yew/vlist: optimize diffing and patching

[bakape]

Description

• Optimized keyed VList patching for node addition or removal from either end of the list and localized changes in the middle
• Changed unkeyed VList patching algorithm to much cheaper reverse vector itteration
• Added opt-in more verbose test logging
• Improved Debug output of some types

Checklist:

• I have run ./ci/run_stable_checks.sh
• I have reviewed my own code
• I have added tests

• I have opened a PR against https://github.com/yewstack/js-framework-benchmark: Feature/opt keyed diff js-framework-benchmark#13

[bakape]

yew-functional CI tests seem to have failed due to chromedriver.

[jstarry]

@bakape perf looks good for 7b00d04e but not for 0c8a2715. Do you mind taking a look?

7b00d04e

result yew-05_swap1k.json min 49.149 max 52.446 mean 50.693000000000005 median 50.484 stddev 1.6584067655433616
result yew-baseline-05_swap1k.json min 49.25 max 52.606 mean 51.334 median 52.146 stddev 1.8193933054730091

0c8a2715

result yew-05_swap1k.json min 411.713 max 443.661 mean 429.0826666666667 median 431.874 stddev 16.155876093029836
result yew-baseline-05_swap1k.json min 51.105 max 60.054 mean 56.18866666666667 median 57.407 stddev 4.597216802950819

[jstarry]

Also, I fixed the macro tests on master fyi

[bakape]

The regression was introduced in e6c965b, which is very odd. I don't see anything there that could have caused it. I'll rebase and merge master anew.

[bakape]

@jstarry Sorry, I've been busy with other matters and projects for a while. I've fixed the swap issue and added a few other optimizations.

[jstarry]

No worries, I've been busy as well. Will review soon

[bakape]

@siku2
Merged master and reran benchmarks using bakape/js-framework-benchmark

@jstarry
yewstack/js-framework-benchmark is currently broken, so the action will fail. My PR should fix that.

[cecton]

I'm looking forward to see this merged \o/

[bakape]

It's been forgotten, it seems. A lot to review, so understandable finding time might be hard.

[teymour-aldridge]

Just some thoughts

[teymour-aldridge]

I'm not sure about the benefit of using a macro here (especially with an eye towards compile times).

[bakape]

It saves us from copy-pasting and I think the compile costs are negligible. Especially so, with how much the html! procedural macro does.

[ranile]

Apart from this comment, this looks good to me.

[ranile]

Nit: maybe also print the VText's content? Or is that not possible?

[bakape]

This was intentional, as it made the debug log output much more unreadable. Especially so in already verbose log output, like the one I added to the layout tests.