"Eric Niebler" <eric@boost-consulting.com> wrote in message news:413A5044.6020604@boost-consulting.com...
Jonathan Turkanis wrote:
- Have you considered allowing format strings to be bound statically?
That hadn't occured to me. Certainly possible, but I think the returns would be small. The complexity of doing string replacements is linear, and so is not as sensitive to optimization as pattern matching. And the format string has very simple syntax, so static syntax checking isn't as big of a win here, either.
- Have you considered using _1, _2, _3 instead of s1, ... for capturing paretheses? You could also use them as placeholders in statically-bound
Makes sense. format
strings.
Yes, I used _1, _2, _3 for a long time. I abandoned it because of name conflicts with the identically-named placeholders from other boost libraries.
I know this problem well :(
I picked s1, s2, ... for two reasons: 1) The "s" in "s1" stands for "sub-match", which is what these thigs represent. 2) s1 kind of looks like $1, which is the perl equivalent.
I didn't think of 2). Did you put in in the docs? FWIW, capital 'S' looks more like '$' to me. E.g., '<' >> (S1= +_w) >> '>' >> -*_ >> "</" >> S1 >> '>'
That said, I'm open to suggestions for avoiding the name conflicts. I would consider switching back to _1 _2 _3 if the technical problems were overcome and if people liked it better.
May I assume you have considered reusing the placeholders from boost::bind? There doesn't seem to be much operator overloading involving boost::arg<>.
- have you considered overloading operator/ to emulate perl syntax for substitution. E.g.,
s/expr/fmt, sub/expr/fmt, or just expr/fmt
could return a function object with templated operator which invokes regex_replace.
A little too cute, IMO.
Oh well ;-) Jonathan