#Testing generated pdf

8 messages · Page 1 of 1 (latest)

frozen shale
#

I'm using typst to generate a pdf , I want test if the pdf generated is generated as expected or not.
I'm thinking of having a pre generated correct sample pdf , which I can compare with the generated report and know if the pdf generated is correct or not...
I'm not yet sure how to achieve this , but if you have some good idea on how to test thhe pdf , please shahre.

grave garden
#

depends - what is "correct"?

If you want complete equality, then comparing the md5sum of two documents would work. I usually use this strategy in my libraries, but in a somewhat manual way:

  • I have some examples in a gallery/ folder
  • after a change to the library, I regenerate the example pdfs
  • if git status reports that any PDFs have changed, I know that the change I did was not a clean refactoring; if that was unexpected, I will investigate
    generally, this works if your documents use set document(date: ...) (so that the timestamp is repeatable) and as long as the Typst version doesn't change. It may fail for more visual features, where what counts is not byte-for-byte equality.

If you want a visual comparison, you can have a look at typst-test. It doesn't wok on PDFs but on pngs and does a visual comparison, exactly like you described. In my experience, it also requires the Typst version not to change to work properly.

frozen shale
#

Thanks for the reply @grave garden ...
I cannot use typst-test as of now , since it's not stable as of now.

for the first approach I was trying something similar by generating hashes of both the files and comparing them , but that was failing due to meta data features like time etc

grave garden
frozen shale
#

no that's not me

grave garden
#

well, what a coincidence 😅 anyway, is there still a problem with the metadata or does it work now? if the document is not reproducible, I'd say that's considered a bug.

frozen shale
#

that'll be much more reliable