>>> hunk <file> <byte#> <line#> <#old bytes> <#old lines> <#new bytes> <#new
>>> lines>
>>> old:
>>> <the old data>
>>> new:
>>> <the new data>
>>> end.
>> Question: do we want to allow for arbitrary amounts of padding in
>> order to page-align the data? This might provide a slight performance
>> benefit when mmapping.
> That's already allowed by the lexing function dropping arbitrary
> whitespace before "old" and "new".
Hmm... my understanding is that with the new patch format, you just
mmap/read/gzread the patch file, check for the strings ``old:'',
``new:'' and ``end.'' at the right file offsets (which are implied by
the <#old bytes> and <#new bytes>), then extract the actual data. There
is no lexing whatsoever -- you never even touch the mmapped pages
until you need the data.
What I'm thinking about is allowing two new fields, called say
<#old padding> and <#new padding>, which allow arbitrary number of
bytes of padding before the strings ``old:'' and ``new:''. This would
allow to optionally page align the data, which would save a page of
memory when mmapping.
> It would hurt us any time we needed lots of small hunks from one
> file at once, though.
We can always use 0 bytes of padding for small hunks.
> I'm doubtful it will be worth the added complexity of the patch
> writing code.
Yeah, you're probably right.
Juliusz
_______________________________________________
darcs-devel mailing list
[email protected]
http://www.abridgegame.org/cgi-bin/mailman/listinfo/darcs-devel
