semanticate: i inside a tag is not a Roman numeral

[gvtulder]

se semanticate has a regexp that wraps lowercase i to a Roman numeral <span>. The regexp also does this to is inside attributes:

<a href="appendix-i.xhtml">

becomes

<a href="appendix-<span epub:type="z3998:roman">i</span>.xhtml">

This patch adds a lookahead clause to check that the i is not inside an open tag.

[alerque]

Shouldn't this sort of RegEx only be run on the content of tags via an XML parser that then iterates through the AST, not run directly on the raw XML?

[gvtulder]

Shouldn't this sort of RegEx only be run on the content of tags via an XML parser that then iterates through the AST, not run directly on the raw XML?

That might be better, yes. I noticed that the raw XML-regex also adds <abbr> tags inside alt attributes, for example, so this happens to other rules as well.

[acabal]

Shouldn't this sort of RegEx only be run on the content of tags via an XML parser that then iterates through the AST, not run directly on the raw XML?

Yes, but that becomes extremely complicated to implement extremely quickly. Regex is simple and fast, and has acceptable results 95% of the time for this narrow use case. When it does get something wrong, it's easy to undo.