ad_proc reform:
- use nsf::proc for defining ad_procs, when nsf::proc is available
- nsf::proc is part of the environment of XOTcl2 and provides
an efficient, C-implemented argument parser that can replace
the scripted *__arg_parser of OpenACS.
- The benefits are performance and reduced memory consumption.
* Simple functions (where the calling overhead was a large
part of the execution times can become several times faster
(e.g. db_name is 3x faster, cookie handling is 2x faster),
for functions with much db-activities the percentage will
be less.
* The memory consumption (RSS) of a site like OpenACS.org
is reduced by about 15%.
NEW (with nsf::proc)
::ad_get_cookie x: 10.93 microseconds per iteration
::ad_get_cookie ad_session_id: 21.95 microseconds per iteration
::dt_systime: 24.89 microseconds per iteration
::db_name: 8.70 microseconds per iteration
::db_string dbqd.null.g "select version() from dual": 316.66 microseconds per iteration
::ad_convert_to_html "hello world": 102.29 microseconds per iteration
OLD (without nsf::proc)
::ad_get_cookie x: 19.42 microseconds per iteration
::ad_get_cookie ad_session_id: 30.10 microseconds per iteration
::dt_systime: 27.55 microseconds per iteration
::db_name: 32.28 microseconds per iteration
::db_string dbqd.null.g "select version() from dual": 366.69 microseconds per iteration
::ad_convert_to_html "hello world": 160.51 microseconds per iteration
======================================================================
proc mtime cmd {set c 5000
set t [time [list time $cmd $c]];
regexp {^(-?[0-9]+) +} $t _ mS1;
set ms [expr {$mS1*1.0/$c}];
return "[format %6.2f $ms] microseconds per iteration"
}
set tests {
{::ad_get_cookie x} {::ad_get_cookie ad_session_id} {::dt_systime}
{::db_name} {::db_string dbqd.null.g "select version() from dual"}
{::ad_convert_to_html "hello world"}
}
foreach t $tests {append ::_ $t ": " [mtime $t] \n}
set ::_
======================================================================
