Community Wishlist Survey 2020/Wiktionary/More Lua memory for Wiktionary

From Meta, a Wikimedia project coordination wiki

More Lua memory for Wiktionary

  • Problem: Lack of Lua memory for basic words. See wikt:CAT:E for currently affected words.
  • Who would benefit: All users and readers of Wiktionary.
  • Proposed solution: More Lua memory for Wiktionary.
  • More comments:
    • Pages that lack memory are not being properly categorized and the sortkey is not working properly.
    • Standard information such as citations, semantically related terms are being removed as a temporary solution and this is a loss of information for our readers.
  • Phabricator tickets: phab:T188492
  • Proposer: KevinUp (talk) 16:20, 11 November 2019 (UTC)[reply]


Alternatively, consider implementing a tool so that the source of each language can have its own separate page, like how all the proposals have its own individual page. KevinUp (talk) 16:20, 11 November 2019 (UTC)[reply]

The most parsimonious solution is to raise the cap on Lua memory. This cap seems arbitrarily placed, and it is crippling for long pages. Metaknowledge (talk) 18:04, 11 November 2019 (UTC)[reply]
Indeed. If all were like in a company we would already have more memory, because the time spent to circumvent the cap is dearer than the trifling amount of more memory that in total would be used (since it concerns but some dozens of pages, for which much dust has been raised). Fay Freak (talk) 21:52, 15 November 2019 (UTC)[reply]
@Noé: Hi. Does the French Wiktionary have the same issue (lack of Lua memory) for entries with short words? I think if we were to migrate information from English Wiktionary to French Wiktionary, the same situation would occur. KevinUp (talk) 23:32, 16 November 2019 (UTC)[reply]
Local community discussion regarding lack of Lua memory can be found here, here and here. If only the Community Tech team would be graceful enough to inform us the actual memory that is needed by wikt:do, wikt:一, wikt:人, wikt:水, wikt:月, wikt:生, wikt:我 which are basic words. KevinUp (talk) 20:15, 18 November 2019 (UTC)[reply]
@Pamputt: Hi. Does the French Wiktionary currently have issues with lack of Lua memory? Do you think the same issue would occur if information from English Wiktionary for entries such as wikt:do, wikt:一, wikt:人, wikt:水 were copied to the French Wiktionary? KevinUp (talk) 11:56, 19 November 2019 (UTC)[reply]
@KevinUp: actually we do not use that much Lua module on the French Wiktionary so I am not aware of such limitations. Yet, maybe JackPotte or Darkdadaah can say more. Pamputt (talk) 18:50, 19 November 2019 (UTC)[reply]
We do use Lua modules quite a bit, although not as much as the English Wiktionary. We don't have as much metadata (languages in particular) and automated content, so I believe we are not at the point where memory is an issue... yet. Our main issue at the moment I believe may be for pages with lots of translations, as in this demonstration page which purports to list the word for "water" in all languages: wikt:fr:Utilisateur:Pamputt/eau. In that case though the limit is execution time (>10s).
For the memory issue, it would be nice to have an idea of what takes so much memory, so that we can make an informed decision on how much memory will be needed. Darkdadaah (talk) 16:12, 21 November 2019 (UTC)[reply]
Thanks for the reply. For translations, some entries in English Wiktionary use wikt:Category:Translation subpages to redirect content and reduce Lua memory. Yes, it would be nice if we knew how much memory is actually needed by pages in wikt:Category:Pages with module errors. KevinUp (talk) 07:15, 22 November 2019 (UTC)[reply]
@Lo Ximiendo: Does the Chinese Wiktionary use a lot of Lua memory as well? I noticed that it uses similar modules from English Wiktionary. KevinUp (talk) 05:13, 30 November 2019 (UTC)[reply]
@KevinUp: All I know is, that the Lua modules in the Chinese were imported from the English Wiktionary, with modifications made of course. --Lo Ximiendo (talk) 05:29, 30 November 2019 (UTC)[reply]
There's also a phabricator task for better memory profiling support, which would allow us to do targeted optimizations instead of blind guesswork. – Jberkel (talk) 08:28, 2 December 2019 (UTC)[reply]