consider a warning when secheadings are created

[pkra]

I was revisiting AmerMathSoc/texml-to-html#439 (cf. also #194) which made me realize that a secheading seems to be a reliable indicator of problematic authoring.

Some examples (from AmerMathSoc/texml-to-html#439 (comment))

• splitting a very large proof into sections - in which case it should be a section instead of a proof environment (where further sectioning would of course be fine). This usually coincides with a level skip (likely because the proof is considered a sectioning element).
• accidentally nested sectioning in a theorem
• abusing theorems as sections to hack counters but keeping sections within
• things that ought to be nested theorems (e.g,. a claim with subproof inside a proof)
• emphasis (e.g., "From now on"...)

Additionally, sectioning levels are often chosen for visual appearance (e.g., a section in a proof in a subsection) and often skip levels (likely because the proof is considered a sectioning element, cf. above).

We cannot do anything about them downstream at this point. But it might be worthwhile to flag these so they can be reviewed/fixed early on. They don't appear very often, most of them have obvious fixes (i.e. make the outer proof a section or make the inner section a theorem env).

[pkra]

For the record, I currently find secheading in ~3% of publications with ~6 occurrences per publication. More than 90% of them are in proof environments. The only other noticeable environment names are "step"/"case" (~5%).