Bug: missing data in north-west New Zealand.

[rhorho]

In the far north west of New Zealand (between Kerikeri, Kaitaia and Dargaville), there are no networks shown on the map at all, even in urban areas. Networks are shown either side of this area, which appears to be roughly the 173 degrees east area (173.0000-173.9999 E), although 173 E much further south is OK.

I've been in this area (Kerikeri to Kaitaia) for a few days. Lots of networks get logged to the android app, and the app map shows them. When I upload them, the GPS and cell counts are always zero (they are not counted). I've had no problem uploading in other areas.

So either the uploader is rejecting everything in this area, and always has done (hence the empty areas), or they are stored and something else decides it's impossible for any networks to be here, so isn't counting them and isn't mapping them.

I thought initially you might have a bounding box for NZ clipping off this area, but the entries way to the north west (within 172 E) indicate it's not a box being clipped, but a 1 degree slice missing (or something that gives the appearance of that).

[arkasha]

Hi rhorho,

thanks for your interest and attention to detail. We agree that data that should be getting trilaterated over inside the triangle defined in your post is not getting sent to the correct spot. Casual inspection suggests to me that it's related to some quirks in Android GPS reporting we've been tracking through a few other threads in support email and the forums. Would you be kind enough to PM us with a few transaction IDs whose data is notably absent? You can check your export data using the KML export feature for your Uploads list (just click on the transaction ID for a KML summary of the file).

Casual inspection using the some NetIDs you've uploaded that show up correctly in the KML (export algorithm does a simple "find-the-best-location" election per-netId, per-file, rather than trying to use as much data as possible to trilaterate), but the preponderance of the data appears to shove the points over toward the dreaded global origin (0,0).

The good news: no data has been lost. The bad news: we need to figure out why Android GPS libraries are so confidently reporting bad information, and how to filter it better.

[rhorho]

PM sent with link to a KML file from yesterday.

I agree, probably no data lost as the raw records for an example I looked at show the correct GPS, but it gets triangulated to 0,0.

Interesting puzzle! I think the way the data stops at the longitude line each side of 173E is a clue. The networks I found in 172E (I went to Cape Reinga) were triangulated and counted OK.