From Andrew Patterson:
With respect to ADL quoting rules...
...what should the behaviour be when a
\ is used without a valid 'quotation' character following i.e. \\
and \" should be interpreted as \ and " respectively, but should
\. be interpreted a one character . or two characters \.? I'd suggest
that if the quotation character is not recongised, then it should
be an error.
Furthermore, what are the rules around the &#xHHHH escape
sequence? Surely the & will have to be quoted as well? Otherwise
how will the parser know when it is being used for a mnenomic?
Wouldn't a unicode escape technique such as used in Java and C#
(\uAABB) be a better approach (i.e. one that fits in more
comfortably with the standard \" quoting rules).
Finally, what about the other standard string escapes used
in Java and C# (\t, \b etc). Is there any room for them (maybe
in ADL 2.0?)
The basic problem here is to have a consistent way of doing the following:
- quoting literals like the " and ' characters (currently uses a backslash)
- allowing other quoted characters such as \r, \n, \t etc - what is the list?
- quoting unicode characters. Is this needed given that real unicode can be used?
It may be needed for systems and transmission in situations where unicode (or
any non-ascii) cannot be tolerated. But we probably want a way of differentiating
between proper unicode files and ASCII fles containing quoted characters.