Most efficient way to convert a run of Perl substitutions til boost::regex?

From: Anders Johansen (yg-boost-users_at_[hidden])
Date: 2003-03-24 15:08:44

Hi!

I'm converting a Perl program that basically consists of a succession of 
regex transformations of the input string, eg.:

#Case 1 - mutually exclusive endings
s/erens$/\@r\@ns_/;
s/ernes$/\@rn\@s_/;
s/endes$/\@n\@s_/;
s/estes$/\@st\@s_/;

#Case 2: - a load of manipulations done in sequence
s/jou/sju/g;
s/jau/sjo/g;
s/nch/ngk/g;

Now the question is, how do I convert this to boost::regex code that 
runs as fast as possible. I expect to have to call the function 
implementing these transformations (there are /many/) about half a 
million times at runtime, and as I'm concerned about runtime 
performance, I'd obviously like to use the most efficient method available.

I expect that Case #2 is simple, as I can use the regex_merge function, 
and nest the calls as in:

static const boost::regex re1("jou");
static const boost::regex re1("jau");
static const boost::regex re1("nch");
static const std::string fs1("sju");
static const std::string fs2("sjo");
static const std::string fs3("ngk");

std::string res = regex_merge(
regex_merge(
regex_merge(input_string, re1, fs1)
, re2, fs2)
, re3, fs3);

As to case 1, I guess I'd have to use a combination of the grep and 
format functions, and in the case as shown it'd be easy, as there is 
either one match or none for each line.

Any suggestions wrt. improving run-time performance?

Sincerely
   Anders S. Johansen