Made diff about *.pbf ignore features order

[sharkAndshark]

CI failed due to the features order changed.

feature	test on latest(ubuntu 24.04 + postgis/postgis:16-3.5)	other pg version in CI
cmp_0_0_0.pbf.txt, feature of layer 1 which its gid == 19	29th feature	18th feature

A  vt-show which don't care features order might be needed.

[sharkAndshark]

To duplicate this failure just update the docker-compose.yml and run just bless or just test

........
  db-is-ready:
    # This should match the version of postgres used in the CI workflow
    image: postgis/postgis:16-3.5
........
  db:
    # This should match the version of postgres used in the CI workflow
    image: postgis/postgis:16-3.5
........

[sharkAndshark]

@nyurik we should fix this how about make a new vt-show tool? CC @CommanderStorm

[nyurik]

yeah, perhaps a simple tile content dumper, possibly using geozero for the conversion into geojson or csv, and sorting it? I don't think Martin should contain the needed code to decode protobuf itself.

[sharkAndshark]

Agree. It should be a repo outside which read *.pbf, sort the features before writing. CSV or GeoJSON To Be Decided.

How to sort?

Join all the properties key value paris, geometry type and coordiantes and then make a hash?

[nyurik]

I don't think we need a separate repo - this is just a few lines of code we would need, and maybe no code at all - just install the geozero cli tool. See https://github.com/georust/geozero - and if the list of features is exported to CSV, you can run it through ... | sort in a linux shell. Or if you want to store geojson, perhaps use jq to sort a specific array (this one is a bit trickier as you would need to use jq script)

[sharkAndshark]

Not be aware that it's a CLI tool either. I thought it's a lib only.. lollll Yeah. It's better make this in shell script.

[nyurik]

yeah, it is not auto-published (i just created a ticket for that). I guess we can simply do a few line binary if we can't get cli to do what we need, perhaps as a unpublished test tool.

[sharkAndshark]

So let's try shell + csv first. Seems it's the best solution.

[CommanderStorm]

perhaps use jq to sort a specific array (this one is a bit trickier as you would need to use jq script)

For reference: we already have merged something similar ^^
https://github.com/maplibre/martin/blob/26249e217eb1d802a11e511bbaab92372ef91d69/tests/test.sh#L86C18-L86C32