[OSM-talk] namefinder problems

David Earl david at frankieandshadow.com
Sun Dec 28 10:54:12 GMT 2008


On 28/12/2008 04:58, Roman Neumüller wrote:
> I just wonder what the problems with the namefinder are.

There are two particular problems at the moment. (a) I managed to 
introduce a bug where it is always doing a search for all occurrences of 
a search string, even when qualified by place, which means it is doing 
to much work at the moment.

(b) the updates have been taking a long time over the last few days and 
it is a great deal slower when searches have to compete with updates. 
Usually they start at 0230 GMT and often take about 4 hours, but in the 
last few days they've been taking all day. I don't know why (maybe 
there's just been a lot of data, or a particularly common name like 
"Main Street" has been updated a lot). I haven't had a chance to look.

> Isn't it only ruby querying a database?

Well, it's not Ruby, it's PHP, but that hardly matters. And it depends 
what you mean by "only". It has a lot of work to do in dealing with 
variants of spellings and especially sorting by distance. See the wiki 
page http://wiki.openstreetmap.org/index.php/Name_finder for details of 
how it works.

> It should not make that much problems as it should only return some text
> back...

No, does a great deal more than that.

> Is it gazetteer.openstreetmap.org which has sometimes problems?

The search on the home page _is_ the same search as at gazetteer. The 
difference is merely in how the results are displayed (oh, and the home 
page also does a search at geonames.org, which is independent of the 
name finder).


> Another thing is that query results of both gazeteer *and* geonames should  
> be
> shown asynchron - I mean results should not have to wait for each other but
> rather show up independent from each other so that which ever comes first
> can be already listed in the "Search Results" bar.
> 
> And: are there any statistics about how often the namefinder is being used
> per day/week/month?

I keep a record of all searches. When I last looked they were running at 
about 10 per minute typically.

David





More information about the talk mailing list