Novus Convert
Convert files
Data format

FEATHER converter

A .feather file is a fast on-disk table format for data frames. Since version 2 it is literally the Apache Arrow IPC file format under a different extension, so a modern .feather file and a modern .arrow file are the same bytes with different names.

Private for supported formats — processed in your browser

Convert supported files

Runs on your device

Drop FEATHER files here

Batch files can each use a different output. Nothing uploads for local conversions.

Working inputs include camera RAW, browser-local audio/video, PDF, CBZ/CBR comics, office documents, ebooks, markup, 3D models, structured text, images, and archives.

Where FEATHER comes from

Wes McKinney and Hadley Wickham created Feather in 2016 so that Python and R could exchange a data frame without a CSV round trip and without either language's native pickle or RDS format. Feather v1 had its own layout; in 2020 Feather v2 was redefined as the Arrow IPC file format, which retired a separate specification and gave Feather compression, nested types and every Arrow implementation for free. Files written before that change are v1 and need a reader that still handles it.

How FEATHER works

  • A v2 file starts with the ARROW1 magic; a v1 file starts with FEA1, which is how the two are told apart.
  • Columnar layout with the same 64-byte buffer padding and validity bitmaps as Arrow.
  • Optional LZ4 or ZSTD compression is applied per buffer, so individual columns stay independently readable.
  • Reads are effectively memory-mapped, so loading a column is a seek rather than a parse.
  • The schema, including nested and dictionary-encoded types, travels inside the file.

When to use FEATHER

  • Moving a data frame between R and Python at native speed
  • Persisting an analysis step so a notebook can be re-run without recomputing it
  • Sharing a typed table with a collaborator who uses a different language
  • Converting a stored data frame to CSV, JSON or a spreadsheet for a non-technical reader

Strengths and limitations

Strengths

  • Reads far faster than CSV and preserves exact types
  • Identical to Arrow IPC, so the whole Arrow ecosystem can open it
  • Optional per-column compression without giving up random access

Limitations

  • The v1/v2 split still surprises people opening old files
  • Larger than Parquet for archival storage and less widely supported by data warehouses
  • Nothing about the extension tells you which version you have without reading the magic

Compatibility

pyarrow, R's arrow package, Pandas, Polars, DuckDB and Julia all read Feather v2. Novus Convert handles .feather as Arrow IPC in the browser and converts to JSON, CSV, XLSX or .arrow. A v1 file will be rejected rather than silently misread, which is the correct outcome: guessing at a different layout would produce plausible nonsense.

Schema and column-level metadata are read from the Arrow footer; text and spreadsheet targets have no equivalent channel, so type information is reported in the conversion summary.

Convert FEATHER to another format

Convert other formats to FEATHER