[inlinedocs] slow package.skeleton.dx(), probably because of extract.file.parse()

[Julien Moeys]

Hi all inlinedocs developers

I tested the latest inlinedocs code on another package of mine (54 functions), and it appears that the execution of package.skeleton.dx() is really slow (nearly 6 minutes), and much slower than it was before (with version 1.3) on the same code.

It seems that 90% of the time is spend on extract.file.parse() in extra.code.docs() and it seems that this time is spread homogeneously between the different functions in the package. So I don't think that one of my function in the code is causing any slow down.

I tried to see if there was some potential speedup in extract.file.parse() (like getting rid of the for loop), but it seems that the part concerning "setS3Method" needs to be able to look into previous results in the loop, so it will not work with lapply (or it would require to split the loop in to pieces, a first round, and then a second). I am not sure of the best way it could be accelerated (and I am not sure I have correctly identified the problem!).

Any suggestion?

All the best

Julien 
PS: I am using the latest inlinedocs commit, 162, compiled on with computer (Win XP, R 2.12)
PS2: I can send an output from Rprof() if that can help.
PS3: The problem is not critical, so I am not in a hurry...