime arrow pinning (#12522) @vyasr
• Enable metadata transfer for complex types in transpose (#12491) @galipremsagar
• Fix issues with parquet chunked reader (#12488) @nvdbaranec
• Fix missing metadata transfer in concat for ListColumn (#12487) @galipremsagar
• Rename libcudf substring source files to slice (#12484) @davidwendt
• Fix compile issue with arrow 10 (#12465) @ttnghia
• Fix List offsets bug in mixed type list column in nested JSON reader (#12447) @karthikeyann
• Fix xfail incompatibilities (#12423) @vyasr
• Fix bug in Parquet column index encoding (#12404) @etseidl
• When building Arrow shared look for a shared OpenSSL (#12396) @robertmaynard
• Fix get_json_object to return empty column on empty input (#12384) @davidwendt
• Pin arrow 9 in testing dependencies to prevent conda solve issues (#12377) @vyasr
• Fix reductions any/all return value for empty input (#12374) @davidwendt
• Fix debug compile errors in parquet.hpp (#12372) @davidwendt
• Purge non-empty nulls in cudf::make_lists_column (#12370) @ttnghia
• Use correct memory resource in io::make_column (#12364) @vyasr
• Add code to detect possible malformed page data in parquet files. (#12360) @nvdbaranec
• Fail loudly to avoid data corruption with unsupported input in read_orc (#12325) @vuule
• Fix NumericPairIteratorTest for float values (#12306) @davidwendt
• Fixes memory allocation in nested JSON tokenizer (#12300) @elstehle
• Reconstruct dtypes correctly for list aggs of struct columns (#12290) @wence-
• Fix regex \A and \Z to strictly match string begin/end (#12282) @davidwendt
• Fix compile issue in json_chunked_reader.cpp (#12280) @ttnghia
• Change reductions any/all to return valid values for empty input (#12279) @davidwendt
• Only exclude join keys that are indices from key columns (#12271) @wence-
• Fix spill to device limit (#12252) @madsbk
• Correct behaviour of sort in concat for singleton concatenations (#12247) @wence-
• Purge non-empty nulls for superimpose_nulls and push_down_nulls (#12239) @ttnghia
• Patch CUB DeviceSegmentedSort and remove workaround (#12234) @davidwendt
• Fix memory leak in udf_string::assign(&&) function (#12206) @davidwendt
• Workaround thrust-copy-if limit in json get_tree_representation (#12190) @davidwendt
• Fix page size calculation in Parquet writer (#12182) @etseidl
• Add cudf::detail::sizes_to_offsets_iterator to allow checking overflow in offsets (#12180) @davidwendt
• Workaround thrust-copy-if limit in wordpiece-tokenizer (#12168) @davidwendt
• Floor division uses integer division for integral arguments (#12131) @wence-

📖 Documentation

• Fix link to NVTX (#12598) @sameerz
• Include missing groupby functions in documentation (#12580) @quasiben
• Fix documentation author (#12527) @bdice
• Update libcudf reduction docs for casting output types (#12526) @davidwendt
• Add JSON reader page in user guide (#12499) @GregoryKimball
• Link unsupported iteration API docstrings (#12482) @galipremsagar
• strings_udf doc update (#12469) @brandon-b-miller
• Update cudf_assert docs with correct NDEBUG behavior (#12464) @robertmaynard
• Update pre-commit hooks guide (#12395) @bdice
• Update test docs to not use detail comparison utilities (#12332) @PointKernel
• Fix doxygen description for regex_program::compute_working_memory_size (#12329) @davidwendt
• Add eval to docs. (#12322) @vyasr
• Turn on xfail_strict=true (#12244) @wence-
• Update 10 minutes to cuDF (#12114) @wence-

🚀 New Features

• Use kvikIO as the default IO backend (#12574) @vuule
• Use has_nonempty_nulls instead of may_contain_non_empty_nulls in superimpose_nulls and push_down_nulls (#12560) @ttnghia
• Add strings methods removeprefix and removesuffix (#12557) @davidwendt
• Add regex_program java APIs and unit tests (#12548) @cindyyuanjiang
• Default cudf::io::read_json to nested JSON parser (#12544) @vuule
• Make string quoting optional on CSV write (#12539) @mythrocks
• Use new nvCOMP API to optimize the compression temp memory size (#12533) @vuule
• Support "values" orient (array of arrays) in Nested JSON reader (#12498) @karthikeyann
• one_hot_encode to use experimental row comparators (#12478) @divyegala
• Support %W and %w format specifiers in cudf::strings::to_timestamps (#12475) @davidwendt
• Add JSON Writer (#12474) @karthikeyann
• Refactor thrust_copy_if into cudf::detail::copy_if_safe (#12455) @ttnghia
• Add trailing comma support for nested JSON reader (#12448) @karthikeyann
• Extract tokenize_json.hpp detail header from src/io/json/nested_json.hpp (#12432) @ttnghia
• JNI bindings to write CSV (#12425) @mythrocks
• Nested JSON depth benchmark (#12371) @karthikeyann
• Implement lists::reverse (#12336) @ttnghia
• Use device_read in experimental read_json (#12314) @vuule
• Implement JNI for strings::reverse (#12283) @ttnghia
• Null element for parsing error in numeric types in JSON, CSV reader (#12272) @karthikeyann
• Add cudf::strings:like function with multiple patterns (#12269) @davidwendt
• Add environment variable to control host memory allocation in hostdevice_vector (#12251) @vuule
• Add cudf::strings::reverse function (#12227) @davidwendt
• Selectively use dictionary encoding in Parquet writer (#12211) @etseidl
• Support replace in strings_udf (#12207) @brandon-b-miller
• Add support to read binary encoded decimals in parquet (#12205) @PointKernel
• Support regex EOL where the string ends with a new-line character (#12181) @davidwendt
• Updating stream_compaction/unique to use new row comparators (#12159) @divyegala
• Add device buffer datasource (#12024) @PointKernel
• Implement groupby apply with JIT (#11452) @bwyogatama

🛠️ Improvements

• Update shared workflow branches (#12696) @ajschmidt8
• Pin dask and distributed for release (#12695) @galipremsagar
• Don't upload libcudf-example to Anaconda.org (#12671) @ajschmidt8
• Pin wheel dependencies to same RAPIDS release (#12659) @sevagh
• Use CTK 118/cp310 branch of wheel workflows (#12602) @sevagh
• Change ways to access ptr in Buffer (#12587) @galipremsagar
• Version a parquet writer xfail (#12579) @galipremsagar
• Remove column names (#12578) @vuule
• Parquet reader optimization to address V100 regression. (#12577) @nvdbaranec
• Add support for category dtypes in CSV reader (#12571) @galipremsagar
• Remove spill_lock parameter from SpillableBuffer.get_ptr() (#12564) @madsbk
• Optimize cudf::make_lists_column (#12547) @ttnghia
• Remove cudf::strings::repeat_strings_output_sizes from Java and JNI (#12546) @ttnghia
• Test that cuInit is not called when RAPIDS_NO_INITIALIZE is set (#12545) @wence-
• Rework repeat_strings to use sizes-to-offsets utility (#12543) @davidwendt
• Replace exclusive_scan with sizes_to_offsets in cudf::lists::sequences (#12541) @davidwendt
• Rework nvtext::ngrams_tokenize to use sizes-to-offsets utility (#12540) @davidwendt
• Fix binary-ops gtests coded in namespace cudf::test (#12536) @davidwendt
• More @acquire_spill_lock() and as_buffer(..., exposed=False) (#12535) @madsbk
• Guard CUDA runtime APIs with error checking (#12531) @PointKernel
• Update TODOs from issue 10432. (#12528) @bdice
• Update rapids-cmake definitions version in GitHub Actions style checks. (#12511) @bdice
• Switch engine=cudf to the new JSON reader (#12509) @galipremsagar
• Fix SUM/MEAN aggregation type support. (#12503) @bdice
• Stop using pandas._testing (#12492) @vyasr
• Fix ROLLING_TEST gtests coded in namespace cudf::test (#12490) @davidwendt
• Fix erroneously skipped ORC ZSTD test (#12486) @vuule
• Rework nvtext::generate_character_ngrams to use make_strings_children (#12480) @davidwendt
• Raise warnings as errors in the test suite (#12468) @vyasr
• Remove int32 hard-coding in python (#12467) @galipremsagar
• Use cudaMemcpyDefault. (#12466) @bdice
• Update workflows for nightly tests (#12462) @ajschmidt8
• Build CUDA 11.8 and Python 3.10 Packages (#12457) @ajschmidt8
• JNI build image default as cuda11.8 (#12441) @pxLi
• Re-enable Recently Updated Check (#12435) @ajschmidt8
• Rework remaining cudf::strings::from_xyz functions to use make_strings_children (#12434) @vuule
• Build wheels alongside conda CI (#12427) @sevagh
• Remove arguments for checking exception messages in Python (#12424) @vyasr
• Clean up cuco usage (#12421) @PointKernel
• Fix warnings in remaining modules (#12406) @vyasr
• Update ops-bot.yaml (#12402) @ajschmidt8
• Rework cudf::strings::integers_to_ipv4 to use make_strings_children utility (#12401) @davidwendt
• Use numpy.empty() instead of bytearray to allocate host memory for spilling (#12399) @madsbk
• Deprecate chunksize from dask_cudf.read_csv (#12394) @rjzamora
• Expose the RMM pool size in JNI (#12390) @revans2
• Fix COPYING_TEST: gtests coded in namespace cudf::test (#12387) @davidwendt
• Rework cudf::strings::url_encode to use make_strings_children utility (#12385) @davidwendt
• Use make_strings_children in parse_data nested json reader (#12382) @karthikeyann
• Fix warnings in test_datetime.py (#12381) @vyasr
• Mixed Join Benchmarks (#12375) @divyegala
• Fix warnings in dataframe.py (#12369) @vyasr
• Update conda recipes. (#12368) @bdice
• Use gpu-latest-1 runner tag (#12366) @bdice
• Rework cudf::strings::from_booleans to use make_strings_children (#12365) @vuule
• Fix warnings in test modules up to test_dataframe.py (#12355) @vyasr
• JSON column performance optimization - struct column nulls (#12354) @karthikeyann
• Accelerate stable-segmented-sort with CUB segmented sort (#12347) @davidwendt
• Add size check to make_offsets_child_column utility (#12345) @davidwendt
• Enable max compression ratio small block optimization for ZSTD (#12338) @vuule
• Fix warnings in test_monotonic.py (#12334) @vyasr
• Improve JSON column creation performance (list offsets) (#12330) @karthikeyann
• Upgrade to arrow-10.0.1 (#12327) @galipremsagar
• Fix warnings in test_orc.py (#12326) @vyasr
• Fix warnings in test_groupby.py (#12324) @vyasr
• Fix test_notebooks.sh (#12323) @ajschmidt8
• Fix transform gtests coded in namespace cudf::test (#12321) @davidwendt
• Fix check_style.sh script (#12320) @ajschmidt8
• Rework cudf::strings::from_timestamps to use make_strings_children (#12317) @davidwendt
• Fix warnings in test_index.py (#12313) @vyasr
• Fix warnings in test_multiindex.py (#12310) @vyasr
• CSV, JSON reader to infer integer column with nulls as int64 instead of float64 (#12309) @karthikeyann
• Fix warnings in test_indexing.py (#12305) @vyasr
• Fix warnings in test_joining.py (#12304) @vyasr
• Unpin dask and distributed for development (#12302) @galipremsagar
• Re-enable sccache for Jenkins builds (#12297) @ajschmidt8
• Define needs for pr-builder workflow. (#12296) @bdice
• Forward merge 22.12 into 23.02 (#12294) @vyasr
• Fix warnings in test_stats.py (#12293) @vyasr
• Fix table gtests coded in namespace cudf::test (#12292) @davidwendt
• Change cython for regex calls to use cudf::strings::regex_program (#12289) @davidwendt
• Improved error reporting when reading multiple JSON files (#12285) @vuule
• Deprecate Frame.sum_of_squares (#12284) @vyasr
• Remove deprecated code for 23.02 (#12281) @vyasr
• Clean up handling of max_page_size_bytes in Parquet writer (#12277) @etseidl
• Fix replace gtests coded in namespace cudf::test (#12270) @davidwendt
• Add pandas nullable type support in Index.to_pandas (#12268) @galipremsagar
• Rework nvtext::detokenize to use indexalator for row indices (#12267) @davidwendt
• Fix reduction gtests coded in namespace cudf::test (#12257) @davidwendt
• Remove default parameters from cudf::detail::sort function declarations (#12254) @davidwendt
• Add duplicated support for Series, DataFrame and Index (#12246) @galipremsagar
• Replace column/table test utilities with macros (#12242) @PointKernel
• Rework cudf::strings::pad and zfill to use make_strings_children (#12238) @davidwendt
• Fix sort gtests coded in namespace cudf::test (#12237) @davidwendt
• Wrapping concat and file writes in @acquire_spill_lock() (#12232) @madsbk
• Rename cudf::structs::detail::superimpose_parent_nulls APIs (#12230) @ttnghia
• Cover parsing to decimal types in read_json tests (#12229) @vuule
• Spill Statistics (#12223) @madsbk
• Use CUDF_JNI_ENABLE_PROFILING to conditionally enable profiling support. (#12221) @bdice
• Clean up of test_spilling.py (#12220) @madsbk
• Simplify repetitive boolean logic (#12218) @vuule
• Add Series.hasnans and Index.hasnans (#12214) @galipremsagar
• Add cudf::strings:udf::replace function (#12210) @davidwendt
• Adds in new java APIs for appending byte arrays to host columnar data (#12208) @revans2
• Remove Python dependencies from Java CI. (#12193) @bdice
• Fix null order in sort-based groupby and improve groupby tests (#12191) @divyegala
• Move strings children functions from cudf/strings/detail/utilities.cuh to new header (#12185) @davidwendt
• Clean up existing JNI scalar to column code (#12173) @revans2
• Remove JIT type names, refactor id_to_type. (#12158) @bdice
• Update JNI version to 23.02.0-SNAPSHOT (#12129) @pxLi
• Minor refactor of cpp/src/io/parquet/page_data.cu (#12126) @etseidl
• Add codespell as a linter (#12097) @benfred
• Enable specifying exceptions in error macros (#12078) @vyasr
• Move _label_encoding from Series to Column (#12040) @shwina
• Add GitHub Actions Workflows (#12002) @ajschmidt8
• Consolidate dask-cudf groupby_agg calls in one place (#10835) @charlesbluca