propagate translator exceptions (see for_pnorman branch) #21
Conversation
|
$ ogrinfo --version This is what ships in Ubuntu 12.04. BTW, I guessed that you were using a newer version of libgdal. I also get a slight difference in the srs output for the utf8 test:
There seems to be some activity in gdal post-1.7 that may account for this difference: Perhaps it also accounts for the utf8 failure. Ciao, |
While there is only one example of replacejwithi being called, it is clear that the arguments are meant to be type Geometry, not (Geometry, role) tuple. So when replacing j with i, preserve the existing role.
|
Hey Paul, I see that you merged my patch. I added another fix to this for_pnorman branch for replacing children of relations. I also eliminated the utf8 failure patch. Can you pull this new patch? Sorry if I'm veering from github protocol here. Ciao, |
|
|
|
Understood. Do you know of any translators or other code that are currently using the Relation object's replaceiwithj? I don't see one in the existing code. I figured the common case might be that two multipolygon relations have the same (untagged) Way. And this is the case that I was fixing when I discovered this issue. |
|
Thinking about it, I'm not sure if the relation A translation shouldn't be calling these, they're used deep in the OCG -> OSM data model conversion Do you have a testcase for where the relation The current version of that line came about in a086c32 but it was simply a mirror of the changes to the way |
|
Hey Paul, This Relation.replacejwithi issue came up while I was working to eliminate multiple multipolygons from a parcel shapefile. The code I'm using is in preOutputTransform function in https://github.com/ual/parceltools/blob/master/translations/sf.py. It is inspired from the code in ogr2osm that replaces duplicate points. The test file I was using is: https://github.com/ual/parceltools/blob/master/tests/multiparcel.shp and friends. There's a multiparcel-ref.osm file that is the target output. Perhaps a translator should not be invoking replacejwithi. But the translator should probably be invoked when tags on duplicated geometries need to be reconciled, and the detect/replace code should probably be in ogr2osm. This was my first approach. But a lot of questions came up about corner cases. Like what if you have two geometrically identical ways, one wrapped up in a Feature with attributes, and the other that is just a multipolygon Relation's child? My head started spinning on these, so I just focused on getting my specific case working in my specific translator code instead of solving the problem in general. I'm happy to work this up as a contribution to ogr2osm if you think this is worthy. Ciao, |
Hello Paul,
I've been working on a translator and found that exceptions from my translator functions were not appearing. I came up with a solution and would be grateful if you'd except it into the mainline. I also fixed a failing test. It seems that the sp_usinas test data attributes are in Latin-1 not UTF-8. But if the test passes on your machine, maybe I'm wrong about this.
Ciao,
Brian