codebrowse urls are long and confusing

[Michael Hudson-Doyle]

I'd like to work on letting pydoctor, my API doc generation tool, link to source as presented by codebrowse (it already has support for linking to trac's source viewer). Currently the codebrowse urls contain file-ids, which I could dig out of the wc being generated from, but well, it's not nice :)

spiv said on IRC that he thought allowing "filename=..." as an alternative to "file-id=..." looked easy enough.

[James Henstridge (jamesh)]

I talked with Robert about this. The two changes we should probably make include:

1. use the filename as part of the URL. Rather than "/annotate/$REVID?file_id=$FILEID", use "/annotate/$REVID/$FILENAME".

2. allow some short hand to refer to the head revision ID. Robert suggested "head:", which would give a URL like "/annotate/head:/$FILENAME".

[Michael Hudson-Doyle (mwhudson)]

I'm not sure point 2) is especially relevant to my use case, as I think pydoctor documentation will be generated from a particular revision, so may as well link to a particular revision... point 1) seems nice, even just from a human interface point of view.

[Robey Pointer (robey)]

I originally used filename instead of fileid, but I think I had some issues looking up the correct file object from that. I guess if the filename was always pulled from the stated revision, it would work.

Hm, in fact, I just tried patching loggerhead to use James' point #1 above, and it seems to work fine. I'll check it in.

[Robey Pointer (robey)]

rev 126. point #2 was easy to do too.

[Michael Hudson-Doyle (mwhudson)]

This is in production too now.

[Michael Hudson-Doyle (mwhudson)]

Released in 1.2, though we can do better...