MapCSS meaning of 'none'

[frodrigo]

combinations.mapcss now provide this rule

789	*[fixme][count(split(" ", tag("fixme"))) == 1][tag(tag("fixme")) != "none"],
790	*[FIXME][count(split(" ", tag("FIXME"))) == 1][tag(tag("FIXME")) != "none"] {
791	  throwWarning: tr("{0} together with {1}. Is the fixme fixed?", "{0.tag}", "{0.value}");
792	  group: tr("suspicious tag combination");
793	  assertMatch: "way name=\"Florist Gump\" fixme=name";
794	  assertMatch: "way name=\"Florist Gump\" FIXME=name";
795	  assertNoMatch: "way fixme=name";
796	  assertNoMatch: "way name=\"Florist Gump\"";
797	  assertNoMatch: "way name=\"Florist Gump\" fixme=\"the name might have changed\"";
798	}

I concerned about this part

[tag(tag("fixme")) != "none"]

It checks if there is no key like the value of fixme key. But comparing to the "none" string.

It asserted with

795	  assertNoMatch: "way fixme=name";

It strange to me to check "none" string as not existing tags.

This assert fails on Osmose-QA MapsCSS implementation.

[Klumbumbus]

from #17296
Yes, to be safe we could replace "none" by something like "a_value_that_hopefully_never_appears_in_the_osm_database_:o"
But it would still fail in the osmose interpreter, would it?.

[taylor.smock]

It looks like I used none because println was returning none instead of null (System.out.println(o == null ? "none" : o.toString());). I think, at the time I wrote the test, I was using println to see what values were output, and using that to check against.

Should we change the println function to print null when the object is null?

[Klumbumbus]

So we have basically 3 different problems here:

1. "none" in the rules is not a good choice as currently a way with fixme=sidewalk + sidewalk=none is not catched by the rules. This can easily be fixed, see comment:1

2. A simple a!=b selector does not require the presence of the key a to match. But in our case [tag(tag("fixme")) != "none"] requires the presence of tag("fixme") e.g. name=*. I don't know if this is a bug, feature or hack, but in this case it makes the rule work in the first place by not matching solely fixme=name. A simpler implementation could be [(tag("fixme")] instead of [tag(tag("fixme")) != "none"], but that doesn't work in JOSM (it matches solely fixme=name).

3. none vs. null as output of println.

[skyper]

Does an additional count solve it?
[count(tag(tag("fixme"))) > 0]

[Klumbumbus]

Replying to skyper:

Does an additional count solve it?
[count(tag(tag("fixme"))) > 0]

No, but I found another solution which seems to work fine:

[has_tag_key(tag("fixme"))]

[Klumbumbus]

In 17038/josm:

fix #19805, see #17296 - Improve "Is the fixme fixed?" validator warning