[Talk-ca] canvec / shp-to-osm

[Frank Steggink]

Hi Sam,

I've just downloaded some CanVec data, and had a look at sheets 031I07 
and -08. I wonder what you mean by uploading all "sub-residential 
files". I understand that the data is separated over multiple files, 
because of certain limitations. In the residential OSM files I also see 
no polygons with a multipolygon relationship of "outer". So,this means 
that the outlines of places like Trois-Rivieres and others are missing. 
The same issue is going on with wooded areas. The data is converted with 
Canvec2OSM version 0.9.4.

I had a closer look at the raster file (from Toporama) of sheet 031I08, 
because there is much less data, and I looked at the village of Gentilly 
(see [1]). This is in the center of the sheet. The raster file suggests 
that a multipolygon relationship should be in place, but the vector file 
(BS_1370009_2_Residential_area0.osm) shows only the two inner polygons. 
Are the outer polygons stored in a different file, or are they not 
converted at all? The shape of the outer polygon doesn't look to be 
complex, so I don't think the max_nodes threshold would be exceeded. 
Looking at the OSM file: there is only one multipolygon relationship in 
it, but it only refers to the two inner polygons, and not to any outer 
polygon at all.

One note regarding multipolygons: the inner polygons shouldn't have any 
tags at all. See [2].

Anyways, some clarifications about what is going on, and how the data 
should be interpreted would be welcome. I'm reluctant to import data 
which looks not correct. For the rest, keep up your good work :)

Regards,

Frank

[1] http://osm.org/go/cKHX9ApT-
[2] http://wiki.openstreetmap.org/wiki/Multipolygon

Sam Vekemans wrote:
> Hi Richard,
> i think your refering to the large multi-polygons such as
> 'residential_area', and it 'appears' to be inverted.
>
> Here's the majic; when all the sub- residential.osm files are uploaded
> to OSM, it renders correctly.
> In JOSM, you need to zoom out and load the area, to see it.
>
> I think i'll load a region of NFLD in the next cuple days to test my hypothises.
>
> Sam
>
> ps. I cc'd talk-ca as this was mentioned b4.
>
> On 9/22/09, Richard Weait <richard at weait.com> wrote:
>   
>> Dear gentlemen,
>>
>> I've had a look at some of Sam's test areas.  In 1435 files there are
>> zero occurrences of Relation=outer.
>>
>> So at some point we started calling relation=outer, relation=inner or
>> completely dropping outer relations by mistake.
>>
>> I do still see rare nested ways, but both are marked as inner, and are
>> on separate layers after --maxnodes
>>
>> I've run 0.6.1 again with an old rules file and see the same problem
>> so I believe that this is an issue in shp-to-osm.
>>
>> Ian can you check a 0.5.0 - generated file and see if it contains any
>> "outer"?
>>
>> Best regards,
>> Richard
>>
>>     
>
>
>