The only time underscores aren't special in Markdown is when they appear
in preformatted text. We have gotten away with not escaping underscores
where an asterisk-enclosed span or the paragraph ends before the next
underscore appears, but this is fragile and bad practice. Conversely,
there are many places where we have not escaped underscores but needed
to.
Escape all underscores that do not appear in preformatted blocks or
preformatted spans and are not themselves delineating emphasized spans.
The changes in this commit are exactly the result of executing the
following Bash code:
```bash
e=':x;' # begin loop
e+='s/^' # anchor match at beginning of line
e+='(' # begin capturing subexpression
e+='(' # begin list of alternatives
e+='[^`_\\]|' # any mundane character, or
e+='`([^`\\]|\\.)*`|' # backtick-enclosed span, or
e+='\b_|_\b|' # underscore at boundary, or
e+='\\.' # backslash-escaped character
e+=')*' # any number of the preceding alternatives
e+=')' # end capturing subexpression
e+='\B_\B/\1\\_/;' # escape non-formatting underscore
e+='tx' # repeat loop if we escaped an underscore
escape_underscores=(
sed
# use extended regular expressions
-E
# skip over indented blocks (following an empty line)
-e '/^$/{:i;n;/^( {4,}|\t)/bi}'
# skip over preformatted blocks
-e '/^\s*```/,/^\s*```/{p;d}'
# skip over generated sections
-e '/GENERATE-FROM-SCHEMA-START/,/GENERATE-FROM-SCHEMA-END/{p;d}'
# escape underscores
-e "${e}"
)
"${escape_underscores[@]}" -i doc/*.[0-9].md
```
Changelog-None
Matt Whitlock
Rusty Russell