I'm very much of the school... that someone should say what was 4 times slower and why?
:-) On 3 Feb 2008, at 16:52, Richard Fairhurst wrote: > Patrick Weber wrote: > >> Hi >> >> Noticed something very strange just now. Looked at Luxembourg-City >> in Potlatch, and all closed ways (roundabouts, parking lots ... ) >> had been "opened" . By that I mean the last node had been deleted >> so that that there was a gap and the line wasnt closed anymore. Now >> this doesnt seem to have been a manual error, I found dozens of >> errors!! >> >> example : http://www.openstreetmap.org/? >> lat=49.60003&lon=6.10759&zoom=17&layers=B0FT >> >> Go into Potlatch, and the roundabout, the parking lot, the >> graveyard are all not closed anymore (they obviously were before, >> as Mapnik render shows). >> >> Whats goind on ? > > Short version: > > An unintended side-effect of a change to the server code deployed > about one minute before you sent your message. Spotted instantly and > now fixed. Sorry for the inconvenience. > > Long version: > > Since the dawn of time, Potlatch has accessed the database directly > using SQL statements, rather than the Rails object model that the > rest of the site uses. This is principally because I understand a bit > of SQL but not any Rails. (But also because, especially in the old > API 0.4 days, Potlatch abstracted a lot of stuff away from the user - > particularly segments - and that wasn't easily mappable to Rails > objects. Well, not easily for a n00b like me.) > > I'm very much of the school that it doesn't matter if it's written in > Fortran as long as it works. Not everyone else is, and that's fair > enough, they have their own reasons and I'm not going to say who's > right and who's wrong. Anyway, one of those who takes the opposite > view s Steve, and again given that most of the Rails code is his he's > got the right to say that. So as part of ZXV's Week of OSM, Steve > rewrote some of the SQL in Rails. > > Now this is all great and means that Rails developers can understand > the code. _But_ unfortunately it was a full four times slower than > the old SQL. Hence why Potlatch has been running slowly for the last > week. > > So today, Tom did a bit of work to improve this. Tom's improved Rails > code was just over twice as fast as this, which is clearly a win > (i.e. 1.9x slower than original Potlatch SQL). Unfortunately, AIUI > due to peculiarities of Rails, this meant that any node would only be > returned once for a way... which broke circular ways. > > Tom deployed this, 30 seconds later I spotted the problem, two > minutes later Tom had deployed a fix... it's just bad luck that you > were editing in the intervening 2 minutes. :) > > Anyway, more usefully, Potlatch now automatically resizes itself to > your browser window. Still a couple of rough edges to iron out > (particularly with the Yahoo imagery) but hope you like it! > > cheers > Richard > > _______________________________________________ > talk mailing list > talk@openstreetmap.org > http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk > _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
