fixing stats
| From: | Gregory Beaver | Date: | Tue, 17 Oct 2006 16:02:46 +0000 |
| Subject: | fixing stats | ||
| Groups: | php.pear.webmaster | ||
| Request: | Send a blank email to pear-webmaster+get-5301@lists.php.net to get a copy of this message | ||
Hi,
I have fixed stats, but it is a complicated process that also affects
pecl.php.net
First, I created a table of aggregated package statistics that contains
a per-month/per-release counter (each release of every package can have
a counter defining the downloads in a certain month)
Then, I took my offline copy of the stats, and processed all downloads
pre-October 1, 2006 to create the file, and have synced the resulting
table with pear.php.net.
Now, to continue, I need to commit changes to pear-database.php wherein
every time a file is downloaded, the counter in aggregated_package_stats
is updated, but the counter in downloads is not.
Things would be simpler, but the downloads table has no index on
package/dl_when, which is needed to speed up the query, but would kill
the speed of the server on the insert/update end.
So, the only solution is to DELETE FROM downloads where dl_when <
'2006-10-01', and to run my little processing script on the downloads
since october 1 after the new code kicks in, and then remove the
downloads table altogether.
This will save TONS of disk space on the server, and speed downloading
up considerably as well. This will also work well until we have a
better method using some kind of trigger on access.
I have tested the code, and get the same results in statistics for both
pecl and pear.
I'm writing this because this is a major database change, and need
feedback before implementing it, as there was some concern on irc from
the pecl camp.
Thanks,
Greg