faster url parser

[jonathanong]

https://github.com/petkaantonov/urlparser

@petkaantonov said you guys won't merge this in core, but i don't see a reason not to, especially when you guys are/will be focusing on performance improvements, there are no regressions, there are more tests, and the code is easier to read.

last thing i want are PRs for faster node modules (and the consequential maintenance burden) in every repo when node should be improving performance themselves.

related: senchalabs/connect#986

[stefanpenner]

Getting this in core would be ideal. For future reference I am quite interesting in knowing what (if any) the blockers and concerns turn out to be.

[tjfontaine]

To quote my original senchalabs/connect#986 (comment)

For posterity, at the moment I am not interested in doing a substantial rewrite of a core module before 0.12 or for 1.0 which will be coming immediately afterward. That doesn't mean another team member won't want to champion it.

It is a scary proposition for a relatively new module to replace particularly important functionality. That's why I advocate leaving it as an external module for now, where people can opt into using it and it can grow and mature on its own without node standing in its way.

As it grows in popularity and need, it can be integrated into node (as other modules have transitioned historically). Or such time that a core team member has the time, motivation, and priority to give the module the attention it deserves, and integrate it that way.

In the meantime know that we aren't unsympathetic to your needs, but it is certainly possible to work around that for now without requiring a change to node proper.

[jonathanong]

okay cool. thanks for putting it on a milestone.

here's a hacky PR to replace node's url parser globally: petkaantonov/urlparser#2

[jbergstroem]

How about baking a patch that lands additional tests for starters?

[jonathanong]

ask @petkaantonov. i'm not going to make a PR unless i know someone's going to look at it. there's already 169 PRs that have yet to be merged or rejected. i'm not going to add another one.

[jasnell]

Given the excellent work that @petkaantonov has been doing on the io.js side, significant performance improvements will be landing in the converged stream. We can decide whether or not we want to backport those here once that work is complete.