Page MenuHomePhabricator

[Scraper] Update Database Checks: Populated Wikibase Software
Closed, ResolvedPublic

Description

Per database analysis:

populated_wikibase_software_expectation_suite.json

  • description: expect_column_values_to_not_be_null: 70% -> 55%
    • Per spot-checking, several instances have url but no other data, and webpages do have no text. However, some have latest_version data but not descriptions, and websites do appear to have description text. Suggest: periodic re-fetch expectations that have no descriptions but have other data, investigate fetcher for proper description parsing
  • latest_version: expect_column_values_to_not_be_null: 60% -> 49%
    • Per spot-checking: several instances legit do not have this datum. Remove expectation
  • quarterly_download_count: expect_column_values_to_not_be_null: 25% -> 0.1%
    • Per spot-checking: several instances legit do not have this datum. Remove expectation
  • public_wiki_count: expect_column_values_to_not_be_null: 25% -> 0.1%
    • Per spot-checking: several instances legit do not have this datum. Remove expectation
  • mw_bundled: expect_column_values_to_not_be_null: 75% -> 57%
    • Found 1 record where not archived and mw_bundled null
  • archived: expect_column_values_to_not_be_null: 75% -> 57%
    • Per spot-checking: several instances have entirely blank pages
  • Suggest: move all criteria except archived to another validation definition that also filters on not archived

Implement above changes

AC:

  • Create nonarchived_wikibase_software_expectation_suite, with appropriate query and validation definition added to relevant checkpoint
  • Move checks except archived into nonarchived_wikibase_software_expectation_suite - use best judgement re relevance
  • Implement periodic re-check of extensions that have latest_version, quarterly_download_count, public_wiki_count, or mw_bundled populated but not description; check that fetcher does properly fetch description. Test instances: https://www.mediawiki.org/wiki/Extension:UnusedRedirects, https://www.mediawiki.org/wiki/Extension:Wikidata.org
  • latest_version: expect_column_values_to_not_be_null - remove expectation
  • quarterly_download_count: expect_column_values_to_not_be_null - remove expectation
  • public_wiki_count: expect_column_values_to_not_be_null - remove expectation
  • archived: expect_column_values_to_not_be_null - remove expectation
  • mw_bundled: expect_column_values_to_not_be_null - reset expectation to 100%

Event Timeline

RickiJay-WMDE renamed this task from [Scraper] Update Database Checks: Nonzero Wikibase Log Observation Month to [Scraper] Update Database Checks: Populated Wikibase Software.

Attempted to follow software fetcher logic; kept getting 403 responses. Added log warning for responses other than 200. Folo recommended.