Bug Fixes

• Fixed#168: Remove deprecated usage ofdplyr::summarise duringshinyloadtest_report.

• Fixed#163:gtable_trim error duringshinyloadtest_report with newer versions of ggplot2.

Full Changelog:v1.1.0...v1.2.0

shinyloadtest 1.1.0

• record_session() gained a new variableconnect_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 requiresshinycannon >= 1.1.0
  (#133)

• load_runs() now uses vroom instead ofread_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 release included an addin and an R Markdown report. The release took the approach of a single load test, given the # of expected concurrent users, and gave metrics based on the successful number of connections.

v0.2 will be a major re-write, changing the approach to a time window wheren connections are continuously attempted.