As an aside to this post, I learned the hard way a few months ago why -
save `tempfile', replace - should be avoided in most circumstances. If the
macro tempfile is undefined, say because you have misspelled it (for
example, you set - tempfile steve - but then used - save `steven', replace),
this reduces to - save, replace - which will overwrite the dataset you are
working with.
BTW, has statacorp ever thought about altering save so it requires a dataset
name? In the circumstance above it is very dangerous that it does not
require a dataset name. For example, in my case, I overwrote a raw dataset
(which fortunately I had another copy of on a CD).
Cheers,
Steve
-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of William Gould,
StataCorp LP
Sent: Thursday, 30 August 2007 5:10 a.m.
To: [email protected]
Subject: Re: st: tempfile already exists
Sergiy Radyakin <[email protected]> asks a series of questions
concerning Stata's tempfiles.
> Can anyone confirm that Stata does or does nothing to check if the
> 1) temp candidate file already exists
> 2) temp candidate file is open by Stata or another application (Stata
> or not Stata)
> 3) temp folder is writable
> 4) another Stata session is attempting to allocate tempfiles too
> (syncronization to avoid overlaps)
1. whether temp candidate file already exists
----------------------------------------------
Under Windows, a tempfile name looks like
<path>\ST_<instance><seqno>.tmp
where <instance> := 2-digit, base-34 instance number
<seqno> := 6-digit, base-34 sequence number
Sergiy gave the example,
<seqno>
__|___
c:\Documents and Settings\mylogin\Local Settings\Temp\ST_010000jf.tmp
----------------------------------------------------- --
| |
<path> <instance>
The instance number has to do with which Stata is running if multiple
Stata's are running and will be unique for each simultaneously running
Stata.
The sequence number will 000001 to begin, followed by 000002, etc.,
counting in base 34 for which the top digit is "x". You count 1, 2, ...,
9, a, b, ..., w, x, 10, 11, ...
So, to answer the question, "Does Stata check whether a tempfile already
exists?", the answer is No, not at the time the -tempfile- command is given.
At startup time, Stata sets <instance>. When Stata comes up, it creates a
special temporary file (not to be confused with tempfile) named
SQ_<instance>000000.tmp. When Stata shutdown cleanly, it erases that file.
Thus, when Stata comes up, it looks for SQ_<instance>000000.tmp files, find
the first one that does not exist, and that determines this Stata's
<instance>. Stata immediately creates SQ_<instance>000000.tmp.
By this logic, files of the form ST_<instance><seqno>.tmp should not
exist and therefore it is not necessary to check whether they do.
2. Whether temp candidate file is open by Stata or another application
-----------------------------------------------------------------------
Stata does not check for this. See (1) above.
3. Whether temp folder is writable
-----------------------------------
This is checked when Stata comes up. It is NOT checked everytime
a tempfile is generated.
4. Another Stata session is attempting to allocate tempfiles too
-----------------------------------------------------------------
See (1) above. It is not necessary to check because any other
running Stata will have a different <instance>.
Sergiy goes on to ask,
> [...] what might be an explanation for the following:
>
> tempfile tfile
> save `"`tfile'"'
> file c:\Documents and Settings\mylogin\Local
> Settings\Temp\ST_010000jf.tmp already exists
> r(602);
>
> The situation occurs not always, but with a non-zero probability.
> (Stata 9.2 up-to-date, Windows XP). The temp folder indeed contains
> dozens of abandoned temporary files, but please do not just suggest to
> clean it up regularly.
There are three possiblities:
1. You are using -tempfile- interactively (which is to say, outside of
do- and ado-files). In this case, it is your responsibility
to erase tempfiles you create.
2. You are passing a tempfile name out of a program for use by
a calling program. (That is inappropriate use of -tempfile-.
Each program, do-file, and ado-file knows only to erase
tempfiles that it itself obtained from -tempfile-).
3. Stata previously crashed. In this case, you should clean up.
Concerning (1), due to Sergiy's posting we are now considering changing
Stata's logic to erase ALL temporary files on exit. Is anyone taking
advantage of (1) as it currently works and passing temporary files from
one Stata session to the next? If so, we need to hear about it.
Concerning (3), cleaning up is easy. With no Stata's running, erase all
ST_*.tmp files in the temporary directory, which in Sergiy's case is
c:\Documents and Settings\mylogin\Local Settings\Temp\
We recommend Sergiy do this. If Sergiy has found some way to crash
Stata repeatedly, he should contact Technical Services because there
are other files in the temporary directory that Sergiy may need to
erase. For most users, the need to erase temporary files never arises.
Finally, Sergiy says,
> So far I have to enforce ",replace" everywhere in the program. This
> doesn't seem to be a good idea, [...]
We agree; do not use -, replace-. Clean up the temporary directory.
-- Bill
[email protected]
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
