Recently a discussion arose concerning a way of determining how many
occurances of a character appeared in a string.  Several methods were proposed:
   * [regexp]
   * [split]
   * [foreach]/[split]/[incr]
   * [string map]

 #! /usr/tcl84/bin/tclsh

 set mystring "line1\nline2\nline3\nline4\n"

 proc splitline {string} {
        set rc [llength [split $string "\n"]] 
        incr rc -1
        return $rc
 }

 proc regline {string} {
        set rc [regexp -line -all "\n" $string]
        return $rc
 }

 proc splitchar {string} {
 foreach x [split $string ""] {if {[catch {incr counters($x)}]} {set counters($x) 1}}
        set rc $counters(\n)
        return $rc
 } 

 puts [time {splitline $mystring} 10000]
 puts [time {regline $mystring} 10000]
 puts [time {splitchar $mystring} 10000]


On my SPARC Solaris, using Tcl 8.4, I get

 49 microseconds per iteration
 154 microseconds per iteration
 963 microseconds per iteration

as the result. - [RS]: Surprising, but good to know, that [split] is by far the fastest way...

[Brett Schwarz] added this proc as well:

 proc strmap {str} {
 return [expr {[string length $str]-[string length [string map {"\n" ""} $str]]}]
 };

on my SuSE Linux 7.3 machine (2X333MHz PII) with Tcl 8.4, I get these results:

 34 microseconds per iteration
 87 microseconds per iteration
 782 microseconds per iteration
 34 microseconds per iteration

So, it gives the same performance of the split version. However, if the string length is increased, to lets say 100 lineX's (i.e. 100 \n), then this proc using [string map] actually scales better. Here are the numbers with 100:

 263 microseconds per iteration
 1729 microseconds per iteration
 2573 microseconds per iteration
 141 microseconds per iteration

There is quite an improvement....