forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Crossley <>
Subject Re: Reducing Forrest build time
Date Tue, 09 Aug 2005 07:55:23 GMT
Diwaker Gupta wrote:
> As Forrest grows and becomes more complex, we really have to watch out for the 
> build times, and think of ways to reduce the build time. I was doing a build 
> today of site-author, and it took ** 35 minutes **
> Now I might not be running the most powerful machine out there, but its decent 
> -- 1.7GHz centrino, 512 RAM. Seeing the build process eat up ~100% of the CPU 
> for a half hour straight was not a pretty sight.
> What do other devs feel about this? Is this something we should be seriously 
> looking at? Atleast I feel we should be.

Same for me, 31 minutes. However ...

I rolled back my SVN to (r227321) just before the
recent Cocoon upgrade. Only 8 minutes (still not good, but).

So something in the Cocoon upgrade (or associated libs)
is the problem.

Also, we have already identified a number of other
problems, but no answers yet ...
"run a memory profiler while forrest is operating"
"PDFs for the generated DTD documentation is going haywire"

> What are the best ways of 
> approaching it? (avoiding building unchanged document is the obvious first 
> step I guess)

Well yes, but that issue is complex. Evidently
Ferdinand talked with Cocoon people at ApacheCon
and might now have some answers.

One thing to bear in mind is that each time we
enhance the pelt skin, then *all* pages need to
be re-generated even though their content hasn't changed.

Also adding a new page means changes to the navigation menu
which means regenerating lots of pages that haven't changed.

> Also, the site-author build currently generates 783 pages. Do we really need 
> all those pages? (I'm just thinking aloud here, so feel free to disagree). I 
> understand that we need to demonstrate Forrest's power so its good to have 
> different output formats available, but can't we just have a "demo" area 
> showing different output formats for different input formats (like we have in 
> the fresh-site). Do we really need to carry around all of the other PDFs 
> (other than the fact that Google might have indexed them) -- perhaps an 
> analysis of the server logs will help us make those decisions.

We have all of 0.6 and 0.7 and 0.8-dev documents being
rebuilt each time. The plan is to always have three
versions available.


> Just testing waters here :)
> -- 
> Web/Blog/Gallery:

View raw message