[jbearak]

Summary

A planned arithmetic tree of a grouped mutate() or filter() takes cumsum(), cummax() and cummin() of a Stata numeric operand, as ADR 0085 allows. Of a comparison, such as the running count cumsum(b5 == 1), the tree went to the loop. The comparison gives a bare logical vector, and each group's evaluation called Ops.dta_numeric() and base's cumsum(). Over the 42,739 women in the first 100,000 rows of the India 2021 DHS births file, mutate(b, c = cumsum(b5 == 1), .by = caseid) took about 0.30 seconds, where a tibble takes about 0.060.

This PR lets the tree take cumsum(), cummax() and cummin() of a logical tree, which gives a bare integer vector for each group. The call must find base's primitive, which dispatches on no bare logical vector. The native evaluator computes the running count, maximum or minimum of the logical values as integers, missing from the first missing value on, as the primitive does. A logical vector no longer than INT_MAX cannot overflow cumsum()'s integer.

The result may be the whole expression of a mutate(), or compare with a bare number, as the positions from row_number() do under ADR 0099. The comparison gives a logical tree that mutate() and filter() take.

Behavior

Results are unchanged. ADR 0123 records the decision. Arithmetic of the counts, a summary of them, and abs() or the rounding functions of a logical vector, which give other types, still go to the loop.

Timing

Best of three runs in fresh R sessions, in seconds, with the builds interleaved, over the first 100,000 rows of the India 2021 DHS births file. Before is main.

Call	Before	After	Tibble
mutate(b, c = cumsum(b5 == 1), .by = caseid)	0.30	0.050	0.060
filter(b, cumsum(b5 == 0) <= 1, .by = caseid)	0.43	0.059	0.059

Tests

• A test runs each of the three functions of a comparison of each column of the plan tests' data, by .by and by group_by(). The columns' missing and tagged missing values make the rest of a group's results missing. It also runs other logical trees, nested counts, comparisons of the counts with a number in mutate() and filter(), and between() as the tree. Each case must give what the loop gives, and the plan must take every one.
• A test checks that the plan leaves these to the loop: another function of the name where R finds it first, a column named cumsum, arithmetic of the counts, abs() of a comparison, cumsum(n() > 1), a summarise() and rowwise data.
• The full suite passes: 169 files, 2,278 tests. The native manifest and repository guard tests pass. Codex review raised only a stale manifest, which the refresh fixed.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: jbearak/dta-parser
• Branch: claude/cumulative-logical
• File: .coderabbit.yaml
• Commit: 041ca2a
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Team
• Run ID: 2433e3ef-908c-4a94-91a7-c0f1b1a9dfe2

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}