Aliases should help with searching for objects
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Add aliases to the labels table and return in label search requests | mediawiki/extensions/WikiLambda | master | +321 -8 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Jdforrester-WMF | T262091 Introduce aliases for objects | |||
| Resolved | gengh | T285368 Use aliases in search |
Change 713484 had a related patch set uploaded (by Genoveva Galarza; author: Genoveva Galarza):
[mediawiki/extensions/WikiLambda@master] [WIP] Add alias to labels table
Some doubts about alias/label uniqueness arose as part of Jame's comments on my patch, and I'd like to bring the conversation here so that we can discuss it more comfortably.
The comment is: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/WikiLambda/+/713484/comment/0626bfe7_524c1a2b/
Currently the table is indexed by ["wlzl_type", "wlzl_language", "wlzl_label"] but this index does not impose a uniqueness constraint (for that, I believe we would need to declare the JSON schema with unique=true)
There are uniqueness constraints on the PHP code though, that make sure that primary labels are unique (there cannot be two identical labels for the same type and the same language). However I'm not sure about what uniqueness constraints we want to impose over aliases. Here's a summary with examples:
Label uniqueness:
Labels are quite restrictive because they are the identifying string. They must be unique by type, language and label, which means:
Alias uniqueness:
Alias are not so restrictive, as they do not identify ZObjects but enrich them. That's why I believe aliases don't need to be unique, but let's think case by case.
Alias and Label clashes:
I'm very confused about this!
Yes.
- Can we have two identical aliases for different languages?
- [type=Z61, language=EN, label="javascript", is_primary=false]
- [type=Z61, language=FR, label="javascript", is_primary=false]
Yes.
Alias and Label clashes:
I'm very confused about this!
- Can we have an alias that is identical to its primary label? I understand it makes no sense as it adds no information, but does that mean we should forbid it?
- [zid=100, type=Z61, language=EN, label="javascript", is_primary=false]
- [zid=100, type=Z61, language=EN, label="javascript", is_primary=true]
No, I think we want to prevent this.
- Can we have an alias be identical to other primary labels in the same language? I think it does make sense in here and we should allow it:
- [zid=100, type=Z61, language=EN, label="JavaScript", is_primary=true]
- [zid=200, type=Z61, language=EN, label="JavaScript (ES6)", is_primary=true]
- [zid=200, type=Z61, language=EN, label="JavaScript", is_primary=true]
No. How would the user know which to select if they're both going to be shown as "JavaScript" and of the same type?
Change 713484 merged by jenkins-bot:
[mediawiki/extensions/WikiLambda@master] Add aliases to the labels table and return in label search requests