Geojson polygons are losing specificity, recording values as "inf"

[alemosie]

Two places to see this in action:

Snapshot of breakpoint in app:

In geometry \"inf\" debugging.ipynb on the debug-inf branch. Explore both raw and processed data. The.to_crs(fiona.crs.from_epsg(2263)) is the root of the polygon value conversions.

This occurs in both master and more-headings (#9 pr) branches.

EDIT

Not sure if the infs are invalid, or merely the representation for real numbers. Either way, when I try to extract a single, transformed polygon through the Jupyter notebook, it kills my kernel.

Unrelated, but potentially relevant: https://github.com/WikiWatershed/rapid-watershed-delineation/issues/27.

• "Just checked and inf is the actual value, it's just not encoding the string properly."
• "Inf is also a phrase (D-Infinity) for flow direction, and though it was a string of the method used, but doing an ogrinfo -al on the shapefile shows that it's a Real number."

Not sure if that applies to us. Maybe someone that worked with the geopandas library will have insight?

[interglobalmedia]

What I got in Heroku was an application break, and then later when I tried locally with our changes, I got "no data received" and nothing was sent. But the files did upload. They just didn't get transmitted. I'm also going to share the two posts I sent you here to keep things in one place:

http://flask.pocoo.org/docs/0.12/patterns/fileuploads/#uploading-files (Uploading Files: A General Introduction)

http://flask.pocoo.org/docs/0.12/security/#xss (Security Considerations: Cross Site Scripting XSS)

I'm definitely going to try and help out.

[alemosie]

When I ran the nycha.json file from our sample data through, this is what the crs conversion looked like:

Unfortunately, at this stage, the geojson crs conversion was still producing infs.

[alemosie]

Taking a look at the differences between raw sample json vs the NYCHA.geojson that Sonya provided (represented as raw_places below), I noticed that there is a huge discrepancy in the significant digits of the values. One has precise floats, the other ints. Example:

Also of note -- the sample json geometry objects are polygons, whereas the geojson objects are multipolygons.

[afeld]

I remember seeing infs when we didn't have the EPSG set properly in the code/files...any chance that's the issue here? In other words, perhaps the files you're trying are from a different geographic region? /cc #13