[Talk-ca] broken forests in eastern Canada

[Michael Reichert]

Hi,

Am 01.09.2016 um 01:21 schrieb dega:
> On Aug 31, Daniel Bégin wrote:
>> On the same topic, it has been suggested to split wooded areas in smaller 
>> chunks by using features on the ground as outer limits (mostly roads,
>> streams, rivers) and get rid of arbitrary rectangles from Canvec.
>> Is it something we are aiming at?
> 
> The grid is an important source of problems. Here are some examples:
> 1) If a lake is on the grid, it will be split in 2 parts. Each part will have 
> a name tag and and 2 identical names will be displayed on the map, one on each 
> part. This problem exist in thousands of lakes. I even saw a lake with a 
> triplicated name.
> Merging the parts would require modifying 2 or more relations and many 
> importers don't do it (even if they use JOSM) because of the complexity. Some 
> have used a quick fix where they remove names from the parts and put it in a 
> POI. It looks fine but that's bad for database integrity.

If someone does not merge two lakes because it is too "complex", he
should evaluate if he is the right person to import such data. If an
import contains much multipolygon relations, the people how import the
data should know how they work, what can be done wrong, how to edit and
how to fix them. :-/

> 2) A addr:interp way may be split in 2 parts. 2 consequences:
> - the interpolation way become useless because it's now 2 different objects.
> - the mid-point becomes 2 superposed nodes. Useless duplication.
> 
> 3) A grid tile has a fixed size. It may be appropriate for unpopulated areas 
> but it is too large for urban areas where editors work at a high zoom level 
> (17 and up). It's easy to damage a relation without knowing it.
> 
> But there are other problems (not related to the grid):
> 4) the relations seem to be designed to be stand-alone. Thus the relation 
> borders don't share a way. They use 2 superposed ways. Useless duplication. 
> It's very confusing and we frequently alter the wrong way.
> 
> 5) lakes are represented by 2 superposed identical objects, one for the hole 
> and one for the lake. If the hole happen to be on top, the name will not be 
> displayed. It's an unjustifiable complexity for the casual user.
> I've also seen triplicated contour (one for the lake, on for the inner role 
> and one for the outer role)
> 
> Yes, all these quirks can be fixed manually but it's time-consuming and error-
> prone.

What about reverting the tiles which have all these issues and seem to
be uploaded with too few checks beforehand, run a better version of the
preprocessor on the CanVec raw data and reimport them again? (That
causes a loss of OSM history but an import changeset is not as much
valueable than a manual changeset)

> Ideally, the contour of a forest must not split any object and it's not 
> possible with a grid.
> The sole advantage of a grid IMHO is to simplify the CanVec exports.

What about replacing the grid by less artificial borders, e.g. roads,
rivers, powerlines etc. If an area has too few of theses objects,
artifical borders could be automatically drawn which are optimized to
cut as few objects as possible into two parts.

> Some years ago I would have proposed that someone write a guide "How to fix a 
> CanVec import". But now I would rather propose that someone write a "How to 
> pre-process a CanVec export before importing it into OSM".

+1

All ongoing changesets which import CanVec data should either use an
improved version of the preprocessor or should undergo the manual
quality checks I described in my other emails and the changeset comments.

Best regards

Michael


-- 
Per E-Mail kommuniziere ich bevorzugt GPG-verschlüsselt. (Mailinglisten
ausgenommen)
I prefer GPG encryption of emails. (does not apply on mailing lists)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: OpenPGP digital signature
URL: <http://lists.openstreetmap.org/pipermail/talk-ca/attachments/20160901/a13efef9/attachment.sig>