tladeras’s avatartladeras’s Twitter Archive—№ 1,894

                1. Joe Cheng talking about using Shiny in production by looking at “whales” who download lots at RStudio CRAN mirror #Rstudioconf
                  oh my god twitter doesn’t include alt text from images in their API
                  Permalink On twitter.com Twitter logo tladeras’s avatar ♻️ 1 Retweets ❤️ 5 Favorites Mood 0
              1. …in reply to @tladeras
                Production environments: relied on by real users with real consequences if they go wrong. Goals are uptime, security, correctness, responsiveness. #Rstudioconf
                Permalink On twitter.com Twitter logo tladeras’s avatar Mood -2 🙁
            1. …in reply to @tladeras
              Can shiny be used in production? Yes, possible and now easy. Challenges are organizational, culural and technical. #Rstudioconf
              Permalink On twitter.com Twitter logo tladeras’s avatar ❤️ 1 Favorite Mood +2 🙂
          1. …in reply to @tladeras
            Cultural: many app designers are users and don’t necessarily know best practices, performance and scale testing, security testing. #Rstudioconf
            Permalink On twitter.com Twitter logo tladeras’s avatar Mood +3 🙂
        1. …in reply to @tladeras
          Organizational: IT/management are skeptical, credibility gap. Engineering department may be biased against R. #Rstudioconf
          Permalink On twitter.com Twitter logo tladeras’s avatar ❤️ 1 Favorite Mood -4 🙁
      1. …in reply to @tladeras
        Technical: hard to debug and profile , single threaded execution. #Rstudioconf
        Permalink On twitter.com Twitter logo tladeras’s avatar Mood -1 🙁
    1. …in reply to @tladeras
      shinytest, shinyloadtest for testing and profiling apps, profvis (R Profiler), plot caching, async are tools to tackle technical challenges #Rstudioconf
      Permalink On twitter.com Twitter logo tladeras’s avatar Mood 0
  1. …in reply to @tladeras
    shinyloadtest to see how fast apps are. Use profvis to improve R processes. Improve by moving work out of shiny, make code faster, caching, async. Recheck in shinyloadtest. #Rstudioconf
    On twitter.com Twitter logo tladeras’s avatar ♻️ 1 Retweets ❤️ 1 Favorite Mood +4 🙂
    1. …in reply to @tladeras
      Shinyloadtest: record average app use case, run concurrent load tests on shiny app. shinycannon runs concurrent load tests, produce report to understand performance. #Rstudioconf
      Permalink On twitter.com Twitter logo tladeras’s avatar Mood 0
      1. …in reply to @tladeras
        How to improve? Use profvis to id bottlenecks in your app. Cranwhales spends too much time on loading in app. Load/process data ahead of time. Save as feather files for quick loading. Automate this using scheduling. #Rstudioconf
        Permalink On twitter.com Twitter logo tladeras’s avatar ❤️ 1 Favorite Mood +4 🙂
        1. …in reply to @tladeras
          Good candiates are slow plots, and plots that users frequently request. Add caching in renderPlot and tell shiny what input variables matter #rstudioconf
          Permalink On twitter.com Twitter logo tladeras’s avatar Mood +4 🙂
          1. …in reply to @tladeras
            Er, renderCahchedPlot #rstudiocomf
            Permalink On twitter.com Twitter logo tladeras’s avatar Mood 0
            1. …in reply to @tladeras
              I can’t spell or my thimbs are too large. #rstudioconf
              Permalink On twitter.com Twitter logo tladeras’s avatar Mood 0
              1. …in reply to @tladeras
                @kellrstats working on shiny in production book! bit.ly/shiny-prod-book #rstudioconf
                Permalink On twitter.com Twitter logo tladeras’s avatar ❤️ 4 Favorites Mood 0
        2. …in reply to @tladeras
          After this fix, max completion time goes way down. ETL (extract load transfer) speeds things up. Can we do better? ggplot2 is now bottleneck. Plot caching can save result and serve it up in shiny 1.2 #Rstudioconf
          Permalink On twitter.com Twitter logo tladeras’s avatar ❤️ 2 Favorites Mood +4 🙂