sum & mean group by operation not optimised for factor types

Open
#2,458 9 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
38/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Stale
Tech stack
r
Domain
data, performance

Research direction

Start by reproducing the integer-versus-factor group-by benchmarks linked in the issue, then inspect the source paths implementing grouped sum and mean. Compare the two cases and use the benchmark results to verify that factor grouping receives the intended optimization.

Written by the indexing model from the issue text.

Description

performance

Seems like group by are not optimised where the by-variable is of factor type. Essentially one can perform group-by faster for factors given that it's represented by integers from 1 to n where n is the number of groups.

See Python implementation discussion http://wesmckinney.com/blog/mastering-high-performance-data-algorithms-i-group-by/
and Julia implementation discussion https://www.codementor.io/zhuojiadai/an-empirical-study-of-group-by-strategies-in-julia-dagnosell

Update - Benchmarks

Here are some benchmarks to show that the factors as group-by doesn't seem to receive extra optimisation.

So I ran two benchmarks one for group-by variable with 2.5million groups, and one for 100millions groups. I ran one where the group-by variable is an integer and one where the group-by variable is a factor. In both cases the run time did not get faster in the group by variable is factor case. One can probably confirm by looking into the source code that sum group-by a factor is not optimised for.

Dominant language
R
Stars
3.9k
Forks
1.1k
Avg merge
14h 4m
Merged PRs (30d)
4

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from Rdatatable/data.table

All issues in Rdatatable/data.table

Similar issues

More R issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.