Alex Lam S.L.

IMHO the numbers look rather unappealing as it stands - I'd be happier if the uglified gain is much larger than the gzip loss.

I think part of the issue here is with `mangle`, c.f. #2219 Introducing cross-scope variables for this feature probably won't do a lot of good, since it'd mess with variable...

@Skalman now that #2948 is in, please have another go at this and see if your previous strategies would be more compressible with the new `mangle`.

@Skalman as I mentioned in https://github.com/mishoo/UglifyJS2/pull/2945#issuecomment-367797185, no measurable performance difference is observed through JetStream. In general we need something like 10x slowdown across a majority number of platforms to justify...

I'm actually surprised the numbers don't improve with the latest `mangle` - I'll take another look at the changes in this PR after some sleep to see what I have...

Hmm - looks like the upper-level variables still don't like to be kicked around by those newly introduced `void 0` substitutes: ```diff --- a/lib/compress.js +++ b/lib/compress.js @@ -59,6 +59,7 @@...

@dabretin hope these help: http://gitready.com/advanced/2009/02/10/squashing-commits-with-rebase.html https://help.github.com/articles/about-git-rebase/

`node test/benchmark.js` is a good way to test if your changes make for smaller (gzipped) output.

Simply put, the code base is monolithic − source files are separated the same way the Web used to parse them: ```html ``` Conversion to a modular approach would incur...

AFAICT this project started off before `require()` existed, so no questioning hysterical raisins here :smirk: Just importing (& maintaining) all the `AST_*` classes alone would be a major headache. But...