0

I have created a markdown script which is producing some ggplot graphs. I also wrote a test script to automatically testing the whole markdown by first extracting the code chunks and then executing the script.

With some test data, I get warnings like Removed 1 rows containing missing values (geom_point) which I normally ignore and since updating testthat to 1.0.0, it is complaining about these warnings because I don't "expect" them.

runAllChunks <- function(rmd, envir=globalenv()){
  # as found here http://stackoverflow.com/questions/24753969
  tempR <- tempfile(tmpdir = '.', fileext = ".R")
  on.exit(unlink(tempR))
  knitr::purl(rmd, output=tempR, quiet=TRUE)
  sys.source(tempR, envir=envir)
}

expect_that(f <- runAllChunks('analysis.Rmd'), not(throws_error()))

testthat is reporting the following:

testing analysis script: WWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWW1.............

Warnings ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1. analysis works with "nurias_testdata" folder. (@test_analysis.R#48) - Removed 1164 rows containing non-finite values (stat_boxplot).
... truncated here ...

In older versions of testthat I got rid of the warnings by saving the result into f. But now testing throws warning messages and also the not is deprecated. I changes this to check for expect_success, but with the same result.

expect_success(f <- runAllChunks('analysis.Rmd'))

So not sure how to test my script and take care of the warnings. I wouldn't like to suppress all warnings inside the markdown, also to always filter before plotting is a bit exhausting.

Update

Thanks to comment ov bVa I fixed my geom_* sections in the plot to include na.rm=TRUE and got rid of the messages.

I am now just wondering, if I expect multiple messages or warning while testing a markdown script, how can I check for those without running the markdown chunks several times? Is there a way of doing this?

I am asking because I have now still a warning message I expect, but I would also would love to check for success with expect_success. Or is the way to go by using complicated regex like 

expect_warning(f <- runAllChunks('analysis.Rmd'), 'warningA|warningB|warningC')
halfer
  • 19,824
  • 17
  • 99
  • 186
drmariod
  • 11,106
  • 16
  • 64
  • 110
  • 1
    If the warnings are only related to missing data, just add a `na.rm = T` in the arguments of `geom` layers. – bVa Sep 16 '16 at 07:21
  • Thanks for the hint with `na.rm=TRUE`! I changed this and there is now only one expected warning. How can I check for success AND a warning? Or would be `expect_warning` sufficient since it will complain if it is not successful? – drmariod Sep 16 '16 at 08:45

0 Answers0