Greg Smith <gsmith@gregsmith.com> writes:
> After spending some time assembling a list of special characters, I had an
> ah-ha moment when I realized they are all listed in the "Sections" section
> as "section title adornment characters":
> ! " # $ % & ' ( ) * + , - . / : ; < = > ? @ [ \ ] ^ _ ` { | } ~
I'll give you another ah-hah moment: that is exactly the list of all the
non-letter non-digit ASCII characters. (In order, even.) So this seems
to boil down to "if (ispunct(ch)) add backslash".
> I also note that there are some bullet and arrow inputs it will treat as
> special, see "Bullet Lists" in
> http://docutils.sourceforge.net/docs/ref/rst/restructuredtext.html
Ick.
> Seems reasonable to just document that non-ASCII input is somewhat
> perilious as a known limitation.
Perhaps so, especially in view of all the encoding dependencies that
would arise if we tried to fix that. Still, I'm getting a stronger and
stronger impression that ReST is a good standard to stay away from.
Somewhere along the line its designers lost track of the notions of
simplicity and predictability. It's probably peachy for hand-generated
text, but for machine-generated output there are way too many cute special
cases.
regards, tom lane