Closed missinglink closed 3 years ago
Nice, this makes sense. Definitely good for consistency.
I recall from our investigation into abbreviations a few months back that there aren't too many cases of expanded directionals in OA data, meaning this won't have too much effect, right? But the effect it does have is good as it will make sure all the directionals are the same across OA and OSM.
Oh, also, this replaces https://github.com/pelias/openaddresses/pull/479, right?
Oh funny, I totally forgot I already did something similar in https://github.com/pelias/openaddresses/pull/479
Looking at that PR it differs slightly in that:
[north, south, east, west]
includeSelfReferences: true
so it will uppercase existing versions of the contraction which are lowercased (ie. 'Se' and 'se' becomes 'SE')I think in both cases there is a risk of introducing some error, things like 'N Street' and 'S Road' come time mind, and any usage of se
which doesn't mean southeast
.
I think this PR is a lot safer and probably why the other one got a bit stuck getting merged, I'll close it.
Nice, this makes sense. Definitely good for consistency.
Yeah, that's the main value of this, so we handle diagonal contractions the same in OA and OSM
as a follow on from https://github.com/pelias/openaddresses/pull/477 and pairing with https://github.com/pelias/openstreetmap/pull/560 this PR contracts English diagonals in street names.
it only targets four terms
["southeast", "southwest", "northeast", "northwest"
], when those tokens are found they are replaced with the contracted form[SE, SW, NE, NW]
.as such it's pretty safe