#3 Parsed store persistence

[Johannes Berg]

http://sourceforge.net/mailarchive/forum.php?thread_id=2058513&forum_id=4647

requested there, will try implementing soon. Read my follow-up once it shows up on the archive.

[Johannes Berg]

Logged In: YES
user_id=32619

Some of this is implemented now, but not really the way
you'd want it. But it should be far easier now that the
saving framework is in place (need timestamp checking etc)

[Johannes Berg]

Logged In: YES
user_id=32619

This is probably never going to fly.

Too much stuff, and its not really that much faster.

I know this would be useful for shipping docs for inclusion,
but then again you'll probably need to ship interfaces files
anyway, so ...

[Michalis Kamburelis]

Logged In: YES
user_id=987895

Actually, it flies already. --cache-dir option was very
useful as you implemented it. And I think that I made it
even more useful by making cache independent of output format :)

Therefore it's tempting for me to just close this RFE,
because I consider cache as a very useful and correctly
working feature of pasdoc. Let me summarize here precisely
what is expected to be implemented to close this RFE.

[http://pasdoc.sipsolutions.net/CacheOption] gives an
overview of what is already done and working, so I will not
repeat it here.

What can be possibly done more:

1. Let the cache specify both things format-dependent and
   format-independent. This way e.g. when generating latex
   output and we have the cache with format-independent and
   HTML-specific info, we will use format-independent info. And
   we will add LaTeX-specific info to the cache. Then if we
   will generate latex output once again, we will have in the
   cache 3 things: format-independent, HTML-specific and
   LaTeX-specific. And we will use LaTeX-specific.

This has low priority: useful, but not terribly useful (at
least for me). Much work, probably little speedup.

1. Probably cache files should be read and written to the
   same directories where pas files are ? Similar to where
   compilers write .o, .ppu, .dcu files. We would add .pduc
   files.

This is easy to do, but really depends on whether people
want it. Suggestions are needed here.

1. CommentMarker and similar settings should be stored in
   cache files, and if user will specify different
   CommentMarker settings, the cache should be automatically
   assumed to be invalid.

This is seldom needed, but is crucial to give users some
sense of "comfort" when using cache dir. After all, I don't
want to care about whether I should manually remove the
cache files (e.g. because I just changed pasdoc settings
used). In the perfect world, pasdoc should use cache files
always (at least by default). And user should not have
to worry about this at all.

1. Timestamps are compared always against .pas unit file.
   But when .pas file includes some .inc file, the
   timestamping can not realize that some units should be
   reparsed. Inside each .pduc file we should store filenames
   of all included *.inc files.

These are my conclusions from reading pasdoc-main thread
mentioned in this bug report. I think that when 4 points
above will be done, this RFE can be closed. It will not
happen tomorrow :), but at least now we precisely know what
must be done here. If someone sees some other lacks of cache
system, comments are of course welcome.