Jump to content

Reading/Web/Projects/A frontend powered by Parsoid/Notes

From mediawiki.org

Notes of the research project through the weeks, and the planning sessions.

Things left to tackle

[edit]
  • loot-ui & loot
    • Cache invalidation strategy (on loot api and loot-ui client)
  • Reports
    • Html
      • Wikipedia parser analysis (analogous to Restbase analysis)
    • Webpagetest
      • Define how to (and do) measure
  • Slides

Dec 28 - Jan 1

[edit]

Dec 21 - 25

[edit]
  • Yes Done Fix a couple of regressions with cached site and broken lazy loaded images
  • In progress In progress Expand report to include other transformations like superficial markup and ids
  • Yes Done Defer loading of initial web app data (to minimize initial HTML payload)
  • Lazy load references in loot-ui
  • Run Webpagetest to generate reports in browser rendering of wp web, wp mobile web, restbase and loot-ui

Dec 14 - 18

[edit]
  • Yes Done Add CI running tests & lint on master & pull requests
  • Yes Done Handle 404s and link tag redirects from restbase
  • Yes Done Show last modified bar on articles.
  • Yes Done Gather data about HTML size with different transformations individually and make report (task T120504)
  • To do To do Add tests for UI components (server side testing of browser components and its interactions) (delayed)

Dec 7 - 11

[edit]
  • To do To do Section collapsing (Delayed)
  • To do To do References popup (Delayed)
  • Yes Done Add tests to loot-ui (reducers, components, actions)

Nov 30 - Dec 4

[edit]
  • Yes Done Add tests to loot-ui (reducers, components, actions)
  • Yes Done Implement initial service worker caching (Pull request)
  • Yes Done Finish search feature
  • If time, section collapsing

Nov 23 - 27

[edit]
  • Research Service worker & learn (Reading list)
  • Yes Done Document current architecture and code structure in loot-ui (link to docs)
  • Yes Done Create application Chrome .
  • Yes Done Start adding search feature.

Nov 16 - 20

[edit]
  • Client side data caching
    • Use indexedDB + fallbacks and cache in the client the state and fetched data to avoid future round trips.
    • Visit an article, click a link. Navigate back. Refresh the page as if you visited later, click on the same link, load should be instant.
  • Lazy loading images
    • Improve transformation of images markup in loot.
    • Lazy load images in viewport in loot-ui.
    • Visit an article, you should see images in viewport loaded. As you scroll you'll see the rest of the images appear instead of all at once blocking the browser and using bandwidth.
  • Sync with Services team
    • Chat about Apps - Api driven frontend - Loot
    • Shared basic services composed in aggregation services.

Nov 9 - 13

[edit]

Nov 2 - 6

[edit]
  • HTML only experience https://reading-web-research.wmflabs.org/wiki/Struggle_session
    • UI urls: https://github.com/joakin/loot/issues/12
      • /wiki/:title
        • Chrome + slim
      • /wiki/full/:title
        • Chrome + full
    • Api:
      • Main transformations
        • Sectioning
        • data-mw
        • comments
        • img -> link
        • style attrs: 
          • whitespace: nowrap
          • width: XXX ???
        • strip <base> element from head/only serve body contents
        • Rewrite links (parsoid outpus "./:title")
      • /api/slim/lead/:title
        • Infobox
        • Lead section
        •  With a "Read more..." button
        • References (only lead)
      • /api/slim/:title
        • Infobox
        • All sections
        • All references
  • Connect with Peter for performance measuring
  • Sync with mobileapps content service devs.

Oct 26 - 30

[edit]
  • Clean up loot server produced in offsite's hackathon.
    • Move to more efficient parsing library (libxmljs).
    • Add more transforms than the basic ones.
    • Add basic sectioning.
  • Reproduce research and produce output report.
  • Deploy loot to vm in wmflabs and set up git deployments.
  • Plan next steps, html only version.

Oct 19 - 23

[edit]

Offsite

  • Hacked a middleman node server to test parsoid html performance with & without stripping content.
  • Anecdotal results from hacking time show very interesting data.
  • Plan to formalize the server and write up a more informed report.

Oct 12 - 16

[edit]
  • Adapt existing web app prototype to use parsoid html and review possible problems
  • Plan to test standalone parsoid html to get performance results next week.