Dead links to categories / Support categories

[Popolechien]

Wikisource home page directly links to categories for books - however this is not correctly handled and most links are unclickable.

[kelson42]

Yes, categories are not mirrored. This is a work to do in mwoffliner. Probablyt the top priority.

[kelson42]

@ISNIT0 Here is currently the TOP priority topic on mwoffliner. It's not extremly complicated but need a bit work. Let me know if you are interested to have a look so I can explain you a bit.

[ISNIT0]

@kelson42 I'm interested :) What's the best place to start looking?

[kelson42]

@ISNIT0 Let's make a video conf about that. Let me know when you have time.

[tim-moody]

Was 'probably the top priority' in Sept of 2016 yet still not implemented. Any time frame?

[kelson42]

No, it is still the top priority, but there is nobody to work on this so far.

[WikiDocJames]

Thanks for pointing me to this. Hope to see it fixed sometime soon. Maybe a google summer of code project for someone?

[kelson42]

@WikiDocJames Maybe even if this first GSoC we are managing is focus on Kiwix-Android. That said if someone comes to me and is motivated and capable, I might consider to mentor it myself.

[holta]

Further Context: this issue directly affects Haiti schools who've made clear they would use Vikidia IF its link ("84 super articles") were clickable in the top right, as seen in the current Vikidia ZIM here:
http:// iiab . me : 3000 /vikidia_fr_all_novid_2018-03/

Current Vikidia ZIM downloaded from:
http://download.kiwix.org/zim/vikidia/vikidia_fr_all_2018-03.zim

Compare the original (online) version at https://fr.vikidia.org works far better. However the offline version (above ZIM file) is extremely frustrating to educators or children, when the most important link ("84 super articles") is not yet fixed — in future these essential materials should appear much like they do online here:
https://fr.vikidia.org/wiki/Cat%C3%A9gorie:Super_article

PS @kelson42 has clarified that he's hopeful this will be fixed before the end of 2018.

[kelson42]

Things to do (the ones I can see):

• Include "Category" namespace to the namespace to scrape per default
• Verify the category pages are scrapped properly
• Secure that links within articles pages to categories work properly
• Secure the category links at the bottom of the page are displayed properly
• Secure the list of articles are displayed properly like online (sorted alphabetically) are displayed also offline
• Secure the category pagination works properly
• Remove articles which are not mirrored from the category list of articles.
• Mirrors categories which only have at least one article in it.

[ISNIT0]

What is the best thing to do for an articleList selection? Keep all the many parent categories? Not keep categories? Keep only one level of categories? Something else?

[kelson42]

@ISNIT0 Keep each category with at least one non-category child and merge all categories (to the top one) if there is only one sub-category.

[ISNIT0]

What about categories with media? e.g. https://commons.wikimedia.org/wiki/Category:Birds_in_art

[ISNIT0]

There doesn't seem to be a way to get the structured data of what order to show the sub-categories in. It's not just Alphabetical:
e.g. https://bm.wikipedia.org/wiki/Cat%C3%A9gorie:Lien_th%C3%A9matique_pour_cat%C3%A9gories
The single category is in the "G" namespace

and
https://en.wikipedia.org/wiki/Category:London
There is a *, Β (greek letter), Ι, Ξ, and Σ

Any suggestions here @Popolechien?

The query I'm currently using is this: https://bm.wikipedia.org/w/api.php?action=query&list=categorymembers&cmtype=subcat&cmlimit=500&format=json&cmtitle=Cat%C3%A9gorie%3ALien_th%C3%A9matique_pour_cat%C3%A9gories
Which only gives back the article namespace, pageid, and title.

[ISNIT0]

Progress:
I've added a --getCategories work in progress flag which enables the category scraping.
There are certainly issues with the current implementation, so it should not be used yet.

Each article has a Categories section added to the bottom with a list of links to category pages, each Category page has a Sub-categories section which links to sub-category pages.

TODO:

• Display categories as on wikipedia.org
• List pages within a category
• Improve efficiency of category page scraping

[ISNIT0]

It seems to display the categories in the same way as MediaWiki displays them, we need information that isn't available through the API. Instead I'm just grouping them Alphabetically which is pretty close

[ISNIT0]

Progress so far:
https://framadrop.org/r/R1A5MJwaey#hxz6gNnGy7mFqv23Hf9SJNQSVL9JPS+pHOGQgVcyDvc=

Known issues:

• Categories/Sub-Categories/Pages sections don't collapse/expand
  (resolved by implementing Use <details> element for collapsible sections instead of JS #677)
• "Hidden" Are not separated from normal categories
  (solved by not downloading "hidden" categories)

[Popolechien]

Yeah, the hidden categories not being weeded out is a real blocker. These are useless and take up quite some space. @ISNIT0 what's your plan about those?

[ISNIT0]

@Popolechien I've just updated the comment above, we're now not scraping them at all. Is this okay?

[Popolechien]

perfect.

[ISNIT0]

@kelson42 @Popolechien
For review:

• EN top 100 (nopic): https://framadrop.org/r/T3rMJ5Lb24#tIg7V9ceXfFjT/96FJpKZ7gu1e70GSsuDDi1+nlKlLk=
• FR top 100 (nopic): https://framadrop.org/r/-7DezLe9kR#X/SpyR0FxJjyoT8Nzvxh0M/OkwqMwOJA3r4udgEg/Yc=

[Popolechien]

Niiiice.
Am I right to understand that all categories within the categories will also be showcased (ie not only the categories in the articles themselves)?
Either way, good job!

[ISNIT0]

BM Full nopic: https://framadrop.org/r/cyk0sHthFk#vjOsZMdLvq9vqrulrpSOO/WUqSAlZ7ehMf6Zv36aVy0=

No, the current logic is to check each article for categories as it's downloaded. Then we only end up with categories that contain at least one article as per @kelson42's spec:

Mirrors categories which only have at least one article in it.

[kelson42]

I have tested with https://framadrop.org/r/D1EE0C6YxL#SwJO6719lYGfukNN1i71HHy1glAK4MaJTdKiifDHBlo=:

• In Category pages, if an article is not in the selection it should not be listed (currently in black)
• I think we should choose an other namespace to put category pages. The ZIM specs talks about "U" see https://wiki.openzim.org/wiki/ZIM_file_format#Namespaces
• Up categories should be migrated too, it is not the case in "Category:2010s in Austria".

[ISNIT0]

@kelson42 What do you mean by "Up categories should be migrated too"?

[kelson42]

@ISNIT0 In mean "categories parent categories", the full ancestor three should be downloaded (but of course in a simplified version).

[ISNIT0]

@kelson42 You previously said:

Mirrors categories which only have at least one article in it.

[samkellerhals]

@ISNIT0 @kelson42 @Popolechien thanks a lot for working on this - I think the addition of live category links will make for a huge improvement! So far I've been working with Kirundi/Kynarwanda/French zims for use in refugee camps and they also appeared with dead links on the index.html page. Are you thinking of applying these changes (active category links) to all zims currently available for download via the kiwix website?

[kelson42]

@samkellerhals This is the goal, might take a few additional months to see it happening everywhere.

[ISNIT0]

@kelson42 This is now doing the tree-shaking/graph simplification:
https://framadrop.org/r/dIaIeQVRtO#zZjY9W6s5P6ukctJPxU8GDvEQpzAUPdsqSKXbQohwII=

Because this is done using the top 100 articles, there is not a lot of shared categorisation, but Mantis is a good example

[ISNIT0]

@kelson42 I'd like to move the namespacing item you mentioned into a separate ticket and add it to 2.0

I can see it causing lots of back-and-forth with routing edge-cases

[kelson42]

@ISNIT0 From what I can see from last file you have proposed https://framadrop.org/r/P1S5xi6PRm#A6fiUMsysQsdZzr72yXsT6i/QaYm/Dc97iJZZtYktVg= This looks quite good :) That said I was not able to check if the pagination works fine! Do you have a demo ZIM for that?

[kelson42]

AFAIK everything has now been implemented in 1.9, except #762 to be done in 2.0