Skip to main content

Apache Arrow/Feather Format

Description​

Apache Arrow is a columnar in-memory data format designed for efficient data transfer and analytics. Feather is a file format built on Arrow that provides fast, language-agnostic columnar storage. It's optimized for speed and is ideal for data exchange between Python, R, and other languages.

File Extensions​

  • .arrow - Apache Arrow files
  • .feather - Feather format files (alias for Arrow)

Implementation Details​

Reading​

The Arrow/Feather implementation:

  • Uses PyArrow library for reading
  • Loads entire table into memory
  • Supports batch reading for efficient iteration
  • Preserves data types and schema

Writing​

Writing support:

  • Buffers records before writing
  • Uses PyArrow Feather writer
  • Flushes data on close
  • Maintains schema from data

Key Features​

  • Columnar format: Efficient for analytical workloads
  • Fast I/O: Optimized for speed
  • Type preservation: Maintains data types
  • Cross-language: Compatible with R, Java, and other languages
  • Totals support: Can count total rows
  • Batch processing: Efficient batch reading

Usage​

from iterable import open_iterable

# Basic reading
with open_iterable('data.arrow') as source:
    for row in source:
        print(row)

# Writing
dest = open_iterable('output.feather', mode='w', iterableargs={
    'batch_size': 10000
})
dest.write({'id': 1, 'name': 'John'})
dest.close()  # Important: closes and flushes data

Parameters​

  • batch_size (int): Batch size for reading (default: 1024)

Limitations​

  1. PyArrow dependency: Requires pyarrow package
  2. Memory usage: Entire file is loaded into memory when reading
  3. Flat data only: Designed for tabular/flat data structures
  4. Write buffering: Must call close() to flush buffered data
  5. File path requirement: Feather writing may require file path rather than stream

Compression Support​

Arrow/Feather files can be compressed with all supported codecs:

  • GZip (.arrow.gz)
  • BZip2 (.arrow.bz2)
  • LZMA (.arrow.xz)
  • LZ4 (.arrow.lz4)
  • ZIP (.arrow.zip)
  • Brotli (.arrow.br)
  • ZStandard (.arrow.zst)

Performance Considerations​

  • Fast I/O: Feather format is optimized for speed
  • Memory: Entire file loaded into memory for reading
  • Batch size: Larger batches improve iteration performance

Use Cases​

  • Data exchange: Fast transfer between Python and R
  • Temporary storage: Intermediate format in data pipelines
  • Analytics: Fast columnar data access

Error Handling​

  • Missing dependency: optional libraries raise ImportError with an install hint (pip install 'iterabledata[<extra>]' when an extra exists).
  • Write mode: read-only formats raise WriteNotSupportedError or ValueError when opened with mode="w".
  • Bad or unsupported input: may raise ValueError, OSError, or library-specific errors.
  • See Troubleshooting for decoding, detection, and engine issues.
  • Parquet - Similar columnar format with compression
  • ORC - Another columnar format