martin-cp doc source param

[BigBoulard]

Hi team,

I think there's something wrong with the martin-cp command description cause there's no my_table in the command line. After reading the other sections of the doc, I understand that source_name can be a comma-separated list of tables from postgres, and probably a comma-separated list of .mbtiles or (strict) .pmtiles files as well.

source: https://maplibre.org/martin/martin-cp.html

I have 4 tables in postgres (created by osm2pgsql):

• planet_osm_point
• planet_osm_line
• planet_osm_polygon
• planet_osm_roads

martin-cp seems to take a single table as a parameter. Does it mean I need to create a new table that combines data from the 4 existing ones and pass it to martin-cp to generate a valid tileset?

This copies tiles from a PostGIS table my_table into an MBTiles file tileset.mbtiles using normalized schema, with zoom levels from 0 to 10, and bounds of the whole world.

martin-cp  --output-file tileset.mbtiles \
           --mbtiles-type normalized     \
           "--bbox=-180,-90,180,90"      \
           --min-zoom 0                  \
           --max-zoom 10                 \
           --source source_name          \
           postgresql://postgres@localhost:5432/db

[nyurik]

martin-cp can take any source as a parameter, which includes a combined comma-separated source as well. Do note that the default osm2pgsql output may not be what you want - take a look at the planetiler tool to generate your tiles directly from OSM dump in a much better schema and much faster.

[BigBoulard]

Do note that the default osm2pgsql output may not be what you want - take a look at the planetiler tool to generate your tiles directly from OSM dump in a much better schema and much faster.

Trying to create a public repo to illustrate something I struggle to do, I wondered why martin-cp needs postgres to generate a mbtiles file and if it's something mandatory. Because I understand martin can serve mbtiles files to render a map in a browser or another GIS client without postgres https://maplibre.org/martin/sources-files.html, hence this question. There is some context I'm missing about all these tools and their use cases, sorry if my question sounds stupid. I understand that martin-cp is still under development and cannot be considered 100% ready for mbtiles file generation, correct? But, still for serving tiles with Martin, I used to feed postgres with a map file from <download.geofabrik.de> using osm2pgsql and let Martin serve the tiles from postgres, is it still the advised procedure? Thank you so much.

[ricofritzsche]

Generating the mbtiles directly on the fly from the Postgis database sounded very promising to me, as it would also solve the problem of OSM replications. It works well with small amounts of data or countries, but with a country the size of Germany alone it is extremely slow. Maybe I'm doing something wrong.

[nyurik]

Yes, for large datasets / basemaps, you want to use planetiler

[ricofritzsche]

Exactly, that confirms what I also found out. The question is, how do I deal with daily OSM updates?

[nyurik]

All depends on your usecase, as always. If you are building "base map" - use something like planetiler to pre-generate it into an mbtiles/pmtiles file. Generating tiles on the fly for the base map could be expensive (slow). Think of it as a tile cache. Martin can serve that file - no postgres needed.

If you are generating custom data on top of a base map, e.g. taxi info like in the demo, martin could get that data from a database, possibly even with custom parameters (like in the demo). If you are serving tiles and need some extras - like fonts or sprites - martin can handle that for you too.

[BigBoulard]

Thank you, It makes sense. I want to display a "classic" interactive map for customers to select POIs. I guess base layers like continents, forests, and water could be updated yearly or in case of a natural disaster. Shops and roads could be updated fortnightly. All these could be pre-rendered with batch scripts using Planetiler or another tool without Martin which is probably a better tool for serving "live data" like statistics. Am I correct in saying that?