simplify decision tree for `\beamer@howtotreatframe`

[amonakov]

Expanding the frame environment relies on one of the following macros:

• \beamer@doseveralframes handles the "common case"

• \beamer@doexternalframe handles fragile frames, internally invoking the above macro when \frame is invoked recursively with fragile=false

• \beamer@dosingleframe handles fragile=singleslide frames without writing an intermediate file, so it doesn't invoke @doseveralframes and the frame cannot be reused via \againframe

• \beamer@autobreakframe handles allowframebreaks frames

• \beamer@donoframe typesets the frame in a temporary box, which is then discarded

Beamer used to choose between them by first branching on presence of a label, and when a label is NOT present, also by testing overlay specification to see if it yields no slides, via the following tree:

if (no label)
    if (overlay specification not empty)
        if fragile=singleslide
            \beamer@dosingleframe
        elif allowframebreaks
            \beamer@autobreakframe
        elif fragile
            \beamer@doexternalframe
        else
            \beamer@doseveralframes
        fi
    else # empty overlay specification
        \beamer@donoframe
    fi
else # has label
    if fragile
        \beamer@doexternalframe
    elif fragile=singleslide
        \beamer@dosingleframe
    else
        \beamer@doseveralframes
    fi
fi

This structure is confusing and leads to bugs such as #872 (allowframebreaks is not handled in presence of a label) and #873 (labeled fragile=singleslide frame shown despite <all:0> ospec).

It also uses \beamer@donoframe to suppress unlabeled slides, but it is inefficient (bug #866) compared to \beamer@doseveralframes which does not typeset the skipped content.

Use the following structure instead:

take \beamer@doseveralframes unless overridden below:
if allowframebreaks
    if ospec not empty
        \beamer@autobreakframe
    # else keep \beamer@doseveralframes
    fi
fi
if fragile
    \beamer@doexternalframe
fi
if fragile=singleslide
    if ospec not empty
        \beamer@dosingleframe
    else
        \beamer@donoframe
    fi
fi

Fixes #866. Fixes #872. Fixes #873.

[samcarter]

Thanks for your PR! As this is like an open-heart-surgery on beamer, I'll do a bit of testing before merging.

[josephwright]

I think this PR needs both @samcarter and I to read carefully. What we definitely should do is tidy up coding in such changes: for example, % should only be at the end of the lines where it's needed.

[samcarter]

There might be some difference in the handling of againframes, unfortunately the git history does not go back far enough to see why this was introduced...

[josephwright]

There might be some difference in the handling of againframes, unfortunately the git history does not go back far enough to see why this was introduced...

You reading Till's commits ;)

[samcarter]

There might be some difference in the handling of againframes, unfortunately the git history does not go back far enough to see why this was introduced...

You reading Till's commits ;)

The ones with *** empty log message *** are the most helpful ones :)

[amonakov]

Can you share your thoughts so far, please?

[samcarter]

My thoughts so far are that it seems that there is a difference how again frames are treated, but I haven't yet found an example where this causes a problem.

My proposal would be to merge this into the main branch and keep testing it until after the TL24 code freeze.

@josephwright What do you think?

[amonakov]

@josephwright What do you think?

Let me point out he thumbed up your response (Github doesn't send out notifications in such cases afaik).