implement end-to-end-tests

[a-maldet]

It would be great to have End-To-End-Tests for the package.
The results could be compared by a image distance measurement in order to classify success or failure.

[GregorDeCillia]

After some googling, these packages could be useful for measuring image similarity

• colordistance
• RNiftyReg::similarity

It should be enough to test the outputs of write_png as a first step.

For xlsx files, we could go down the rabbit hole with the headless version of libreoffice. However, the headless testing can wait until after the CRAN release.

[GregorDeCillia]

Seems like we can even make a bitwise comparison with png files. This doesn't work with pdf though.

library(styledTables)

st <- styled_table(mtcars)
write_png(st, "ex1.png")
write_png(st, "ex2.png")

tools::md5sum("ex1.png")
#>                            ex1.png 
#> "d6611369188488028b862cda981136ca"
tools::md5sum("ex2.png")
#>                            ex2.png 
#> "d6611369188488028b862cda981136ca"

Testing against hashes would be nice since

• it avoids a repository bloat and
• it doesn't create any additional dependencies.

On the other hand, this test is very strict and could lead to incorrect failures

[GregorDeCillia]

Update: Bitwise comparison fails if the testing environment is not the same as the environment generating the checksums. I get different checksums for ubuntu 16.04 and ubuntu 18.04.

In other words, we need a similarity measure for png files in order to run the checks on travis.

[GregorDeCillia]

I am not sure if it makes sense to put much effort into this for the CRAN release since it is currently not clear how they deal with the CTAN dependencies. Therefore, I'll remove the CRAN label for this issue for now. @a-maldet feel free to disagree.