Unknown turn restriction for restriction:hgv:conditional=no_left_turn

[simon04]

What steps will reproduce the problem?

1. Validate ​https://www.openstreetmap.org/relation/12657026 (type=restriction restriction:hgv:conditional=no_left_turn @ (20:00 - 08:00))

What is the expected result?

No validation error?

What happens instead?

Error "Unknown turn restriction"

Please provide any additional information below. Attach a screenshot if possible.

Reported via ​https://app.element.io/#/room/#_oftc_#josm:matrix.org/$1619901120674395AzAOS:matrix.org

Revision:17855
Is-Local-Build:true
Build-Date:2021-05-02 11:46:51

Identification: JOSM/1.5 (17855 SVN en) Mac OS X 11.3
OS Build number: macOS 11.3 (20E232)
Memory Usage: 235 MB / 2048 MB (55 MB allocated, but free)
Java version: 16.0.1+9, Azul Systems, Inc., OpenJDK 64-Bit Server VM
Look and Feel: com.formdev.flatlaf.FlatLightLaf
Screen: Display 1 1440×900 (scaling 2.00×2.00) Display 2 3008×1692 (scaling 2.00×2.00)
Maximum Screen Size: 3008×1692
Best cursor sizes: 16×16→16×16, 32×32→32×32
Environment variable LANG: en_IE.UTF-8
System property file.encoding: UTF-8
System property sun.jnu.encoding: UTF-8
Locale info: en_AT
Numbers with default locale: 1234567890 -> 1234567890

[simon04]

Related: #17709

[Famlam]

Also: #19722

[skyper]

Is it really useful to have a tag for every vehicle type?
E.g. restriction:conditional=no_left_turn @ (hgv AND 20:00-24:00,00:00-08:00) does say the same or does it not?

Probably will introduce restriction:conditional in the scope of #20833.

[Famlam]

Is it really useful to have a tag for every vehicle type?

This is the same as oneway. Hypothetically all i.e. oneway:bicycle= X could be converted to oneway:conditional = X @ (bicycle) (provided it fits the 255 char length limitation). Nevertheless, for clarity, I prefer i.e. oneway:bicycle over oneway:conditional. I guess the same applies to restriction:xxx and restriction:xxx:conditional versus restriction:conditional = ... @ (xxx).

So if you'd ask: is it really strictly necessary, then: no, there are alternatives, indeed. However, is it useful: I'd say definitely yes.

Example:
restriction:hgv -> 1038 cases
restriction:hgv:conditional -> 75 cases
restriction:conditional containing hgv -> only 6 cases
(similar ratios for i.e. bicycle and bus)

[skyper]

Interesting, restriction:bicycle=give_way is ​dominant (1,686 and 95.58% key's value).

If we would cover all possibility the preset will get too big or we need to split it per vehicle type as we need two additional tags per transport mode, e.g. restriction:motor_vehicle=* and restriction:motor_vehicle:conditional=*

Another question, are :conditional type values valid for restriction:[vehicle_type]=*, e.g. restriction:hgv=no_left_turn @ (00:00-08:00,10:00-24:00) ?

[Famlam]

If we would cover all possibility the preset will get too big

I don't think the preset has to cover all possibilities as they're quite rare. I think it's sufficient to silence the validator (and definitely don't throw an error) for those cases.

[skyper]

Ticket #20109 has been marked as a duplicate of this ticket.

[skyper]

Ticket #21363 has been marked as a duplicate of this ticket.

[taylor.smock]

Modify TurnrestrictionTest to account for restriction:mode:conditional and restriction:mode. Note: supported modes are currently hardcoded (see #18383, needs update) to a limited set and no tests are written yet.

[skyper]

I think it makes more sense to (re-)write this test in mapcss.

[GerdP]

I think I added this test to avoid confusing detail messages for unknown restrictions, so the test itself is probably needed anyhow. If we code this again in MapCSS we just have duplicate code.

[skyper]

I thought following should work

relation[type=restriction][/^restriction(:|$)/][/^restriction(:(agricultural|bicycle|bus|caravan|foot|hazmat|hgv|horse|motor_vehicle|motorcar|motorcycle|psv|vehicle))?(:condition)?$/ !~ /^(only|no)_(straight_on|u_turn|(left|right)_turn|no_(entry|exit)/] {
  throwWarning: tr("Unknown turn restriction");

We also have an error (java) and a warning (mapcss) for missing ^restriction(:|$), atm. Should raise the warning to error if the java part is dropped.

[skyper]

I see, we need both lists for the role checks in the java test. To get rid of duplicates how about an error warning (mapcss) for missing restriction*=* (#19722) and the rest in java.