The thing is that for most characters, \N markers don't behave quite like normal identifiers, e.g.
//echo -a $regsubex(ab,/(a)(b)/,<\1-+\2>)
echoes "<a-+b>" which means \1 and \2 are evaluated, even though they touch <, >, - and +.
Judging by the results of this: //echo -a $regsubex(a,/(a)/,$bits\1$bits)
which echoes "32a32"
and this: //var %z$1 = TEST | echo -s $regsubex(a,/a/,%z [\1 ])
which echoes "TEST" (meaning "%z [\1 ]" was interpreted as "%z [ $+ $1 ]")
it looks like mirc internally replaces strings like "foo\1bar" with something like "foo $+ $1 $+ bar". Can't this be made to work when foo = ( and bar = ) ?
I suspect (and this is just a guess) that ( and ) are excluded from this treatment because of their special role as parameter list delimiters in identifiers, but stuff like $upper( $+ $me $+ ) work fine already, so this special treatment may not be necessary
Last edited by qwerty; 04/08/06 12:46 PM.