JSDoc support

[RyanCavanaugh]

Adds support for adding type information in JavaScript files using jsdoc tags

[DanielRosenwasser]

Will we ever manifest this name to the user?

[RyanCavanaugh]

Yes

[ahejlsberg]

The parseJSDocOptionalType function in parser.ts is never called.

[DanielRosenwasser]

I'm not seeing where you skip this check.

[ahejlsberg]

Hmm, so an identifier can start with anything but the characters checked for above, but then after that must contain proper identifier characters. That seems a bit odd.

[ahejlsberg]

Should be parseJSDocIdentifier. The scanning is handled by the scanner.

[ahejlsberg]

Also, it should call nextJSDocToken(), similar to what parseIdentifier in the main parser does.

[billti]

This should never happen though right? As you've eaten the preceding white space, and there's nothing in here where you would consume white space inside a token.

[billti]

It seems a little fragile to use a method (setText - which calls setTextPos) to modify the state prior to callback execution, then manually restore each field individually after. If anyone added further state changes in setText (or setTextPos), they'd need to know to come and manually undo those same changes here. Can this set/restore be made more symmetrical/colocated somehow?

[RyanCavanaugh]

This is the same pattern we use elsewhere (e.g. in speculationHelper). I don't see any obvious improvement.

[ahejlsberg]

The skipSignatureCheck parameter never appears to be used.

[billti]

All these edits to server.ts look like changes I made in my branch to help with debugging Sublime but hadn't pushed to 'master' (and are part of bigger changes on both Sublime and tsserver to use TCP over stdin/stdout if configured to do so via TSS_DEBUG). Any idea how the changes in this file got in this code review?

[RyanCavanaugh]

Undone.

[ahejlsberg]

Made it all the way through. I think we're good once my comments are addressed.