They sound like great ideas!

- Split maps along country borders. This would allow maps to be fully
   customised to country conventions. I'd always thought of this being
   a splitter function and it would be ideal to perform it there, but
   it could perhaps be done using the multi-polygon code.

This would be awesome for international tramps like me. In particular being able to mix drive-on-left and drive-on-right in one .img!

- Doing something with values that are separated with ';'. This is
   clearly desirable, but knowing what exactly to do is less clear.
   I think it would have to be enabled on a tag-by-tag basis.

I had already been chewing this one over. If we are able to treat the values as lists, where the normal case degenerates to a list of one value, then we could think of simple list manipulation functions such as "first element", "last element",  "count of elements", "element N". In practice the count of elements will be quite low (often 2-3 will be enough I think), so allowing a reference to a non-existing element (returning an empty string) would allow quite powerful style-file constructions with only a couple of lines. I'm sure we can get by without "for each element" constructions which sound more difficult to implement.

Nuts to crack:

• how to handle a ";" in the data itself (escape with preceding backslash? enclose value in quotes?) Really this will need standardising across all editors/consumers, but I'm under no illusions about the feasibility of that!
• the "lanes" syntax effectively has lists embedded in lists, using a "|" as another separator. Can we support this with a generic syntax? Could this simply be 2-dimensional data?

Later we could think of extended functions such as "delete element", "append element" etc. This would open the door to hand-constructed labels driven by the styles. I recall that there is a maximum of 4 labels per road and this is put together somewhere in the Java code, so it's difficult to "customise". If this can affect the address search index, then we would be able to add alternate names etc to the index. Imagine a multi-part road name which can be "converted" or "cast" to a list of values, split into parts on the space character.

Example: if a road has multiple ref values, ignore all except the first

(highway=* & ref=* & list_count(${ref}) > 1) {set ref=${ref[0]};}

Example: if a road has an alt_name, add it to the list of names:

(highway=* & alt_name=*) {list_append name, ${alt_name};)

Example: if a road has a multipart name, add the final part as well as the full name

(highway=* & name=*) {set x=split(${name}, " ");}

(highway=* & name=* & list_count(x) > 1) {list_append name, list_last(${x});}

Colin