iRule regex to normalize phone numbers?

OK, two steps got it to work.

Code
when  ACCESS_POLICY_AGENT_EVENT {
       expr srand([clock clicks])
       set otp [string range [format "%08d" [expr int(rand() * 1e9)]] 1 6 ]
       set mail [ACCESS::session data get "session.ldap.last.attr.mail"]
       set mobile [ACCESS::session data get "session.ldap.last.attr.mobile"]
       set mobile2 [string map {"." "" " " "" "-" "" "(" "" ")" "" "+" ""} $mobile]
       ACCESS::session data set session.user.otp.pw $otp
       ACCESS::session data set session.user.otp.mobile $mobile2
       ACCESS::session data set session.user.otp.username [ACCESS::session data get "session.logon.last.username"]
}

Hi Kevin,

That worked as well, thanks for throwing that out there as it makes me realize the order can make a big difference there.

One more thing. Because you originally asked about a regex function:

% set num1 "1234567890"
% set num2 "123-456-7890"
% set num3 "123.456.7890"
% set num4 "123 456 7890"
% set num5 "(123) 456-7890"
% set num6 "123foo456bar7890"

% time { regsub -all {\D} $num1 "" newval; puts $newval }
1234567890
26 microseconds per iteration

% time { regsub -all {\D} $num2 "" newval; puts $newval }
1234567890
29 microseconds per iteration

% time { regsub -all {\D} $num3 "" newval; puts $newval }
1234567890
31 microseconds per iteration

% time { regsub -all {\D} $num4 "" newval; puts $newval }
1234567890
37 microseconds per iteration

% time { regsub -all {\D} $num5 "" newval; puts $newval }
1234567890
32 microseconds per iteration

% time { regsub -all {\D} $num6 "" newval; puts $newval }
1234567890
35 microseconds per iteration

% time { puts [string map {"." "" " " "" "-" "" "(" "" ")" "" "+" ""} $num3] }
1234567890
45 microseconds per iteration

The "\D" regex option used in regsub will replace ANY non-numeric character (see num6), so it'll technically do a better job, and is more flexible than the string map version.

You can see the penalty you pay for using regex. If stripping all non-numeric is the goal, better off with the proc.

Indeed:

% set num1 "123-456-7890"

% time { regsub -all {\D} $num1 "" } 10000
3.172 microseconds per iteration

% time { string map {"." "" " " "" "-" "" "(" "" ")" ""} $num1 } 10000
1.1674 microseconds per iteration

So then it's really a trade-off between speed and flexibility. Regex will catch more anomalies (ie. "123foo456bar7890") at the expense of CPU time. That said, the goal here was convert a string during access policy evaluation, which should only happen once per session.