just saw that CI failed. parts of it are unrelated to my changes I think: https://github.com/TalkingCatSW/MathCAT/actions/runs/21458974397/job/61806814200#step:3:175

error: called `unwrap` on `operator_versions.postfix` after checking its variant with `is_some`
    --> src/canonicalize.rs:3561:12

Using a feature is a good way to go.

I'll need to look into the error you got.

For the rule coverage, I think you are only recording that a rule is used, not which parts of it gets used. Recording that info is definitely better than nothing, but it isn't saying (for example) whether the "then" branch or the "else" is used. I didn't look in close detail, but maybe that is why the line coverage is so low?

yep, this not treating rules on line-level.
I'll look at that at some later point.

MathCAT's rule interpreter is basically just a recursive descent interpreter. Most things will have a build and a replace function. The replace functions are where you could add some instrumentation. For example, The TestArray class's (- test:) replace function is

    fn replace<'c, 's:'c, 'm:'c, T:TreeOrString<'c, 'm, T>>(&self, rules_with_context: &mut SpeechRulesWithContext<'c, 's,'m>, mathml: Element<'c>) -> Result<T> {
        for test in &self.tests {
            if test.is_true(&rules_with_context.context_stack.base, mathml)? {
                assert!(test.then_part.is_some());
                return test.then_part.as_ref().unwrap().replace(rules_with_context, mathml);
            } else if let Some(else_part) = test.else_part.as_ref() {
                return else_part.replace(rules_with_context, mathml);
            }
        }
        return T::from_string("".to_string(), rules_with_context.doc);
    }

If a count were kept, you would be able to record the 'i'th then/else was taken (if then ... else_if ...then ... else_if ...). I don't know how you relate that back to the line number though.