Re: How to rewrite a regexp without word boundaries?

> > I have been wondering, with limited success, how to rewrite a regexp
> > without word boundaries.
>
> Why do you want to? Most likely, the answer is that your regexps are
> getting too clever and thus too unreadable/bug-prone, so you should
> break them up and use more ordinary programming instead.


The regexps are not mine... Sorry, I should have explained. I am
actually writing a tool that takes regexps as input and transforms
them internally into NFAs/DFAs. Since the regexps are not really in my
hands, I should be ready for weird regexps - for example, regexps with
"\b" preceded or followed by other regexps. And I don't know how to
transform a regexp that contains "\b" at an arbitrary position into an
equivalent NFA/DFA.




> > Now, if "\b" occurs at the beginning (or end) of the regexp, I think
> > it's easy to rewrite the regexp without using "\b". For example,
> > "\bex" could be rewritten as "\Wex".
>
> No, that would not match "ex" at the beginning of the string being
> matched. You could use (?:^|\W)ex.


You are right. My mistake.




> But the matched substring (Perl's
> $&) will differ from with \bex, so it depends on how the regexp is used.


I am not interested in the matched substring, so "(?:^|\W)ex" (or
rather the NFA/DFA corresponding to "(?:^|\W)ex") is fine... in this
particular case.




> > But what if "\b" occurs within the regexp? For example, how to get rid
> > of "\b" in "<RE>\bex" (with "<RE>" being any regexp)? "<RE>\Wex"
> > wouldn't work here: for example (with "<RE>" = "\W"), "\W\Wex" is not
> > equivalent to "\W\bex".
>
> Rewrite <RE> to a regexp which ends with \W or \W|^ and is equivalent
> in the cases where it is followed by \b\w.


Hmm, not so simple in the general case, but I will have to think about
this possibility.




-- dave

Return to the comp.compilers page.
Search the comp.compilers archives again.