[Talk-ca] [Imports] Importing CanVec better?

[Richard Weait]

Three-part lakes.

Canvec data is split at canvec (NTS) tile boundaries, so a lake the
crosses such a boundary will be split in several pieces on those
straight lines.

There are other reasons to split lakes and other large objects.  We
split roads when the name changes, or when the highway classification
changes, etc.

We split lakes when they contain too many nodes.  The API limits a way
to less than 2000 nodes.  As a practical matter, many mappers restrict
ways to under 1000 nodes.  Given a thousand nodes, a detailed lake
boundary could require several ways.

Some mappers will choose to split the ways along geographical
features; "here's a nice big bay; make it a way!" others will just
divide the lake into four roughly even-length ways.  Who knows?
Mappers are individuals!  Then the lake is "combined" by a
multipolygon relation.  Each of the outer boundary ways become
multipolygon members with 'outer" roles.  Island ways become
multipolygon members with role = inner.

Starting with a three-piece canvec lake, you might consider,
rationalizing it for an OSM multipolygon.  All else being the same:

- Cut the loops and remove the duplicate (vertical / horizontal) NTS
tile boundaries.
- Check that the remaining linear ways form a loop.
- Join, or sever the loop in ways of the right size.
- Check and fix islands if needed.
- Select all of those ways and create the new multipolygon
- add the name, and other tags, etc to the multipolygon.
- Set the inner and outer roles.
- Check the validator to see if you've missed anything.
- Cross your fingers and save. :-)
- Have an experienced mapper check your work until you feel
comfortable with multipolygons.  :-)

Bear in mind that "coastlines" in OSM are a special an beautiful
beast.  If a way is tagged as natural=coastline, then the way
direction is important!  Water goes on the right.  Other ways don't
have to follow this restriction.  And several large lakes in Canada
are treated as coastline, including the Great Lakes.