Issue with stations.json

[whryan]

It looks like the stations.json file isn't quite accurate? My apologies if I just misread this, I'm not amazing w/ Node, but it looks like that JSON uses the remote unit to match entries to station/line combinations.

The issue with this is that remote units aren't always unique to one station -- they sometimes encompass multiple stations, albeit generally multiple stations on the same line. So, technically you're arbitrarily choosing one station out of a few stations and just routing all the traffic through that station.

Is this an intentional choice? If so, it would probably be helpful to document it. If not, this might be something to look into -- I think it wouldn't be incredibly difficult to correct.

One suggested approach (this might be bad, but it seems logical to me):

• Take remote unit, line, and station from rows in the MTA dataset
• Match those to a dictionary which lists remote unit, line, and a list of possible stations, matching on just remote unit and line
• Use some fuzzy string matching function to choose the option for stations in the dictionary which is closest to the one in the raw mta data. This would help account for the fact that the raw data seems to have very inconsistent station names.
• Do that enough, and you should be able to use the results to make a more complete dictionary to use going forward.

I don't know node or javascript well enough to implement this myself, but am working on it in python. If I manage to get a more extensive/better dictionary I can let you know.

[whryan]

Quick update - you can actually get this mostly correct just by matching on line, unit, booth, and division. Then you could go ahead and fuzzy string match the last few, but that should be close enough already.

[r-shekhar]

@whryan Would you have a csv or json file of the results of the match? It would be super helpful to have a mapping for a subway turnstile I'm working on https://github.com/r-shekhar/NYC-transport