disable Terser's "reduce_funcs" option for performance

[kfule]

Terser's “reduce_funcs” option seems to degrade performance. So, disable it. This commit will increase file size of "mithril.min.js" slightly, but performance appears to improve.

Description

This PR makes mithril.js and mithril.min.js code equivalent (except mangling) by disabling Terser's “compress” option.
This will increase the file size of “mithirl.min.js” by about 0.01KB (gzipped), but seems to improve performance.
Edit: The code in mithril.js and mithril.min.js is not equivalent because only reduce_funcs is disabled.

results of npm run perf (nodejs, using mithril.min.js)

• before

construct large vnode tree x 21,998 ops/sec ±0.49% (127 runs sampled)
rerender identical vnode x 3,876,899 ops/sec ±0.47% (122 runs sampled)
rerender same tree x 83,591 ops/sec ±0.19% (126 runs sampled)
add large nested tree x 6,869 ops/sec ±1.55% (122 runs sampled)
mutate styles/properties x 123 ops/sec ±1.29% (91 runs sampled)
repeated add/removal x 3,007 ops/sec ±0.69% (124 runs sampled)
Completed perf tests in 56568ms

• after (improves "rerender same tree" and "add large nested tree")

construct large vnode tree x 20,641 ops/sec ±0.51% (128 runs sampled)
rerender identical vnode x 4,029,056 ops/sec ±0.56% (129 runs sampled)
rerender same tree x 90,134 ops/sec ±0.25% (125 runs sampled)
add large nested tree x 7,452 ops/sec ±0.84% (123 runs sampled)
mutate styles/properties x 130 ops/sec ±1.41% (93 runs sampled)
repeated add/removal x 3,047 ops/sec ±0.34% (128 runs sampled)
Completed perf tests in 57709ms

Motivation and Context

While measuring performance under various conditions, I noticed the performance was different between "mithril.min.js" and "mithril.js" ("mithril.min.js" seems slower).
Looking into the cause, I found that compressing the code with Terser's “compress” option seemed to affect performance as well as it reduced the amount of code.

I use mithril.min.js as is for prototyping, so I prefer mithril.min.js to be faster.

How Has This Been Tested?

npm run test

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist

• My code follows the code style of this project.
• I have added tests to cover my changes.
• All new and existing tests passed.
• My change requires a documentation update, and I've opened a pull request to update it already:
• I have read https://mithril.js.org/contributing.html.

[dead-claudia]

I'm 90% sure it's related to the fact Terser inlines larger functions by creating IIFEs it doesn't proceed to flatten.

Better would be finding the offending function(s) and either inlining it manually (if it's easy) or marking it with /*@__NOINLINE_*/ (if it's complicated).

This isn't a merge blocker, though.

[dead-claudia]

Could you see if just disabling reduce_funcs helps, and keeping the rest of compression on? terser/terser#1305 among others are related.

[kfule]

I tried the following setting where only reduce_funcs is false, but the output result seems to be the same as in v2.2.11...

const minified = Terser.minify(original, {compress: {reduce_funcs: false}, mangle: true})

[kfule]

It seems that a newer version of terser works well with reduce_funcs: false and produces output that is both small file size and performance. However, installing a newer version of terser seems to install packages that do not support node v18.

[kfule]

@dead-claudia Thank you for your review comments.
In my opinion, I don't want to make changes to the mithril code itself because of terser's behavior.
So I want to improve only terser options. But it seems that the version of terser needs to be upgraded for reduce_funcs to work well, and it may not support node 18...
(It seemed that reduce_funcs, if it worked, could be a good solution for both file size and performance...)

[kfule]

@dead-claudia
Very sorry, the node 18 issue was not related to the Terser's upgrade.
So, I upgraded Terser to make reduce_funcs work well. This allowed better performance with little increase in file size (0.01KB gzipped).