Resolved! Categories double counting

[Aberrant]

Weird issue, some categories are double counting entries. Can be seen here in the Rogue Isles Alignment Missions Lvl 20-29, where the listed number of pages is 20 when it should be 10. This causes issues on pages like Alignment Missions where it displays the contents of a category in a list. Here it shows that each page is for some reason listed twice. I'm not sure if this only has to do with prefixed pages like "Mission:", can't find the cause. May just be that the categories aren't correctly removing old entries when changes are made. As far as I've seen this is affecting the three Rogue Alignment Mission categories 20-29 30-39 40-50, not sure if there are others. The only page in these that isn't currently duplicated in the category is 8UFF00N's Left Hand for whatever reason. This issue is not found on the same pages on the Paragon Wiki. 

 

Edit: This is also likely responsible for Mission:Common - Take package to Stephanie Peebles still appearing in the Category:Candidates for Deletion

 

Edit2: Also noticed these appear to be showing up in the Special Page Uncategorized Pages, despite them having categories.

Edited January 6, 2021 by Aberrant
Added Edit 2. Changed title to Resolved

[Aberrant]

Okay, apologies for the travelogue writing style, but this post was originally written as an edit to the above post while I was figuring this out, but I just kept finding issues so it's a new post. Basically, this isn't just relegated to a handful of Tip missions as I originally thought. As a mini conclusion:

Every page in the "Mission:" "Mission talk:" and "Property:" Namespaces on the wiki at time of original import has an invisible, un-editable duplicate page, with the same title but differing Page ID, found outside of their respective Namespace. They were created as a result of the import from the Paragon Wiki and it's screwing up Categories throughout the wiki making them very inaccurate. Admin intervention will be required to extract and delete these. @Michiyo This appears to cover Page IDs 47140 to 48373.

 

Part 1 - Curid

Looking up the manual to dpl has led me to find what I believe to be the root of the issue. I thought the Category was at fault, when actually it's the pages themselves. By using the dpl parameter 'showcurid=true' on the Alignment Missions page it can be seen that these listed pages appear to have two Page IDs. For instance 'A Long Line of Spirits' has two IDs 48057 and 49308. 49308 is the correct working page, and is properly part of the Mission: Namespace, which we'll call here the 'True' page. Whereas, 48057 has the same Page Title, but is not under the Mission: Namespace and thus doesn't display the Infobox template (as it checks for a Mission: Namespace so it doesn't show if the page is transcluded) and has not been updated by minor edits in 2020, which we'll call here the 'False' page.

 

Attempting to edit these False pages using the edit button on the page makes you redirect and start editing the True page. And editing the False page directly by adding '&action=edit' to the curid in the address bar causes any changes made to instead affect the True page, making these False pages unmodifiable by normal users, since I assume the system applies edits based on Page Title not Page ID.

 

By adding '&action=info' to an address with a curid, you can view the Page Information for a False page without being redirected to the True page. For instance here is the page information for 48057, and you see it has a longer Sort Key, since it isn't in the Mission: Namespace, compared to the True Page's information, which has the truncated Sort Key, and a more recent edit history.

 

So my theory for potentially fixing this issue is, if the function exists, to either merge the False Pages with the True Pages, or to delete these False Page IDs directly rather than by Page Title. The list of False Page IDs for Rogue Isles (Villains and Rogues) missions is as follows:

 

• 48070
• 48074
• 48078
• 48093
• 48129
• 48142
• 48166
• 48167
• 48172
• 48191

 

• 48057
• 48059
• 48077
• 48094
• 48097
• 48105
• 48131
• 48137
• 48139
• 48169
• 48173
• 48181

 

• 48039
• 48044
• 48054
• 48082
• 48087
• 48110
• 48111
• 48113
• 48114
• 48130
• 48132
• 48147
• 48162
• 48164
• 48174
• 48175
• 48189
• 48193

Part 2 - The Issues

Most unfortunately, having had a look about at other mission categories such as Shauna Stockwell Missions, you'll see the same issue of twice as many pages counted as there should be, compared to its Paragon Wiki counterpart. It looks like this is an extremely widespread issue where most Mission pages have a duplicate non Mission: Namespace page that shares it's title, created when the pages were imported. Oof.

 

Which brings us to the true cause of this bug in the case of missions. From what I have observed, it looks like any page in a Namespace that had only a <noinclude>[[Category]]</noinclude> appears to also have a Main Namespace False Page equivalent also in that category, which is basically every mission on the wiki. This is why the other Tip Missions Categories are fine, since they use the conditional Infobox to specify the category, not a <noinclude> tag, so their duplicate pages are uncategorized. At time of Import all of the Rogue Isles Villain/Rogue tip missions had the standard <noinclude> tags method in addition to their infobox. Edit: Actually it appears unrelated to noinclude or categories. I suspect it is simply any page that was in "Mission:", "Mission talk:" and "Property:" namespaces, copied and not put in the correct namespace, and copied again into the correct namespace slightly later.

 

It looks like nowhere else on the wiki uses a lookup to display categories in the way Alignment Missions does so its at least a mostly hidden problem to standard users. With this information, it's a simple matter of adding "namespace=Mission" to the Alignment Missions pages DPLs to only show True Pages, thus fixing my initial issue. However, the true problem lies with Categories which can't do such omissions. It makes it impossible to update existing Mission categories correctly, with False pages with old information left inaccessible and polluting the categories. It also substantially messes with tracking of these pages in Categories like Missing Dialogue, Work In Progress and Incomplete. These will in their current state, never be able to be brought to 0 without these Fake Pages being removed.

 

For instance, here is a False Page of a mission I moved and updated just days ago to instead reference the 5th Column. False Page. True Page. True Page moved. The False Page will remain forever standing in for it's True Page in categories, even when that True Page has been deleted. An extreme case demonstrated by Mission: Common - Take Package to Stephanie Peebles, a deleted page that remains in the category Candidates for Deletion, as it has a False Page still there with the delete template that had been applied back on Paragon Wiki, keeping a dead page in an important category. If you search Common - Take Package to Stephanie Peebles, this False page shows up as the second result.

 

Part 3 - Root Them Out

So to go even further than those IDs above, any Page ID with Mission: in the title that is not in the Mission: Namespace needs to be deleted, without removing it's corresponding same-titled Page that is in the Mission: Namespace. One way these can be distinguished is by their Sort Key. True Pages will have a truncated Sort Key that excludes 'Mission:', whereas False pages have a long Sort Key that starts with 'Mission:'. Interestingly this prevents them showing up in Searches that start with "Mission:", as that assumes you are then looking within the Mission: Namespace, only showing those results.

 

Using DPL with "namespace=", "showcurid=true" and a "titlematch=Mission:%" shows pages with Mission: in the title, but only in the Main namespace. This can reveal the first 500 of these False Pages IDs in the linked urls since DPL seems to only be able to show 500 results at a time. Here is a link to a test page with the below dpl running.

{{#dpl: 
|namespace=
|resultsheader=There are %PAGES% False Pages in Main Namespace displayable on this page.
|titlematch=Mission:%
|allowcachedresults=false
|shownamespace=false
|showcurid=true
|noresultsheader=''No False Pages found''
}}

Edit: Part 4 - Range

Since the pages seem to have been ported in alphabetical order, it's looks like all Page IDs from 47142 to and including 48333 may be these False Mission Pages

 

48334 to and including 48373 seem to be False Mission Talk: pages that are duplicates in the Main namespace. These had no category at all, since they shouldn't ever be categorised, perhaps the reason why they were also duped.

 

Also IDs 47140 and 47141 are dupes of the Property Namespace. So it looks like 47140 to 48373 could be safely deleted to fix this issue completely.

 

Edit: Having run some quick tests, that's all I can find. Three False groups, all within 47140 to and including 48373. All the other namespaces seem unaffected.

 

Edited January 6, 2021 by Aberrant
Added Full Possible Page ID Range. Added Range of False Mission talk: pages. Added False Property pages. Edit: Fixed Thunderhead links.

[Michiyo]

"No False Pages found"  I grabbed a snapshot of the DB, and ran a query to delete all pages between the 2 indexes you indicated above.  Thank you for the investigative work!

[Aberrant]

Hooray! Thanks very much for the swift response!