Re: st: RE : st: RE : st: Do-file no result, command line OK

[Nick Cox <njcoxstata@gmail.com>]

I see.

In that case my only guess -- on this information -- is that something
else is happening in the rest of the do-file that you are not showing
us.

No doubt it is likely to be long and complicated, but we can't easily
make progress on this without an indication of the smallest dataset
and .do file that appear to produce this problem. Even better would be
a reproduction of your problem on data we all can look at, such as
some elaboration of

sysuse auto
gen make_2 = subinstr(make, ".", " ", .)
l make if make != make_2

Nick

On Tue, Mar 29, 2011 at 1:53 PM,  <X.Y.Vollenweider@lse.ac.uk> wrote:
> Thanks for your answer. To be more precise, my concern is that in the best case not all the changes have been made and in the worst case no change at all has been made. Indeed, I found many observations just by skimming through the data with some dot(".") although they should all have been uniformly removed by the command line:
>
> replace companyname_short=subinstr(companyname,"."," ",.)
>
> best,
>
> Xavier
>
>
> -------- Message d'origine--------
> De: owner-statalist@hsphsun2.harvard.edu de la part de Nick Cox
> Date: mar. 29/03/2011 13:39
> À: statalist@hsphsun2.harvard.edu
> Objet : Re: st: RE : st: Do-file no result, command line OK
>
> I don't know how looking in the Data Editor can convince you that
> nothing has changed among 5 million observations. Did you check all of
> them?
>
> The only convincing evidence that nothing has changed is that you
> clone the variable, issue the -replace- command on the clone, and then
> compare with the original.
>
> Nick
>
> On Tue, Mar 29, 2011 at 1:31 PM,  <X.Y.Vollenweider@lse.ac.uk> wrote:
>
>> Thanks Uli, but no -preserve- nor -restore-. Quite weird indeed.
>
> Ulrich Kohler
>
>> This sounds pretty weird.
>>
>> I sometimes experience something like this in Do-Files with a
>> -preserve-/-restore- part in it. When running those parts of the Do-File
>> that contain the -preserve- command but does not contain the -restore-
>> command, the end of the Do-file brings you back to the preserved data
>> set.
>
>> X.Y.Vollenweider@lse.ac.uk:
>
>>> I am running a do-file on a large dataset (4'999'453 observations). One of my command is:
>>>
>>> replace companyname_short=subinstr(companyname,"."," ",.)
>>>
>>> That is, I replace the dot in the variable "companyname" by a blank in the variable "companyname_short". When I run the do-file, stata tells me that x hundred thousands of "real changes" have been made, but actually nothing have changed in my dataset when I open the data viewer. However, when I copy and past the command from the do-file into the command line, everything works fine.
>>>
>>> Any idea to make it run smoothly with the do-file?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/