Pin location error: no location or outside boundary

[vonbearshark]

We should collect the pins that don't have a location given in datasets that do provide locations. By default they are set to something like 0,0, putting them somewhere in Ghana. Also, some pins fall outside the boundary, so we want to put them in Cathy, too, probably. In our WPRDC data collection function (#17), we should have a check for the lat and long data, and default the location to the cathedral of learning.

[brlodi]

I agree with collecting the "unmapped" pins (i.e. ones with garbage location data) onto Cathy, probably with a faded icon or other indication that their map location is not accurate. I'm hesitant to do that to pins that are merely outside of Oakland but still in reasonable locations in Pittsburgh. Some of them are certainly miscoded in the WPRDC data, but others indicate cases where the crime occurred within Oakland but the arrest was made elsewhere. I think that's reasonable to display as-is, but if we really want them mapped in Oakland I think it would be the least disingenuous to place them on the closest point of the Oakland border.

Also we should consider broadening the bounds to include at least a couple of blocks around Oakland (once we have GIS up and running instead of just filtering by reported neighborhood) so we're not cutting off points that are just on the wrong side of the street or the like.

[RitwikGupta]

I don't think we should put them into Cathy, it would confuse a person not familiar with the details. I think we should just discard them since they don't really matter.

[brlodi]

Do they really not matter, though? I'd rather report everything, even if some have (clearly marked) incorrect locations, than simply ignore a report because the location was incorrectly coded.

[vonbearshark]

^ same. I think we might need to have something like a little pocket in the bottom left corner that could expand to show all the pins without data information.

As for pins that start outside of Oakland and manage to make it in--this really stretches the boundary of our map. We want to limit it the Oakland area: to pretend that's the only neighborhood that exists. My vote would be to either collect them with the unlabelled pins, or to, as you suggested, map them on the border such that it was clear it spawned outside Oakland (rather than that spot on the border).

And, yeah, we'll keep track of GIS edge cases, but that's also on the WPRDC

[brlodi]

@vonbearshark I was thinking less edge cases and more a feathered border. For example, right now if someone was mugged at the bus stop outside Seoul Mart (at 5th and Neville, NE corner) we wouldn't be reporting it because that's technically Shadyside according to the data, but if they were mugged at the bus stop on the opposite corner (5th and Neville, SW corner) we would report it as an Oakland incident, and that seems counterintuitive (and actually disagrees with the neighborhood labels on the street signs, which say Oakland stops at Morewood.

[vonbearshark]

Look, we just need to make it more illegal to mug people in those locations.

If we set a sort of padding around the Oakland neighborhood, wouldn't that fix the zoom problem and this problem? Also for this problem specifically, wouldn't we still report it, it just wouldn't be mapped (would be in our limbo tab thing proposed)? Or are you talking about how we get the Oakland data from the WPRDC server? Because that could be trickier?

[brlodi]

It's more forward thinking for our plan to use GIS instead of relying on the NEIGHBORHOOD field, in that our GIS boundaries should be padded slightly. It doesn't matter in the meantime.

[vonbearshark]

I agree completely. Worth keeping in mind.

[vonbearshark]

Right now we run a light check for null at least and default to Cathy. We could check for out of bounds and null in the same place and do something with these pins (put in some kind of UI container, maybe a pin aggregation?) instead: https://github.com/Pitt-CSC/club_project_2017/blob/master/main.js#L215

[vonbearshark]

@mbilker do you think you could tackle this one? Would fit right into that awesome function you whipped up