Don't change document graphics state when rendering with option dry_run in Text::Formatted::Box

[ingk]

Rendering a Text::Formatted::Box with the dry_run option led to output in the rendered PDF.

[practicingruby]

It looks like this behavior change is desirable, but I'm not comfortable introducing another instance variable here. Maybe pass an optional value into process_options instead?

[ingk]

You're right. I changed it to an optional parameter.

[practicingruby]

We probably don't need this line, as flags[:dry_run] can be used directly in the two other places it is mentioned (213, 224)

[practicingruby]

There are many test failures on this pull request... please investigate those and let me know when the build is green again.

[ingk]

I just updated the pull request and now the build passes without errors.

[practicingruby]

From the looks of it, this will mutate the input data, meaning you can't safely pass the same array to the method twice (i.e. once for a dry run and then again to actually render).

Please work on a solution that does not mutate the inputs.

[ingk]

I don't know if I get you right, but the variable text gets a copy of the input data.

def original_text @original_array.collect { |hash| hash.dup } end

normalize_encoding also relies on original_text. So deleting the key color shouldn't be a problem, since in the next iteration without the dry_run option we get a fresh copy of the original input data.

[practicingruby]

I see, we do copy the input arguments. This is surprising because original_text is actually a copy, not the original text. 😦

I'm still not sure how good I feel about this solution, though. It's extremely implicit behavior, and so I'm worried about breakage in future changes to this code. Really we should be having explicitly defined behavior for dry run vs. actual rendering, and not adding various checks for if flags[:dry_run] throughout the code.

This isn't your fault, because it is how the code is designed currently. But I think we should give this more thought before merging.

[ingk]

I get your point. Is there anything I can do here? Unfortunately I'm not that familiar with the codebase to do such a change.

[practicingruby]

Sorry for the delay in response. I'd say if this patch is working for you, just use your fork until we get something similar merged upstream.

[practicingruby]

I looked into fixing the underlying design issues for the dry run mechanism, but it's really complicated because the entire formatted text system is pretty tangled up.

So as an alternative I basically applied your same patch in dd92069, but moved the code into a helper method just to make it less likely to be lost or broken during future refactorings.

[practicingruby]

@ingk: Because your pull request was accepted, you now have commit access to Prawn. Please see our contributor guidelines, and happy hacking!