theoremstyle "plain": differences amsmath/legacy/new block code

[u-fischer]

Theorems in "plain" style differ between standard LaTeX and amsmath: amsmath adds a period and amsmath ignores \partopsep:

\documentclass{report}
%\usepackage{amsthm}
\newtheorem{theorem}{Theorem}
\newtheorem{lemma}[theorem]{Lemma}

\begin{document}
\partopsep=20pt
xxxxxxxxxxx
\begin{theorem}\label{theorem1}some text
\end{theorem}
\begin{lemma}\label{lemma1}some text
\end{lemma}
\end{document}

standard latex:

with amsthm:

with block code

The block code currently produces by default a mix of these settings: a period is added but \partopsep is not ignored

To reproduce the legacy settings, the block code should use an empty punct by default and reset it and begin-extra-vspace if amsthm is loaded:

\DocumentMetadata{}
\documentclass{report}
\IfDocumentMetadataT{
  \EditInstance{thmstyle}{plain}{punct=}
  \AddToHook{package/amsthm/after}{%
    \EditInstance{thmstyle}{plain}{punct=.}
    \EditInstance{block}{thm-plain-1}{begin-extra-vspace=0pt}
    \EditInstance{block}{thm-plain-2}{begin-extra-vspace=0pt}}}
\usepackage{amsthm}
\newtheorem{theorem}{Theorem}
\newtheorem{lemma}[theorem]{Lemma}

\begin{document}
\partopsep=20pt
xxxxxxxxxxx
\begin{theorem}\label{theorem1}some text
\end{theorem}
\begin{lemma}\label{lemma1}some text
\end{lemma}
\end{document}

[FrankMittelbach]

hmm. the value with amsthm are much saner than those of pure 2e. And so I would like to have thosae as default when \DocumentMetadata is used. I don't think it is a good idea to require that in new documents people use amsthm just to get these defaults. That means one would need to have some compatibility flag that can be used for old documents to get the old style even if tagging is turned on.

I think I will provide some \theoremstyle{latex2e} that implements the old settings and worry how that gets enabled later when we start to implement compatibility setups.

[u-fischer]

I'm fine with deviating from legacy values if they are saner, but we should document such changes. And it would be good to have some option to force legacy behaviour as it makes it easier to test for unwanted differences if there aren't other changes around.

[FrankMittelbach]

I'm fine with deviating from legacy values if they are saner, but we should document such changes. And it would be good to have some option to force legacy behaviour as it makes it easier to test for unwanted differences if there aren't other changes around.

agreed, but that was what I was trying to say too. The legacy would be available using \theoremstyle{latex2e} and how that it issued would be up to a legacy strategy that we have to come up with (I still favor some kind of compatibility={A,B,C...} in \DocumentMetadata where you can force certain compatibility options A B C one of which could be thmstyle). But that is a more general strategy decision.

[FrankMittelbach]

The plain theoremstyle now matches what amsthm is doing and in addition I added 'legacy2e' as atheoremstyle matching what LaTeX2e does when \theoremstyle is not provided (by amsthm). How this is managed when processing old documents with compatibility setting needs some decision, but that can happen elsewhere and later.