generate metadata.json for mvt tiles output

[timonkuiters]

I'd like to use the folder of .mvt tiles to build a .pmtiles archive. Using the pmtiles python package to convert (using the disk_to_pmtiles function) requires a metadata.json file at the root of the tiles directory, which for .mvt tiles should contain vector_layers information detailing the tile layers (see https://github.com/protomaps/PMTiles/blob/main/spec/v3/spec.md#5-json-metadata)

Could this metadata.json file be generated by LOOM, or could I know in some other way what vector layers and bounds to expect from LOOM to generate the file myself?

[patrickbr]

I think a PMTiles file would be an interesting additional output format for loom. Glancing over the standard, I think the format is basically a single file with a lookup directory storing the offsets for all tiles, ordered by a space filling curve per zoom level so that you can quickly jump to the offset for a specific tile. After that, all the tiles are simply stored consecutively, in whatever format one desires (vector, some raster formats, or even "unknown"), with additional compression.

Is my understanding correct?

[timonkuiters]

Good to see you think PMTiles would be an interesting output format, that would be even more convenient! And from my own very basic understanding of the format your understanding sounds correct to me, although I'm not very familiar with the workings of the format myself. Maybe the creator of PMTiles, @bdon, could be of help?

[bdon]

Here is an example using the pmtiles.hpp C++ header to write an archive: https://github.com/felt/tippecanoe/blob/main/pmtiles_file.cpp#L264

lancing over the standard, I think the format is basically a single file with a lookup directory storing the offsets for all tiles, ordered by a space filling curve per zoom level so that you can quickly jump to the offset for a specific tile. After that, all the tiles are simply stored consecutively

The ordering on the space filling curve is not strictly necessary but the size of the index is much more compact if it is (e.g. when clustered) since it can store only deltas.

Given that the LOOM tiles seem to be mostly LineStrings and there is little chance of duplication in the tileset it should be straightforward to output.