Skip to content

[WIP] Parquet Go ALP implementation support#704

Open
ArnavBalyan wants to merge 1 commit intoapache:mainfrom
ArnavBalyan:arnavb/alp-go-impl
Open

[WIP] Parquet Go ALP implementation support#704
ArnavBalyan wants to merge 1 commit intoapache:mainfrom
ArnavBalyan:arnavb/alp-go-impl

Conversation

@ArnavBalyan
Copy link
Member

cc @julienledem @alamb

Rationale for this change

  • Implements ALP (Adaptive Lossless floating point) encoding for float and double columns as per Prateek's spec. Related to GH-533: Adaptive Lossless Floating-Point (ALP) Encoding parquet-format#548.
  • ALP converts floating point values to integers via decimal scaling, then applies For and bit packing. Values that don't round trip exactly are stored as exceptions.
  • The encoder is incremental and flushes per vector as values arrive. The decoder is lazy with vectors decoded on demand via offset array.

What changes are included in this PR?

  • Wired ALP implementation through encoder.go and decoder.go based on the original spec.
  • Registered a new encoding as Encodings.ALP.
  • Added unit tests/cross compat tests

Are these changes tested?

  • Yes new unit tests were added, all of which pass in addition to the existing tests.
  • Cross compat tests were added, the external encoded file can be provided as env var which triggers arrow-go ALP decoding.

Are there any user-facing changes?

  • Yes

@ArnavBalyan ArnavBalyan requested a review from zeroshade as a code owner March 11, 2026 13:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant