Add support for %parent.%parent in config schema dynamic type names

I don't think we can call this a regression, we did not touch the parsing much since it was committed and this was not even in Drupal 7. Should be investigated nonetheless. I don't even know if this was supposed to work in the first place.

Looks like one can cross-post with himself...

Looking at the committed patch for schemas (http://drupal.org/files/1866610.47.config-schema_0.patch), there is no documentation nor code to make the kind of things like %parent.%parent work. The replacement handling code would replace any instance of %parent with only the immediate parent, there is no concept of going further up in nesting levels.

Therefore I'd consider this a task or feature request, not a bug (also retitled for this). Maybe some discussion indicated this would be useful to add, but this was not added or expected in the initial patch. My understanding in #1953404: Add config schema to field and instance config entities is that we can at least temporarily avoid doing this and have a little more verbose schemas as a result.

we can at least temporarily avoid doing this and have a little more verbose schemas as a result

A little more verbose but also architecturally incorrect, since this forces schema snippets to redefine over and over again some schema properties about a structure that's actually enforced by the container structure.

So we have to trust 10s of schema snippets to properly expose a schema that's actually not their responsibility but the one of the container - only the container cannnot expose it itself because of syntactic limitations in the include mechanism.

This definitely qualifies as a bug IMO. A bug for which there is a known workaround, but still a bug.

I think we should be able to replace repeated instances of %parent with %parent0, %parent1, etc, and then traverse the config tree to build up the replacement array with parents?

It is a lie that we have people working on this.

Moving back to active, since there doesn't seem to be a patch yet.

Patch "should" work.

Did not test it on an actual example yet, plus the %parent replacement logic doesn't seem to be covered by tests currently.
ConfigSchemaTest only checks "typed definitions" that still contain unreplaced [%parent.id] tokens. It still passes if I comment out the code that does the actual replacement in TypedConfigManager.

Hints welcome: how do you retrieve the fully resolved schema for a config file, so that you can you can check replacements were correct ?

OK, figured it out.

Now with tests. I think we're good here.

Of course, you only notice typos after posting the patch...

Woot, this looks amazing. I did not even expect this may be resolved without a huge refactoring. Wow :) Thanks!

Looks good. Committed/pushed to 8.x, thanks!

Yay thanks!