Better handling of semicolon in values: make it harder to add such values

[skyper]

The handling of semicolons could be enhanced. For background please have a look at:

• ​thread at talk@osm
• ​Jochen's blog an analyses

Two points came to my mind:

1. make it harder to add semicolon separated values when combining by asking the user for interaction for each value with conflict and not automatically adding both values with semicolon as separator.
2. Validator warnings/errors about semicolons in values where it makes no sense as many keys need single value to work.

[simon04]

For item 1, the attached patch drops any default tag combining decision. The contained comment states:

// Do not suggest to keep all values in order to reduce the wrong usage of semicolon values, see #9104!
// Do not suggest to keep the single value in order to avoid long highways to become tunnels+bridges+...

For item 2, I would suggest to have a blacklist of keys resulting in a validation warning (like for amenity, …). For all other keys (except for a few whitelisted ones as ref, source) a validation information should be generated.

Since I do expect some discussion for both items, lets keep it for the 14.02 release.

Please assist in providing blacklist+whitelist keys.

[simon04]

In 6770/josm:

see #9104 - Do not suggest to combine conflicting values in order to reduce the wrong usage of semicolon values

[simon04]

For item 2, here's an analysis based on the TagInfo database: ​https://gist.github.com/anonymous/9058556

select key, sum(count_all), min(value), max(value) from tags 
where value like '%;%'
group by key
having sum(count_all) > 100

Using this list, we can identify some keys to whitelist.

[Don-vip]

That's a huge list. Can we sort it by number of total occurrences for each key?

[simon04]

Here you are: ​https://gist.github.com/9076516

This has been sorted with the very intuitive command sort --numeric-sort --key 4,4 --field-separator='"' --reverse semicolon.csv ;-)

[Don-vip]

Point 1 fixed. See #9757 for point 2 in next milestone.