Different data shown on map for linestring vs geojson

[genebean]

These two URLs show different data for some reason:

- /map/index.html?from=2023-11-19T00:43:33&to=2023-11-27T15:43:33&format=geojson&user=gene&device=foo
+ /map/index.html?from=2023-11-19T00:43:33&to=2023-11-27T15:43:33&format=linestring&user=gene&device=foo

For context, I have backfilled some data via http, but created_at is set to the time I ran a script to backfill the data and tst is set to when the data was actually acquired. The backfilled data is what is missing from the linestring version.

$ apt info ot-recorder
Package: ot-recorder
Version: 0.8.6-0-deb10.1
Priority: extra
Section: default
Maintainer: [email protected]
Installed-Size: 1,968 kB
Depends: libcurl3-gnutls, libmosquitto1, liblua5.2-0, libconfig9, libsodium23, liblmdb0
Homepage: http://owntracks.org
License: https://github.com/owntracks/recorder/blob/master/LICENSE
Vendor: OwnTracks.org
Download-Size: 687 kB
APT-Manual-Installed: yes
APT-Sources: http://repo.owntracks.org/debian buster/main amd64 Packages
Description: A lightweight back-end for consuming OwnTracks data from an MQTT broker

[jpmens]

Did you say Recorder version 0.8.6? That must be something like 2.5 years old; a lot has happened since ... Though to be fair, I don't know why there'd be different data. Is there some way you can send us the data for us to see? ***@***.*** would be the right address, and we'll destroy the data once seen, if you agree to.

[genebean]

Sure thing. The email address above seems to have been redacted though

[jpmens]

Ah, I responded via email, that's why. The address is [email protected].

[genebean]

Sent just now.

[jpmens]

Files received.

@ckrey spent some time looking at the data, and he's found the reason: the listsort() routine we use in Recorder when producing linestring seems to be broken: the sort we need to do to get timestamps (tst) in order to produce a continuous "line" doesn't work properly. This has no effect on the geojson format which just draws the drops on the map.

I've no ETA whatsoever for this fix unfortunately.

The only consolation I have at the moment is that our Frontend doesn't suffer from this bug, and it sports a much nicer interface, so you might wish to give that a spin.

[genebean]

Awesome, thanks for the update! I'll try Frontend out