opl file format

[missinglink]

hey @joto I've been playing with using your .opl format recently, are you still actively using this format for anything?

I've been playing with the idea of a human-readable/grep-able format and so I'm using this as a base.

I'd like to offer a suggestion for the existing format.

The tags are exported in order they are created in the database, this makes it hard to ensure consistency between two identical objects because the order is significant.

It might be better to alphabetically order the tags by key? This would ensure that 2 identical objects produced by different systems would produce output that hashes to the same value.

[joto]

I am using it occasionally for quick scripts, nothing mayor.

Tags are written in the order they are read. Libosmium doesn't do any tag reordering and that is on purpose. One thing it would cost some overhead that not everybody wants to pay for. The other is consistency between different file formats and processing methods. And then there is the problem of how to do a defined ordering: people have different ideas on how to order strings.

So I don't see that as something that can be solved on the level of the file format. The file format should just be a true encoding of the data as it is. But you could write a libosmium-based program reasonably easily that reads an OSM file and writes it out again sorting the tags in the process.

If you have identical objects they have the same id/version, so I am not sure I see the problem with that.

[missinglink]

ok that makes sense, I can see how re-ordering the keys is not a requirement for everyone and it will take some more CPU cycles to do it.