Investigate the possibility to port to UglifyJS 2
Probably after #3.
This is technically possible and not so hard to do, but the new output from UglifyJS2 is much more vebrose — this means that the runtime will consume more memory and will require to load larger files from the server when in pre-parsed mode.
We could investigate converting that into a shorter form suitable to use by qmlweb, though, but that will still slow down parsing.
Also, the parser itself became about two times larger (and I included only the required files).
Filed a patch to uglify-js@2 that will allow us to wrap tokenizer() method in the meantime: https://github.com/mishoo/UglifyJS2/pull/1113 (it's somewhy not exported in uglify-js@2, but was exported in uglify-js@1).
Update: merged.
uglify-js-2 branch has some initial work, mostly broken. It also requires the abovementioned UglifyJS2 patch.