Heuristic for scanning formulas is faulty

[vstange]

The scanFormulaNode can't recognize formulas not containing "apply"- or "mrow"-nodes. The mistake is based on my false assumption that there should always be a surrounding element.

Instead, the method should examine the semantic elements more closely.

[physikerwelt]

@AndreG-P did you already solve this problem. I vaguely remember that you had a similar issue?

[AndreG-P]

Actually, you did... It was a bug in vmext: gipplab/vmext#23

[physikerwelt]

@vstange can you follow the approach from https://github.com/ag-gipp/vmext/pull/25/files#diff-c9cf159d96357b7583ac5bc9f33564adR45

[vstange]

Noted.

[physikerwelt]

can this ticket be closed @vstange ?

[AndreG-P]

@physikerwelt I'm not sure what the problem is. The code already inspects nodes more closely based on
https://github.com/ag-gipp/MathMLTools/blob/00c269d0c313b317a5b5333434b6983157461a11/mathml-converters/src/main/java/com/formulasearchengine/mathmltools/converters/MathMLConverter.java#L232-L245

For vmext you implemented:

    const content = this.select('//m:annotation-xml[@encoding="MathML-Content"]', this.doc, true);
    if (!content) {
      throw new Error("No content MathML present");
    }
    return this.parseApply(content.firstChild);

As you can see, the method above already checks the same information. So either you can tell me what is broken or give me an example that does not work.

[physikerwelt]

It is probably an old issue that had been resolved before