[Insight-developers] portable temporary file name generator function?

[kent williams]

I¹m sure that would be fine;  the problem is that with respect to provable
correctness and security there is no perfect way.  I don¹t know if GUIDs are
more or less unique than the concatenation of two 8 random integers. My
thought with this code was that if you seed the random number generator with
your process ID, you¹ll be using a number sequence that is locally unique.

The perfect temporary file name generator would

1. Work with some sort of per-user preference for a preferred temporary
directory. 
2. Work atomically, i.e. create a temporary file and open it.
3. Return the filename, not just the open file handle.
4. Be thread safe.

My code assumes that /tmp is writable, and won¹t run out of space, and is
not atomic, nor thread safe.  For the specific program I¹m using it in,
that¹s kind of OK for now.


On 7/30/09 6:38 AM, "Luis Ibanez" <luis.ibanez at kitware.com> wrote:

> Kent, Mathieu,
> 
> Could we use the UUID functionalities of GDCM
> in order to generate a unique filename ?
> 
> 
>    My 2 cents,
> 
> 
>         Luis
> 
> 
> --------------------------------------------------
> On Wed, Jul 29, 2009 at 5:48 PM, kent williams <norman-k-williams at uiowa.edu>
> wrote:
>> Yeah, you found the conundrum.  If you can't come up with a temporary name
>> and create the file in one atomic operation, you're theoretically up against
>> a race condition -- another process could create that file between choosing
>> the filename and opening it.
>> 
>> My 'solution' is robust enough for our purposes, though it's neither
>> thread-safe nor secure. It also does nothing to check for things like /tmp
>> filling up or being unwritable. And it will fail on Windows as it stands
>> unless you've made a '/tmp' directory on every drive. But I will worry about
>> that if we ever deploy on Windows.
>> 
>> namespace {
>> 
>>   vnl_random randgen;
>>   bool randomized(false);
>>   // TODO
>> std::string NewTempName()
>> {
>> #if !defined(_WIN32)
>>   if(!randomized)
>>     {
>>     randomized = true;
>>     randgen.reseed(static_cast<unsigned long>(getpid()));
>>     }
>> #endif
>>   // it needs to be an image name. None of the standard library functions I
>> know
>>   // about generate unique filenames with suffixes.
>>   for(;;)
>>     {
>>     std::ostringstream strvalue;
>>     strvalue << "/tmp/";
>>     strvalue << randgen.lrand32();
>>     strvalue << randgen.lrand32();
>>     strvalue << ".nii.gz";
>>     if(!vtksys::SystemTools::FileExists(strvalue.str().c_str(),true))
>>       {
>>       return strvalue.str().c_str();
>>       }
>>     }
>> }
>> 
>> 
>> On 7/29/09 4:26 PM, "Mathieu Malaterre" <mathieu.malaterre at gmail.com> wrote:
>> 
>>> > On Tue, Jul 28, 2009 at 8:46 PM, kent
>>> > williams<norman-k-williams at uiowa.edu> wrote:
>>>> >> I have need to generate some temporary filenames, in order to write out
>>>> >> images for external programs to use as inputs.
>>>> >>
>>>> >> I notice there's no temporary file name generator in
>>>> itksys::SystemTools.
>>>> >>
>>>> >> Shouldn't there be?
>>>> >>
>>>> >> I'd can write one, no doubt, but don't have Windows builds to test it
>>>> on.
>>>> >>
>>>> >> Is there one buried somewhere in CMake?
>>> >
>>> > The only one I found out to be working on linux and win32 was tempnam:
>>> >
>>> >  char *tempnam(const char *dir, const char *pfx);
>>> >
>>> > Most documentation will tell you this is a bad idea anyway and should
>>> > instead request a file (without knowing the actual filename) from the
>>> > system (eg: tmpfile).
>>> >
>>> >
>>> > HTH
>> 
>> _______________________________________________
>> Powered by www.kitware.com <http://www.kitware.com>
>> 
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>> 
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ
>> 
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-developers
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20090730/7ca7aded/attachment.htm>