perf(es/minifier): Make analyzer not call collect_infects_from recursively

[kdy1]

Description:

We don't need to make time complexity non-linear by recursively visiting the effects of infection. Instead, suppose we stop the infection analyzer when the usage analyzer invokes the infection analyzer. In that case, we will have practically identical analysis results because the points are named using an id, and it points to all the infections that occurred by referencing the id.

[changeset-bot]

⚠️ No Changeset found

Latest commit: d21d068

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[kdy1]

cc @CPunisher Do you have an idea for better caching? *const Function does not look good to me.

[CPunisher]

cc @CPunisher Do you have an idea for better caching? *const Function does not look good to me.

The point is to determine an unique id for a function node. Maybe Ident, Span...Since we don't have node id, we may need to find the unique id based on the context.

BTW does it handle a node many times?

[kdy1]

@CPunisher Yeap. I wrote about them in #9927.

While visiting nodes for usage analysis, it calls collect_infects_from on the node, but while recursing into a node from the usage analyzer, it calls collect_infects_from on a child node again.

[CPunisher]

@CPunisher Yeap. I wrote about them in #9927.

While visiting nodes for usage analysis, it calls collect_infects_from on the node, but while recursing into a node from the usage analyzer, it calls collect_infects_from on a child node again.

I see. If it is hard to cache the result, can we avoid repeated analysis? That means we can stop collect_infects_from going deeper since we know we will handle inner nodes later.

[kdy1]

That's ideal, but I'm not sure about the way. I'm still investigating. Expected gain is so big but actually it's just an issue of my skill level.

[kdy1]

@CPunisher I think I managed to find something close to not recursively calling them. But I want to ensure that my assumption about usage analyzer, and actual code is correct.

[codspeed-hq]

CodSpeed Performance Report

Merging #9924 will degrade performances by 3.64%

_{Comparing kdy1:min-cache-infects (ec1460c) with main (6f781d8)}

Summary

⚡ 3 improvements
❌ 2 regressions
✅ 191 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	es/full/minify/libraries/d3	806.7 ms	779.9 ms	+3.44%
❌	es/full/minify/libraries/moment	125.3 ms	129.2 ms	-3.05%
⚡	es/full/minify/libraries/terser	634.2 ms	609.7 ms	+4.03%
⚡	es/minifier/libs/lodash	216.9 ms	205.6 ms	+5.47%
❌	es/minifier/libs/moment	100.6 ms	104.4 ms	-3.64%

[kdy1]

Diff: