rapidsai/cudf

[FEA] Support more dtypes in JIT GroupBy `apply`

Open

#12,608 opened on Jan 25, 2023

View on GitHub
 (6 comments) (0 reactions) (0 assignees)C++ (735 forks)batch import
Pythonfeature requestgood first issuenumba

Repository metrics

Stars
 (6,000 stars)
PR merge metrics
 (Avg merge 17d 21h) (230 merged PRs in 30d)

Description

Is your feature request related to a problem? Please describe. When https://github.com/rapidsai/cudf/pull/11452 lands, we'll get JIT Groupby.apply for a subset of UDFs and importantly, dtypes. However over the summer we only got as far as writing overloads for float64 and int64 dtypes in the users source data. It'd be nice if we could support more dtypes, starting at least with the rest of the numeric types.

Describe the solution you'd like Extend the existing groupby.apply, engine='jit' framework to support the following additional dtypes:

  • float32
  • int32
  • int16
  • int8
  • uint64
  • uint32
  • uint16
  • uint8
  • bool

A lot of the machinery in the original PR is fairly general and should make adding many of these easy- however there will undoubtedly be edge cases. As such it makes for a pretty good first issue for anyone jumping into the numba extension piece of the codebase.

Contributor guide

Research direction
Review the existing JIT GroupBy apply implementation in PR #11452 and extend the dtype overloads to include the listed types. Focus on the numba extension code and ensure compatibility for each dtype.
Tech stack
pythoncpp
Domain
backenddata
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
2
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
Python
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
80