Validator should warn when key is # or +

[daganzdaanda]

I've just been fixing a number of buildings that had a #=yes tag. ( ​https://www.openstreetmap.org/changeset/29472433 )
I tested adding this to a building, and the validator did not complain when I uploaded it.
Most probably these keys come from missing the "Enter" key on a German keyboard. I imagine that in the same way one could arrive at "+" as a key. There is no warning message for "+" as well.

Could you add a warning when there is a key "#" or "+"?

Maybe other keyboard layouts could be taken into account, too. Or be even more general and warn about all single letter keys, or keys with special letters apart from ":" and "_"

[Klumbumbus]

# has 43 uses and + has 82. So I don't think we need special tests for those, but we can test for uncommon characters in general (which would include # and + then). Also the test for keys which are one character long sounds good. Are there keys with 2 characters?

[Don-vip]

Replying to Klumbumbus:

Are there keys with 2 characters?

See ​http://taginfo.openstreetmap.org/reports/key_lengths

[Klumbumbus]

There are:

• ID - not documented, looks like some local imports on taginfo map
• Id - not documented
• kp - replaced by distance
• pk - replaced by distance
• to - used in public transport routes (from=* to=*)

So I think we could warn with "short key" for all keys with 2 or less characters except "to".

[Don-vip]

Replying to Klumbumbus:

There are:

• ID - not documented, looks like some local imports on taginfo map
• Id - not documented
• kp - replaced by distance
• pk - replaced by distance
• to - used in public transport routes (from=* to=*)

So I think we could warn with "short key" for all keys with 2 or less characters except "to".

Maybe allow "id" in lowercase too.

At the other side we should also issue a warning for extra long keys (>55 chars). The longest key I find somewhat acceptable is railway:milestone:emergency_brake_override:direction (52 chars)

[Klumbumbus]

Someone has the regexp for the counts at hand?

[Klumbumbus]

Replying to Klumbumbus:

Someone has the regexp for the counts at hand?

Or could we use the length() expression?

[Klumbumbus]

For character test I have this:

*[/=|\+|\/|&|<|>|;|'|"|%|#|@|\\|,|\./] {
  throwWarning: tr("key with uncommon character");
}

? | and { do not work, they produce exceptions in the console, even with leading \

Any improvements to this?

[daganzdaanda]

Keys should not have any whitespace, so maybe check for the various possible spaces?
Would a check for unicode ranges be sensible? Either positive or negative?

This is a good overview of what's out there:
​http://taginfo.openstreetmap.org/reports/characters_in_keys

[Klumbumbus]

Replying to daganzdaanda:

Keys should not have any whitespace, so maybe check for the various possible spaces?

There is already a test for white spaces.

[Klumbumbus]

The charadcter count somehow does bot work. I tried the following (and a hundred other possibilities):

*[/.{1,2}/] {
  throwWarning: tr("uncommon short key");
}

*[/.?.?/] {
  throwWarning: tr("uncommon short key");
}

What's wrong?

[Klumbumbus]

In 8185/josm:

see #11234 - add warning for some uncommon characters in keys

[Klumbumbus]

Replying to Klumbumbus:

In 8185/josm:

see #11234 - add warning for some uncommon characters in keys

Improvements to this are very welcome.

[stoecker]

@Klumbumbus:

Please open a defect ticket for the MapCSS issues you found and close this.

[aceman]

Are those keys with Chinese characters valid? Is there a rule for keys only being a-z0-9 or something?

[Klumbumbus]

Replying to stoecker:

@Klumbumbus:

Please open a defect ticket for the MapCSS issues you found and close this.

see #11330

[Klumbumbus]

In 8544/josm:

see #11234 - add { } ? * ^ $ to "key with uncommon character" test (availible since r8535)