Arrow 0.3 release

[asfimport]

As we near the next development milestone, please link issues that block the release so we can keep track of what needs to be done

Reporter: Wes McKinney / @wesm
Assignee: Wes McKinney / @wesm

Related issues:

• [Python] Unit tests in arrow/python fail to link on some OS X platforms (is blocked by)
• [GLib] Glib sources missing from result of 02-source.sh (is blocked by)
• [Python] Review GIL release and acquisition in to_pandas conversion (is blocked by)
• [Python] Support for converting from strided pandas data in Table.from_pandas (is blocked by)
• [Format] Update IPC.md to account for File format changes and Streaming format (is blocked by)
• [C++] IPC unloading for BooleanArray does not account for offset (is blocked by)
• [Python] Convert nanosecond timestamps to pandas.Timestamp when converting from TimestampValue (is blocked by)
• Python: Explicit pandas import makes it a hard requirement (is blocked by)
• [java] arrow should not have a dependency on slf4j bridges in compile (is blocked by)
• Arrow freezes on write if chunk_size=0 (is blocked by)
• [C++] PyBuffer dtor may segfault if constructor passed an object not exporting buffer protocol (is blocked by)
• [Python] Mixed-type object DataFrame columns should not silently coerce to an Arrow type by default (is blocked by)
• [Python] Fix manylinux1 packaging (is blocked by)
• [C++] Fix spurious warning caused by passing dl to add_dependencies (is blocked by)
• [Python] Delete incomplete binary files when writing fails (is blocked by)
• [C++] Do not pass -fPIC when compiling with MSVC (is blocked by)
• [C++] MSVC build fails on googletest when using NMake (is blocked by)
• Kerberos Problem with PyArrow (is blocked by)
• [Java] Resolve getObject behavior per changes / discussion in ARROW-729 (is blocked by)
• Integration tests fail for length-0 record batch (is blocked by)
• [C++] Check whether data is contiguous in ipc::WriteTensor (is blocked by)
• [Python] Add updated pyarrow.* public API listing in Sphinx docs (is blocked by)
• listing empty HDFS directory returns an error instead of returning empty list (is blocked by)
• [Python] StreamWriter fails to open with socket as sink (is blocked by)
• Compilation error on Mac with -DARROW_PYTHON=on (is blocked by)
• Python: Parquet: Dictionary encoding is deactivated if column-wise compression was selected (is blocked by)
• Python: jemalloc is not anymore publicly exposed (is blocked by)
• [C++] Portable alternative to PyDate_to_ms function (is blocked by)
• [Python] It is no longer necessary to modify the RPATH of the Cython extensions on many environments (is blocked by)
• [Python] Error from file object destructor (is blocked by)
• [Python] Miscellaneous pyarrow MSVC fixes (is blocked by)
• VariableLengthVectors don't reAlloc offsets (is blocked by)
• [format] For backward compatibility, new unit fields must have default values matching previous implied unit (is blocked by)
• BitVector transfer() does not transfer ownership (is blocked by)
• [Python] UnboundLocalError in ParquetDatasetPiece (is blocked by)
• C++: Writing sliced record batch to IPC writes the entire array (is blocked by)
• [JAVA] Introduction of new types has shifted Enumerations (is blocked by)
• Problem with reading parquet files saved by Apache Spark (is blocked by)
• [Java] Checkstyle additions causing build failure in some environments (is blocked by)
• [Python/C++] Decimal test failure on MSVC (is blocked by)
• [C++/Python] Provide access to "custom_metadata" Field attribute in IPC setting (is blocked by)
• Add integration tests for date and time types (is blocked by)
• [Python] Support _metadata or _common_metadata files when reading Parquet directories (is blocked by)
• [Python] Support reading Parquet datasets with standard partition directory schemes (is blocked by)
• [C++] Add RecordBatch::Validate method that at least checks that schema matches the array metadata (is blocked by)
• [C++] Remove -g flag in release builds (is blocked by)
• [Python] Add support for FixedWidthBinary type (is blocked by)
• [C++] Implement IO interface that can read and write to a fixed-size mutable buffer (is blocked by)
• [Python] Write and read tensors (with zero copy) into shared memory (is blocked by)
• [Python] Attach proper tzinfo when computing boxed scalars for TimestampArray (is blocked by)
• [C++] Build and run full test suite in Appveyor (is blocked by)
• [C++] Add options to StreamWriter/FileWriter to permit large record batches (is blocked by)
• [Python] Enable Feather APIs to read and write using Python file-like objects (is blocked by)
• [C++] IPC zero-copy round trips for arrow::Tensor (is blocked by)
• [Python] Expose arrow::Tensor with conversions to/from NumPy arrays (is blocked by)
• [Python] pandas conversions for new date and time types/metadata (is blocked by)
• [Java] Add vector type for 32-bit date as days since UNIX epoch (is blocked by)
• [C++/Format] Change name of Fixed Width Binary to Fixed Size Binary for consistency (is blocked by)
• [Python] Support for pyarrow on Windows / MSVC (is blocked by)
• [C++] Support obtaining mutable slices of mutable buffers (is blocked by)
• [Java] Change the "boxed" object representation of date and time types (is blocked by)
• [C++] Add function to create arrow::Table with column appended to existing table (is blocked by)
• [Python] Verify Parquet roundtrips for new date/time types (is blocked by)
• Cache miniconda packages (is blocked by)
• Implement DecimalArray (is blocked by)
• Python: All none-Pandas column should be converted to NullArray (is blocked by)
• Add "How to Contribute" section to README (is blocked by)
• [Python] Write memoryview-compatible objects in NativeFile.write with zero copy (is blocked by)
• [Python] Add Python 3.6 to Travis CI (is blocked by)
• [Python] Rename all Cython extensions to "private" status with leading underscore (is blocked by)
• [Python] Construct pyarrow.DictionaryArray from boxed pyarrow array objects (is blocked by)
• [C++] Fix compiler warnings on MSVC x64 (is blocked by)
• [Python] Add function to compute the total size of tensor payloads, including metadata and padding (is blocked by)
• [Python] Make generic ArrowException subclass value error (is blocked by)
• [C++/Python] Raise exception if old metadata encountered (is blocked by)
• [C++] Change struct to class for objects that meet the criteria in the Google style guide (is blocked by)
• [C++] Combine libarrow/libarrow_io/libarrow_ipc (is blocked by)
• [Java] Allow for expanding underlying buffer size after allocation (is blocked by)
• [Python] Review public pyarrow.* API completeness and update docs (is blocked by)
• [Python] Generalize pyarrow.from_pylist to accept any object implementing the PySequence protocol (is blocked by)
• [Python] Variety of Parquet improvements to support Dask integration (is blocked by)
• [C++] Upgrade thirdparty gtest to 1.8.0 (is blocked by)
• [Python] Add pyarrow build to Appveyor (is blocked by)
• [Format] Revise format/README.md to reflect progress reaching a more complete specification (is blocked by)
• [C++] Add optional $ARROW_BUILD_TOOLCHAIN environment variable option for configuring build environment (is blocked by)
• [Python] Automate publishing Python documentation to arrow-site (is blocked by)
• [C++] Decide if typed Tensor subclasses are worthwhile (is blocked by)
• [Python] Move DEVELOPMENT.md to Sphinx docs (is blocked by)
• [Python] Improve source build instructions in README (is blocked by)
• [C++] Expand metadata support to field level, provide for sharing instances of KeyValueMetadata (is blocked by)
• [C++] Write 0-length EOS indicator at end of stream (is blocked by)
• [Format] Constant length list type (is blocked by)
• [C++] Do not build unit tests by default? (is blocked by)
• Scaffold Main Documentation using asciidoc (is blocked by)
• Java: API documentation (is blocked by)
• [Format] Bump metadata version for 0.3 release (is blocked by)
• [Python] Pin runtime library versions in conda-forge packages to force upgrades (is blocked by)
• [Docs] Publish Format Markdown documents somehow on arrow.apache.org (is blocked by)

_{Note: This issue was originally created as ARROW-670. Please see the migration documentation for further details.}