Fix merging of properties type and description

[progval]

Should fix #103, but not tested.

@meldra could you check this works?

[meldra]

It does not seem to, and that's because the Makefile isn't fetching v3.1 at all...

It probably needs to be rewritten for going past v3 anyway in the forseeable future, but I'm inclined to hardcode in 3.1 just so the docs are correct for now.

[meldra]

This didn't address #103 at all because the property term in question is only in v3 (and the error was in v3.0 which is what the Makefile has hardcoded, not v3.1, so it wasn't a string replacement issue).

This does attempt to address the issue introduced in codemeta#342 where typos in types and descriptions were fixed are causing the script to parse duplicates. It doesn't work though, but should with this patch that I cannot push to this branch. Someone other than me should check the validity of the merge.

[meldra]

This didn't address #103 at all because the property term in question is only in v3 (and the error was in v3.0 which is what the Makefile has hardcoded, not v3.1, so it wasn't a string replacement issue).

This does attempt to address the issue introduced in codemeta#342 where typos in types and descriptions were fixed are causing the script to parse duplicates. It doesn't work though, but should with this patch that I cannot push to this branch. Someone other than me should check the validity of the merge.

Though it occurs to me now that the types should be displayed in a format like "URL (v2) URL or Text (V3)"

[progval]

Though it occurs to me now that the types should be displayed in a format like "URL (v2) URL or Text (V3)"

it may be more readable to make them two separate rows when types differ

[progval]

This does attempt to address the issue introduced in codemeta#342 where typos in types and descriptions were fixed are causing the script to parse duplicates. It doesn't work though, but should with this patch that I cannot push to this branch. Someone other than me should check the validity of the merge.

much better, thanks

[meldra]

Though it occurs to me now that the types should be displayed in a format like "URL (v2) URL or Text (V3)"

it may be more readable to make them two separate rows when types differ

I just had a look, and almost all of the types changes were removing double-spacing... this would be a little bit of a mess of logic.

[progval]

we can normalize with re.sub("\W", "", s).lower() (which strips all non-letters) when comparing

[meldra]

Ok this was fun. New patch.

This puts differing types on separate lines for readability as discussed here yesterday. There were complications.

First, there are multiple Identifier properties. They are obviously not all equal yet they were being overwritten to be identical. This is now mitigated and they do not become clones anymore.

Second, when the property has a change to both type and description, it was resulting in a row with both versions and a row with the one of the version. I'm pretty sure these now work correctly. Changing the value of codeRepository's type will confirm this.

[progval]

First, there are multiple Identifier properties. They are obviously not all equal yet they were being overwritten to be identical.

Good catch!

This is now mitigated and they do not become clones anymore.

v2 did not appear in the list of versions for Thing's identifier. I fixed it.

[meldra]

If you're satisfied that this is good enough (you're quite sure there are no other missing versions, improperly overwritten types/descriptions, etc) then I can flip #105 to ready for review and these can both be merged in.

Given /terms has a lot of things pointing at it, is there anyone else who would want an opportunity to weigh in before we do that?