Codecov Report

Patch and project coverage have no change.

Comparison is base (8bcea81) 98.13% compared to head (51d04be) 98.14%.
Report is 7 commits behind head on master.

@@           Coverage Diff           @@
##           master    #3501   +/-   ##
=======================================
  Coverage   98.13%   98.14%           
=======================================
  Files          37       37           
  Lines        5536     5545    +9     
=======================================
+ Hits         5433     5442    +9     
  Misses        103      103           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@mortenpi Looks like it didn't make a difference. Should I try changing the example size threshold?

This PR

Last master

Should I try changing the example size threshold?

Nah, that's set to 1 KiB. But yeah, I think the issue is that the figures are using the show for text/html MIME, which we can't really write to a file..

Now there are a bunch of

That's progress! I think? The only pages that now seem to warn about being over the size threshold are the API pages and changelog!

but no supported image representation is present as an alternative.

These ones should be debugged -- why isn't there an alternative MIME for these ones?

@example block is above the threshold, falling back to 'image/svg+xml' representation.

I am not sure about the utility of these warning. My point is that you should be aware that we're picking a suboptimal representation due to size (otherwise users will complain "why is Documenter not making pretty plots"). But also, real-world docs like JuMP will have a lot of these. I think in an ideal world, I would like the user to somehow disable the text/html MIME if they can't use it anyway. But I am not convinced that this is the right way to go.

One option to improve this is to collect all of these and print just a single warning saying that N at-example blocks got replaced with mime X on pages y, z....

These ones should be debugged -- why isn't there an alternative MIME for these ones?

I think they're the DataFrames? There are no plots in geographic clustering

https://jump.dev/JuMP.jl/previews/PR3501/tutorials/linear/geographic_clustering/

One option to improve this is to collect all of these and print just a single warning saying that

Yeah that sounds good.

Although I'd also be happy with it just happening. In what situation is text/html a better representation than image/svg+xml?

I wonder if documenter could introduce something that we could pirate. Like default_mime_type(::Any) = MIME("text/html"), and then we could add something like Documenter.default_mime_type(::Plots.Plot) = MIME("image/svg+xml"). Pirating doesn't matter because this is only getting run once. I guess it could also be some sort of argument to Documenter.HTML().

I think they're the DataFrames?

Yeah, that would make sense. It looks like some of them are pretty big as well though, going over the 8K limit.

In what situation is text/html a better representation than image/svg+xml?

I did not check explicitly, but e.g. with PlotlyJS? Since you could get SVG + interactivity with text/html.

I wonder if documenter could introduce something that we could pirate. Like default_mime_type(::Any) = MIME("text/html"), and then we could add something like Documenter.default_mime_type(::Plots.Plot) = MIME("image/svg+xml"). Pirating doesn't matter because this is only getting run once. I guess it could also be some sort of argument to Documenter.HTML().

Yeah, something like that could be an option actually. I'll try to remember to open a follow-up issue with this.

This looks a bit better!

Yip. Just trying the excludes now.

Thoughts @mortenpi? I think this is good to merge

If you wait a day, you might be able to get rid of the Pkg.pkg"add Documenter#2fabe9c3bd40b6288c3d86d96c6031728b2d92ad" line too 😛 But yeah, I think this looks good to me!

If you wait a day

Cool cool. That would be good. So close!