adding support to provide custom parse opts and stop esprima from throwing on script level returns

[thlorenz]

This PR is in two parts.

I put them separately if for whatever reason you have a problem with the 2nd one. The most important is the first commit. However the current implementation has incorrect location information for the first line which is a problem if you want to use detective in order to replace requires instead of just finding them.

tolerant: true stops esprima from throwing and is a better solution IMO than wrapping the code inside a function.

1. adding support to provide custom parse opts
   • necessary if you want parser to include extra info like location info in the AST
2. esprima tolerant: true and removing wrapper
   • tolerant parsing does not throw on script level returns
   • enforcing this option for parse makes the function wrapper no longer needed
   • location info for first line is now correct (see test/parseopts.js)

     Note

0.6 currently fails on travis (certificate error) if you want I can change it to support 0.8 and 0.10 instead and add that to this PR.

[defunctzombie]

does tolerant change any other behavior other than script level returns?

LGTM as long as we understand any behavior changes.

[thlorenz]

Quoting from the docs:

tolerant: attempts to continue parsing when an error is encountered

So a more tolerant/stable parse is what we'd want right? All we are trying to do is find require calls, not validate JavaScript.

@ariya will know more, but as far as I can tell, it fixes the problems that required a function wrapper in the past and yes, mostly that's related to script level returns.

[defunctzombie]

This quote seems to imply that it will ignore all script errors. I guess that is reasonable for this module since it isn't actually interested in validating the JS. Tho maybe a script error could lead to improper detection of a require?

[thlorenz]

From what I understand, that's not how it works. It still has to understand the code. In this example it finds a problem/violation and will now throw or just add it to errors array if tolerant: true.

So all tolerant: true does is turn off throwing on ECMA spec violations, but if you have totally invalid code where it may improperly detect requires, it'll still blow up.

[thlorenz]

@substack would really like to see this land since I don't wanna publish browserify-shim while still depending on a github fork. Is there anything I should change so you can pull this?

[ghost]

merged in 2.3.0

[thlorenz]

Awesome, thanks!