[Taginfo-dev] Fwd: New functions

[Jochen Topf]

On Wed, Mar 04, 2015 at 05:06:10PM +0000, Donal Diamond wrote:
> On 4 March 2015 at 16:36, Jochen Topf <jochen at remote.org> wrote:
> 
> There are a few new functions available in taginfo:
> >
> 
> Thanks for another great update.
> 
> This mailing-list appears to be such low volume as I suspect the other
> users just like me rarely have any update or migration issues :-)

Most people are not as diligent as you with their updates. Many sites are
far out of date unfortunately.

> > - "Similar" tab in key view and "Similar keys" report. Similarity in this
> >   context means one key is a substring of the other (ignoring case) or
> >   the Levenshtein distance between keys is <= 3. Keys with three
> > characters or
> >   less are ignored. The "Similar" tab reports all those keys similar to the
> >   given key, the "Similar keys" report only shows rarely used keys that are
> >   very similar to common keys. Currently there is no way of tuning the
> >   parameters for these checks. Please tell me if you are having problems
> >   with the settings for your extract.
> >
> 
> I found the similar report useful for finding and quickly fixing any typos.
> (My taginfo also uses a regional overpass info so it was dead easy to use
> the Level0 or Josm links to fix the errors.)

Yes, that was the reason I introduced the report. I want to push this a bit and
hope people will take it up and fix those typos.

> The similar report is less useful now for finding typos as most of the
> reports are now for 'name:ga'  vs 'name:xx'
> 'name:ga:' (Irish language name tag) would have a high count and the other
> name:xx tags would have very low counts as they probably only exist on our
> country relation.
> 
> http://taginfo.openstreetmap.ie/reports/similar_keys

Hm. Any suggestions how to fix this? Basically the problem is that any name:LANG
tag will be similar to any other name:LANG tag whatever we do. The LANG code is
just too short. I could add some explicit exception but I don't like to do that.
Exceptions can be confusing for users and they have to be "maintained" when new
tags are introduced etc.

Specifically to check name tags I could compare against the official language
tag registry and show language tags which should not exist. There is a lot of
hidden code in taginfo already around language tags that I never "activated"
because it isn't finished yet...

> > - "Historic development" report showing numbers of unique keys, tags, and
> >   relation types over time. This report uses data accumulated over time in
> >   the taginfo-history.db database by each taginfo update run. This data
> >   wasn't used until now.
> >
> 
> I actually used taginfo's  history db to quickly generate a graph of object
> (way/node/relation)  growth over time for a friend's 'State of Ireland'
> lightning talk.
> 
> Might be useful to add that as well.

Would be interesting yes, and there are other interesting stats hidden in there
like the average number of tags per object or so. Maybe somebody wants to send
me a pull request. :-)

Jochen
-- 
Jochen Topf  jochen at remote.org  http://www.jochentopf.com/  +49-173-7019282