Talk:Pageviews Analysis
Pageviews Analysis – Documentation • Feedback forum • Issue tracker • GitHub • translatewiki.net • #wikimedia-cloudconnect
| Welcome to the Pageviews Analysis feedback forum | ||
|---|---|---|
|
This is the forum to provide feature requests, bug reports, and general feedback for the Pageviews Analysis suite of tools. Some key notes when making a post:
|
||
Bug when accessing multiple pages in dates one of them didn't exist
editI hit a bug while accidentally attempting to use the tool for articles that were not created yet in the period defined in the "Dates" field. In particular, I tried to access the stats for article Rui Rocha throughout 2021, while the page was only created in 2022.
If I enter a single page, I get an error saying "Rui Rocha: Error querying Pageviews API - Not Found". This is not really user-friendly (an error like "Page Rui Rocha did not exist in the specified period" would be better), but it's at least minimally informative and not too scary.
But if I enter multiple pages including the one that did not exist in the specified period, I get the message "Fatal error: TypeError: i is undefined" along with a red pop-up saying "Fatal error: Please refresh the page to try again or report this issue".
IMO both issues should be handled more gracefully, but the second one in particular appears to be missing a guard in the code. --Waldyrious (talk) 19:32, 5 February 2024 (UTC)
- I will get this fixed. Thanks for reporting and providing detailed steps to reproduce the issue! :) MusikAnimal (WMF) (talk) 16:20, 14 February 2024 (UTC)
- Restoring this thread which is being archived due to age, but hasn't (AFAICT) been resolved yet. --Waldyrious (talk) 14:43, 18 April 2024 (UTC)
- Apologies. I will tend to this when time allows. MusikAnimal (WMF) (talk) 19:32, 24 May 2024 (UTC)
Please allow pageviews for deleted pages
editWhen a page is moved to a new title and we don't need redirect, we cannot see the pageview statistics for the old title. For example, on w:hu:Magyar Péter (politikus) we would also like to view the statistics for w:hu:Magyar Péter (jogász). Bean49 (talk) 15:33, 7 December 2024 (UTC)
- @Bean49 Usually the "include redirects" option does the job. You can change the tool to always include redirects in your "Settings". Hope this helps, MusikAnimal (WMF) (talk) 05:01, 24 December 2024 (UTC)
Bot views? 1 audio download = 1 play?
editThis scan suggests one of the files has been played almost 50 000 times within just a few days which seems quite implausible. What's the cause of this? Are bot views maybe not properly excluded?
Moreover, I usually download spoken Wikipedia audios to then load them in my podcast player since there is no proper audio player on Commons itself and the Commons app. Hence, I wondered whether only downloading a file is included in the view counts (I guess as 1 view/play as there doesn't seem to be a separate Downloads stat). Prototyperspective (talk) 12:42, 11 December 2024 (UTC)
- The file has 133 k play now and I recently noticed something went wrong with the file so it's not the entire Wikipedia article but nobody had complained about it. Similarly, the Anthropology article recently got 6 million views on one day – why was that and shouldn't the tool detect such likely bot/script views?
- How are downloads counted – as 1 play? Where to ask about this if not here?
- Prototyperspective (talk) 00:13, 29 December 2024 (UTC)
- @Prototyperspective:
- Pageview statistics ≠ played statistics:
- Pageview statistics is the statistics of loading from server.
- Pageview philosophy is: all is human if You don't define that it's is not so (as user-agent in HTTP request).
- Pageview statistics ≠ played statistics:
- Dušan Kreheľ (talk) 13:35, 27 January 2025 (UTC)
pageview problem
editSince 15th of january 2025 the pageviews of Isabelle de Charrière go down with lot of languages. Even below zero! What has changed with these strange results? Boss-well63 (talk) 15:43, 20 January 2025 (UTC)
problem with siteviews & pageview_complete total counts
editI have been looking at the pageview files this weekend, and found a strange inconsistency in the pageview_complete data. (I have only tested "user", not "automated", and all numbers below are enwiki only)
For the monthly summary files from January 2022 up to June 2024, if I extract the enwiki lines and sum them, I get a result identical to the figures on the siteviews web service, eg 7,781,530,443 for both sources in December 2023.
However, from July 2024 onwards, the counts are different. For example, August 2024 has 7,588,961,893 on siteviews, and 7,565,367,414 in the count I extract from the dump. It is a relatively small discrepancy, but the web interface ones are always slightly larger.
For the daily files, sampling just 10 October, I get the same value for 10/10/22 (244,072,934) and 10/10/23 (257,384,560), but for 10/10/24 the discrepancy shows up again - 244,769,209 in the dump, 246,269,094 on siteviews. So it looks like it's not just the monthly summaries.
Looking at the datestamps on the dumps, it seems that the files for July to December 2024 were reuploaded in January 2025 - presumably this was to fix an error, but I'm wondering if it has caused the two to get out of sync, and if so which is the "correct" one? Andrew Gray (talk) 17:48, 20 January 2025 (UTC)
- Still baffled by this but have logged it on phabricator - https://phabricator.wikimedia.org/T384636 Andrew Gray (talk) 18:07, 23 January 2025 (UTC)
- @Andrew Gray:
- The statistics are in dump, unfiltered, i.e. RAW.
- In the API are filtered or "censured".
- Each pageview methodology has its pros and cons.
- Look both: 2024 – Top pageviews statistics. Dušan Kreheľ (talk) 10:47, 25 January 2025 (UTC)
- @Dušan Kreheľ Hmm - I think any filtering would have the opposite effect? It would make the API-based statistics be smaller than the dumps, but they are reporting larger numbers. A bit mysterious! Andrew Gray (talk) 23:33, 25 January 2025 (UTC)
- @Andrew Gray: Another: dump is RAW, API is modified.
- I don't analyzed your problem. Dušan Kreheľ (talk) 00:20, 26 January 2025 (UTC)
- @Dušan Kreheľ Hmm - I think any filtering would have the opposite effect? It would make the API-based statistics be smaller than the dumps, but they are reporting larger numbers. A bit mysterious! Andrew Gray (talk) 23:33, 25 January 2025 (UTC)