| 1 |
John Myers wrote: |
| 2 |
>>if you broke up the path in another table, and referenced that to the |
| 3 |
>>basename, you could probably save lots of space - especially since most |
| 4 |
> It would save a lot of space proportional to the number of unique filenames, |
| 5 |
> but it wouln't really matter in practice. Each filename is only ever stored |
| 6 |
> once, so over time, the largest tables are actually going to be the unique |
| 7 |
> files table and the file->install map. It would also require many more |
| 8 |
> queries to execute |
| 9 |
|
| 10 |
it will matter greatly, here is a simple script to show you: |
| 11 |
|
| 12 |
$ /usr/lib/gentoolkit/bin/qpkg -nc -l kdebase-3.4 | wc -c |
| 13 |
255050 |
| 14 |
|
| 15 |
$ /usr/lib/gentoolkit/bin/qpkg -nc -l kdebase-3.4 | perl -n \ |
| 16 |
-MFile::Basename -e 'next unless /^\//; s/->.*//; push(@{$f{dirname |
| 17 |
$_}},basename $_); END{map { print "$_\n"; map { print "\t$_"; } |
| 18 |
@{$f{$_}}} sort {$a cmp $b} keys %f};' | wc -c |
| 19 |
99327 |
| 20 |
|
| 21 |
That's a ~62% savings. You can use a medium INT to reference the |
| 22 |
pathname, and then use inner joins in your queries. |
| 23 |
-- |
| 24 |
gentoo-amd64@g.o mailing list |
Archives