Missing tiles in debug.log separate from missing with looks_like

[johnstoecker]

Summary:

Summary: None

Purpose of change

Currently, when reloading a tileset, debug.log outputs missing tiles, even if their "looks_like" definition is in the tileset.

For example, when the RetroDays tileset is reloaded, debug.log displays "alexandrite_silver_ring" as missing. However, "alexandrite_silver_ring" has copy_from "garnet_silver_ring" which looks_like "silver_ring". The tile_config.json in RetroDays has "silver_ring" defined, so we might want to separately log it.

Cleaning up this output will help tileset maintainers more easily find missing tiles.

Describe the solution

Use the find_tile_looks_like, and if that is defined in the tileset, don't output as missing.

The output would look like:

Missing item: kevlar_silver_ring
Missing item (but looks_like tile exists): alexandrite_silver_ring

Testing

Reloaded a few tilesets, made sure that they worked in game. Checked the output in debug.log for missing tiles. The missing ones are actually missing, and the ones with missing looks_like still appear, but with their generic tile.

[ZhilkinSerg]

Tiles for entities with looks_like are still missing from the tileset and should still be reported. Maybe in !a?a separate line in the report.

[johnstoecker]

Ah, I see -- my motivation was originally to update a tileset. Then, when I looked at the number of missing tiles, it seemed very daunting....

Is the intention of "looks_like" to indicate that something should look exactly like something else (i.e. alexandrite_silver_ring vs garnet_silver_ring), or is that something that we should leave up to the tileset maintainers?

[ZhilkinSerg]

looks_like is just a fallback mechanism - it is used when tileset is missing a tile for an object. We are adding it to objects, but that does not necessary means we don't have to add unique tiles for that objects anymore.

[anothersimulacrum]

Yes, the proper way to say that we want tiles to look the same is to specify them within the tileset definition.

[johnstoecker]

Updated to show missing looks_like in a new line. With that in mind, my proposed change might make things worse, and not better -- tile maintainers would be more likely to rely on looks_like, and not the proper way of copying ids inside the tile_config, or creating new tiles for each looks_like.

I'll leave the PR open, but feel to close it.

[ZhilkinSerg]

I don't see how it can make things worse - you've basically just split output to two rows. Number of reported tiles would be the same.