Bug Fixes

Improvements

  • record_session() gained a new variable connect_api_key to be able to record a session against RStudio Connect using a Connect API key. Please see Load Testing Authenticated Apps for more details. Using a Connect API key also requires shinycannon >= 1.1.0 (#133)

  • load_runs() now uses vroom instead of read_csv(); this substantially improves its performance.

  • The homepage has been rewritten to get to the big picture more quickly. Most of the details about shinycannon have moved to a new article.

  • The analysis vignette has been overhauled to get into the plots more quickly, and provide more advice about interpretation.

  • Plots have be thoroughly overhauled:

    • Waterfall plot now only shows plots in maintenance period, and times are session relative.

    • Sessions in warmup and cooldown are no longer shown in muted colours, because that unfortunately relied on a ggplot2 bug. Instead they’re simply omitted from the plot.

    • The line of best fit in the latency plots has been made less visually prominent, and no longer shows standard errors.

This is the initial release of shinyloadtest to CRAN.

Bug fixes

  • Server type detection is now tolerant of invalid HTML/XML markup produced by the target application (#115, rstudio/shinycannon#38)
  • Fixed a bug that prevented runs from being loaded (#124, #125)

Improvements

  • The application server type is now captured in recording files. A corresponding enhancement was made to shinycannon that will produce a warning if the recording server type differs from the target (#107), rstudio/shinycannon#36)
  • The way URLs are dealt with internally is now more robust, and should be tolerant of a wider variety of Shiny app deployment scenarios (#119).