Smarter link-with-namespace filter

[BoboTiG]

• Wiktionary page: https://fr.wiktionary.org/wiki/halq%E2%80%99em%C3%A9ylem

Wikicode:

[[Stó:lō]]

Output:

None

Expected:

Stó:lō

The wikicode is stripped at

ebook-reader-dict/wikidict/utils.py

Line 391 in fdca309

text = sub(r"\[\[[^:\]]+:[^\]]+\]\]", "", text) # [[foo:b]] -> ''

Not sure what to do for now, just reporting the issue.

[lasconic]

List of namespaces on french wiktionary: https://fr.wiktionary.org/w/api.php?action=query&meta=siteinfo&siprop=namespaces

Probably the same for other languages. Spanish : https://es.wiktionary.org/w/api.php?action=query&meta=siteinfo&siprop=namespaces

So we could create a script to get the namespaces, and only filter known namespaces...

[BoboTiG]

With 1b09061 it will be easier to use locale-specific code then.

[lasconic]

Mmm ok, I would have pass the language code in clean to get the correct array of namespace. If I understand correctly, here we will need to write a "clean" function per language with the current clean being the default. Correct ?

[BoboTiG]

I didn't want to add the new argument to clean() everywhere. We already have the locale name passed to process_template().
So you can still go with your idea, it'll just be a 2-lines change to actually add the new argument to clean() (instead of ~30 changes before the refactoring).

[lasconic]

I have something more or less working but I'm scratching my head on these two (on top of [[Stó:lō]]) or course. Not sure how to handle this yet

[[Fichier:Blason ville fr Petit-Bersac 24.svg|vignette|120px|'''Base''' d’or ''(sens héraldique)'']]
[[Annexe:Principales puissances de 10|10{{e|−6}}]] [[gray#fr-nom|gray]]

[BoboTiG]

Would you like to open a draft PR so that I can try too?

[lasconic]

There you go.
1/ I'm not sure it's really needed to keep all-namespaces.py in scripts. They probably don't change that often.
2/ I can't find a way to diffentiate the Fichier and Annexe namespace but we need to clean them at different time.
3/ The line you reference above is commented in my code but it we used to pass this test...

clean("[[http://www.tv5monde.com/cms/chaine-francophone/lf/Merci-Professeur/p-17081-Une-peur-bleue.htm?episode=10 Voir aussi l’explication de Bernard Cerquiglini en images]]")

[BoboTiG]

1/ You're right, we could move the file to langs.
2/ 3/ Still checking ⌚

[BoboTiG]

OK I made progress, will propose a patch on the PR.

[BoboTiG]

1/ After more thinking, keeping namespaces in scripts would ease our work when adding new locales, and it will handle updates automatically. I would say we can keep it as-is. The only tiny detail is about using ALL_LOCALES instead of having duplicate lists; but it's a minor issue.

[lasconic]

OK but then it's less easy to implement 2 categories of namespaces, the one with text and the one without.

[lasconic]

I believe it breaks

{{recons|lang-mot-vedette=fr|[[Reconstruction:proto-germanique/*berhtaz|berht]]}}

[BoboTiG]

Can you share the word using it 🙏 ?

[lasconic]

Sorry https://fr.wiktionary.org/wiki/Berthe

[lasconic]

Ok. I have a simple solution... It passes all the tests already in the code... I must miss something.

#1530

[lasconic]

Ok. Already found a problem ...

https://fr.wiktionary.org/wiki/Daghestan uses the File namespace ...

A solution could be to always add File and Category, the english namespaces in the pattern list.

[BoboTiG]

A solution could be to always add File and Category, the english namespaces in the pattern list.

👍