← Previous day | Today | Next day → | Search | Index
All times shown according to UTC.
| Time | Nick | Message |
|---|---|---|
| 17:00 | cfouts | Can anyone point out to me where the "REPLACE INTO sessions..." statement occurs? I'm not having much success finding it. |
| 17:25 | ryan | cfouts: must be done by CGI::Session? |
| 17:29 | cfouts | I looked there, too, and didn't see anything in the CGI/Sessions tree. It may be in one of its dependencies. |
| 17:32 | I'm trying to figure out why it uses REPLACE instead of UPDATE. I'm guessing a lot of the table fragmentation (and slowness) that I see regarding the sessions table is the result of using replace. | |
| 17:33 | acmoore | cfouts: in previous projects using the same sessions module, we've put a timestamp on the sessions table and deleted rows not touched for some time. |
| 17:34 | otherwise, it seems to grow without bounds. Does Koha's grow like that? | |
| 17:35 | atz | acmoore: using replace would keep it from growing *in data* because it would only be 1/user. |
| 17:35 | but it could still grow *in disk* | |
| 17:35 | (assuming the session is keyed on user) | |
| 17:36 | cfouts | yup, it grows without bounds. |
| 17:36 | acmoore | perhaps it depends on what a "user" is. Do opac users get a session? |
| 17:36 | atz | yes |
| 17:36 | cfouts | yes, anon opac connections store session data. |
| 17:36 | which is how it gets so unruly. | |
| 17:36 | acmoore | that's the limitation we were hitting, too. One row per person hitting a website gets pretty big. |
| 17:37 | cfouts | and the fragmentation slows things down quite a bit. |
| 17:37 | I've filed a feature-request bug for a timestamp. | |
| 17:37 | atz | yeah, i don't think there is any point to it... logically, we only need one anonymous user session |
| 17:38 | is there anything that an anonymous user can do that specializes their session? | |
| 17:38 | cfouts | but REPLACE INTO is slower than UPDATE, so I'm wondering what the rationale is behind using it. |
| 17:39 | atz | cfouts: i think because it reliably destroys the data in the row, even if the update only writes a subset of fields |
| 17:40 | cfouts | oh, interesting. |
| 17:40 | of course, there's only two columns in koha's use of it... | |
| 17:41 | atz | and a proposed third... but that's the idea anyway. CGI::Session had to decide before knowing what our data looks like |
| 17:42 | acmoore | cfouts: /usr/local/share/perl/5.8.8/CGI/Session/Driver/DBI.pm appears to use "UPDATE", are you seeing "REPLACE" statements in the database? |
| 17:42 | atz | acmoore: from CGI::Session code, we suppose |
| 17:42 | acmoore | ah, /usr/local/share/perl/5.8.8/CGI/Session/Driver/mysql.pm |
| 17:42 | atz | performance difference is likely dependent on storage implementation |
| 17:43 | innoDB vs. myISAM vs. dbd or whatever | |
| 17:43 | cfouts | oh, good find. |
| 17:43 | acmoore | the mysql driver overrides the store() method from the generic DBI for some reason. |
| 17:48 | cfouts | and the two columns are hard coded into the module, though cursory glance shows that one could transparently add a timestamp column. |
| 17:50 | acmoore | I've done it (actually, with Apache:Session, but it's similar) with no problems. We let mysql auto-populate the timestamp and then a cronjob removed old ones periodically. |
| 17:50 | in fact, it's documented somewhere that you're supposed to be able to add columns, from what I recall. | |
| 19:19 | owen | atz around? |
| 19:19 | atz | ? |
| 19:20 | owen | Hi atz. I finally found the receipt printer we used to use at NPL |
| 19:20 | It's a Star SP2000 | |
| 19:20 | atz | oh, cool. |
| 19:20 | i'll add that to our doc.... | |
| 19:22 | owen | It doesn't bode well that the link to their driver download page is broken |
| 19:22 | atz | classic. usually google has a better record of support docs/downloads for that old stuff... |
| 19:25 | thx for the info | |
| 21:00 | owen | logbot and newlogbot, together again! |
| 21:26 | acmoore | I never realized that there was actually a "logbot". I thought newlogbot was the only one. |
| 21:26 | does he log to the same place? | |
| 21:27 | chris | nope logbot is the original |
| 21:27 | logs to a database at katipo | |
| 21:27 | newlogbot is a copy of it, i got running when i was at liblime, which logs to a db there | |
| 21:28 | and thats the one that the webinterface searches | |
| 21:31 | acmoore | ah. |
| 21:32 | so, I started using firebug a little more since I have some javascript that I'm writing now. I notice that many pages seem to have js bugs in them. Am I the only one seeing that, or is that kind of known? | |
| 21:32 | chris | its known |
| 21:33 | acmoore | hmm. that's a drag. I wasn't sure if I was the one that caused the "addLoadEvent is not defined" error on this page or not. I guess I'm not. |
| 21:42 | atz | acmoore: that's a bogus value in a syspref that snuck in (some custom dev's value) |
| 21:43 | acmoore | ah. that's coming from my intranetuserjs syspref or something like that? |
| 21:43 | I think I recall now | |
| 21:43 | atz | right |
| 21:44 | acmoore | geez. I thought I was going crazy after not being able to find that with ack, or grep, or even by opening about 4 dozen files in my editor. |
| 21:44 | thanks | |
| 21:45 | atz | we need an updatedatabase (or Datafix) to go purge it if it equals the crappy value |
| 21:46 | acmoore | or something. |
| 22:57 | liz | can you pull a report based on withdrawn status? |
| 22:57 | it doesn't seem to be in the report builder | |
| 09:05 | chris | hmmm |
← Previous day | Today | Next day → | Search | Index