-timeit- now on SSC: A same-line timer

[Jesse Wursten]

Dear all

-timeit- is now available on SSC, thanks as usual to Kit Baum. It is a simple wrapper program that replaces

timer on 1
reg y x [note: any command here works, this is just an example]
timer off 1

by

timeit 1: reg y x

In my opinion its main advantage is that your code can still look clean even with timers - the timer on/off lines made everything look super messy in my opinion, making it harder to debug or share code.

Currently the timer will keep running if your command returns an error. I will hopefully fix this in a future version, though suggestions on how to do this are welcome.

[Jesse Wursten]

Updated such that named timers progress independently from numbered ones. In the example below, the "regression" timer cumulates to 1500ms, despite being linked to different numbered timers in each call.

Inspired by this post.

Thanks to Kit Baum as usual for the ssc update.

. timeit 1 regression: sleep 1000
. timeit 2 regression: sleep 500
. return list scalars:
   r(delta_regression) =  .509
         r(regression) =  1.512
           r(delta_t2) =  .509
                 r(t2) =  .509
           r(delta_t1) =  1.003
                 r(t1) =  1.003

[Jesse Wursten]

daniel klein Yes!!!!

Sorry, I have been looking for a way to respect the host version in a robust way for a long time now and this does that brilliantly.

I've adapted the code, it's en route to SSC now (and attached)

[daniel klein]

Ali correctly diagnoses the problem; I suggest a different solution, though. Do not tinker with timeit.ado. Instead, change your programs to properly include the desired version, i.e., write

program define my_program
            version 12 // <- add the version
            sysuse "auto.dta"
            ren weigh* weigh
end

Alternatively, type

timeit 1 : version 12 : my_program

Jesse might want to add caller-version support (something that I tend to forget, too). The relevant changes might be as simple as

[...]
nobreak {
    timer on `theTimer'
    capture noisily break version `=_caller()' : `theCommand'
    timer off `theTimer'
}
[...]

[Justin Niakamal]

Does the trick. Thanks, Ali.

[Ali Atia]

I believe the problem is the version statement. The error doesn't appear when the statement is changed to version 12, which is probably because additional functionality was added to rename group at that time. I'm not sure if changing the version breaks anything else, but if you want a temporary workaround that does the trick.

[Justin Niakamal]

Yes, that's covered in the first post of the thread. My question was in the context of timeit.

[Leonardo Guizzetti]

The usual way would be marginally more typing.

timer on 1
...your code...
timer off 1
timer list 1

[Justin Niakamal]

I’m interested in using timeit before running a do file, but I’m getting an error and I think the wildcard asterisk is causing the issue. Here are two simplified examples

clear all
 
program define my_program
            sysuse "auto.dta"
            ren weigh* weigh
end
 
program define my_program_alt
            sysuse "auto.dta"
            rename *, lower
end
 
 
 
timeit 1: my_program
timeit 2: my_program_alt

. timeit 1: my_program
(1978 Automobile Data)
* invalid name
r(198);
 
. timeit 2: my_program_alt
(1978 Automobile Data)
* invalid name
r(198);

Is there a workaround?

[Jesse Wursten]

Nice catch. This should be fixed in the version on its way to SSC now. I've also attached it here.

[Joro Kolev]

-timeit- returns an error when there is empty space before the semicolon.

E.g.,

. clear

. timer clear

. set obs 10
number of observations (_N) was 0, now 10

. timeit 1 : gen u = runiform()
= invalid name
r(198);

.

This is not an end-of-the-world problem, but it is a bit inconsistent with usual Stata behaviour where additional empty spaces typically do not cause issues. I do use empty spaces a lot to improve readability of code.

[Jesse Wursten]

Hmm, you seem to be correct. I should be able to fix this.

EDIT: I think I fixed it. Updated version is en route to SSC, also attached below.

[Santiago Cantillo]

Your right, I tried that and it worked. But if the name of the do file has spaces it won't work.

 
 timeit 1: do "sleep 2000.do"

produces type mismatch "r(109)".
I also received this error message when trying to merge a file that had spaces in the name.
Perhaps the issue occurs when doing any operation involving file names with spaces?

[Jesse Wursten]

Hi Jesse Wursten . Thanks very much for your program.
It just occured to me that it would also be really helpful if one could use timeit before running external do files
(e.g. timeit 1 dofile1 : do foo.do).

timeit 1: do sleep2000.do

That worked for me?

[Santiago Cantillo]

Hi Jesse Wursten . Thanks very much for your program.
It just occured to me that it would also be really helpful if one could use timeit before running external do files
(e.g. timeit 1 dofile1 : do foo.do).