Sidewalks converter messes up a few sidewalks (very few)

[3vivekb]

A few of the sidewalk ways, after running them through OSMIZER, become giant roads (like 3 of 7,818).

Sidewalks geojson before running through osmizer (and maybe untagged?): https://drive.google.com/open?id=0B098fXDVjQOhd1pGWW5Dc25UeFE

Sidewalks osm after run through osmizer fixing duplicates but contain a few giant segments: https://drive.google.com/file/d/0B098fXDVjQOhUDZMQ3o4NENzOWM/view?usp=sharing

[Andrew-Tan]

Are you able to locate the 3 malformed segments? - Like printing out what they look like originally in the geojson and what they become in the output xml.

[3vivekb]

@Andrew-Tan I added IDs and tried to run osmizer to figure out the bad geometrys. But what i'm seeing is that osmizer gives inconsistent results each time I run it on my full sidewalks dataset. Sometimes I get the long sidewalks, sometimes I get different errors. I haven't looked into how it works.

I pulled out the 2 geometrys that were tweaked when running the full set. I ran osmizer on them individually and they seem fine.

It isn't a huge deal. I don't know if errors would appear as I ran it in a batch call on each smaller set of sidewalks which is how I used it. I don't typically want to run it on the full set.

[Andrew-Tan]

I tried converting your file twice and noticed the issue as well. The malformed segment seems to appear randomly. It looks like a race condition while I did not actually create any extra thread.

We also run Osmizer on smaller data sets and rely on human to check for errors before imported into OSM. This is the way recommended by OSM community as well.