Making Timing More Easy

[Cecil Westerhof]

Because of the discussion about what would be more efficient for my
proc splitOnWhiteSpace:
http://wiki.tcl.tk/55362

I decided to write a proc to make timing more easy to do:
http://wiki.tcl.tk/55368

--
Cecil Westerhof
Senior Software Engineer
LinkedIn: http://www.linkedin.com/in/cecilwesterhof

[Mike Griffiths]

What would you say the benefit is of this over just using [time]?

[Cecil Westerhof]

Mike Griffiths <mi...@keyboardzombie.com> writes:

> On Saturday, 9 June 2018 10:59:05 UTC+1, Cecil Westerhof wrote:
>> Because of the discussion about what would be more efficient for my
>> proc splitOnWhiteSpace:
>> http://wiki.tcl.tk/55362
>>
>> I decided to write a proc to make timing more easy to do:
>> http://wiki.tcl.tk/55368
>

> What would you say the benefit is of this over just using [time]?

That I can do something like:
set regexTime [doTime "splitOnWhiteSpace {${currentStr}}" \
${iterations} ${tries}]

and if I only want the average and no output:
set regexTime [doTime "splitOnWhiteSpace {${currentStr}}" \
${iterations} ${tries} False]

and because I want to know the timings of both I also do:
set tclTime [doTime "splitOnWhiteSpace {${currentStr}} -1 -1 True" \
${iterations} ${tries}]


For one reason or another I really like DRY.

[heinrichmartin]

On Saturday, June 9, 2018 at 11:59:05 AM UTC+2, Cecil Westerhof wrote:
> I decided to write a proc to make timing more easy to do:
> http://wiki.tcl.tk/55368

> As always: comments, tips, questions and requests are appreciated.

[time] accepts zero iterations, and Tcl built-ins were improved to "do nothing gracefully"[1]. There is no reason to bail out.

The output of [time] already is an average. There is no reason to introduce more passes (I'd prefer this term over "tries" here.). If you liked more convenience, then provide [expr] in your call to [time $code [expr $iterations]]. This allows invocation [doTime $code 1000*5].

My third comment is about the context. You could do sth like

[uplevel 1 [list time $code [uplevel 1 [list expr $iterations]]]] ;# untested

I.e. resolve everything in the caller's context. This allows using doTime from others than the global context. And it allows using variables in $iterations.

# untested
namespace eval foo {
variable iterations 1000
variable passes 5

proc bar args {# do bar}

proc time_bar {} {
variable iterations
variable passes

# the uplevel of expr is not even needed for this use case ...
doTime bar $iterations*$passes

# ... but for this fancy code.
doTime bar {$iterations*[incr passes]}
}
}

[1] https://core.tcl.tk/tips/doc/trunk/tip/323.md