Peru's import could get some love

Posted by pieleric on 11 August 2011 in English (English)

Looking at Lima's map, I noticed many roads were just lots of unconnected ways. It seems they come from a huge import from "Telcom IP". Comparing it to the Bing imagery is even more scary as at some places it seems it's completely unrelated (although the imagery seems correctly aligned other where). At these places it seems it'd better to delete all of it!

Quite a few streets are named but it's all uppercase. And the streets without (known) name got named "S/N", probably standing for "without name". This could probably be fixed by a script. Moreover, many streets are cut into very small ways with the same name. Sometimes they are also duplicated.

The imagery is precise enough to permit to fix a lot. In some places, like Piura, there has been some clean up already going on. A lot more love could go on. So, to everyone looking to do some armchair mapping during a rainy day, keep Peru in mind :-)

Location: RĂ­mac, Lima, LIMA 25, Peru

Comment from Donald Allwright on 11 August 2011 at 14:30

I did some GPS mapping in Arequipa 3 years ago, the Telecom IP imports seem simply to ignore any data that are already there. I am here on the ground with a GPS available to add points of interest etc., but have somewhat lost the will to live with this import. The person who has done it has done the easy bit, then given up as far as I can tell. If I'm going to armchair map it really doesn't matter where in the world I do it - I really enjoyed doing all the rivers and lakes of Peru from satellite imagery (Yahoo), maybe I'll find another country that needs some armchair love and do that.

Comment from RM87 on 11 August 2011 at 17:32

It does not need a little love but it does need a lot of love. Currently it looks more like a maze than a map.
If someone has time to make a cake on Mapcraft (and time to maintain it) then after posting it to talk list i guess this gets fixed in day or two.
Doing it without organizing mappers means that a lot of mappers get conflicts while editing the area.
I currently don't have that time to make a cake.

Comment from AndrewBuck on 11 August 2011 at 19:32

There is a russian blog post just a bit later than this one discussing this issue. The comments that are there so far seem to say that a rollback of the import is the best way to handle this and I would agree with that. I think the best way to handle this would be to import the data into a separate database and then render tiles of just that data. People can then use those tiles to get street names in conjunction with tracing/gps mapping (which seems to be about the only thing useful in this dataset).

It looks like the city center has better data then the areas around the edges. If this was from the import then that may be useful as well but the better data may be OSM stuff (not sure where it comes from).

I think that the rollback would be a much better way to go about it as it would likely take more work to "clean up" this data then to just make new data ourselves. Anyway, just my two cents.


Comment from Zverik on 11 August 2011 at 21:11

I agree with rollback: won't it be faster to clear everything and then draw from scratch, based on the source data as a background layer? Fixing and cleaning data isn't a fun task: see USA for example.

Comment from RM87 on 12 August 2011 at 07:17

But what about areas that have been cleaned up already? Mapping them twice will give probably a bad feeling for the ones who have cleaned this area.

Comment from RM87 on 12 August 2011 at 07:44

And there are areas without any aerials

Comment from pieleric on 12 August 2011 at 08:59

Indeed, one of the problems of rolling back the import is that it has now been from some time so it's been cleaned up a bit everywhere. But I guess it'd be possible to only remove what has not been modified.

Actually, the most ugly part are the duplicates. When it's not a duplicate, it's rather OK (it's like a GPS track with the additional information on the name of the street). I wonder if there is a tool to detect duplicates: look at the ways committed by "TELCOM IP", check whether the way is close from another way not from this import (within 5m or something), copy the name, and delete the imported way. Does this already exist?

Comment from Pieren on 12 August 2011 at 09:14

Of course : roll back ! This is one of the best example about crapy imports in OSM worldwide.

Comment from marscot on 15 August 2011 at 14:46

oh dear that's a bit of a mess,

Login to leave a comment