unique file ID

baxy77bax has asked for the wisdom of the Perl Monks concerning the following question:

I was wondering if anyone has a better solution...

Problem: Let say I have this program that writes some temporary information on a disk(in a local directory) which is latter removed. And let say i wish to run that program twice in the same directory (because it is more practical then creating a new directory and then rerunning the program). I am under unix env. In such case if I run the program twice at the same time my temporary files will be overwritten and messed up which leads to bad computation, inconsistent results, scandal (for which i am to blame since i didn't make the program "idiot-proof", though i strictly say in the instructions not to do this, but ok, it is my fault) and finally to threats considering my job.

Solution: One solution would be to generate a unique file names by using random characters. However, this does not guaranty that there wouldn't be any collisions in the future, which means that one has to first check if such file already exists or not and the decide if the generated name is ok or not. Second solution would be to have a file where all filenames, connected to that program, are stored. That way program can first look up in that file see the last file name, let say it is a number and by incrementing the number, generate a new file name , save it into that file , use the name and when done delete that name from the file. But to make this "idiot-proof" I wouldn't use this system since there can be collisions when multiple executions do not have the same execution time and are not started at the same time (probably i would need to check if such file already exists or something)

So my question is does anyone have a better solution to this problem?

Cheers

baxy

Comment on unique file ID

Replies are listed 'Best First'.
Re: unique file ID by marto (Cardinal) on Apr 30, 2013 at 11:04 UTC
Another solution would be to use something like File::Temp.	[reply]
Re^2: unique file ID by davido (Cardinal) on Apr 30, 2013 at 19:18 UTC
This is an excellent solution. File::Temp detects and avoids collisions. It has a limited number of re-tries, so it also helps to use a template with a long enough wildcard segment that a collision is as likely as the earth's destruction by asteroid this month. File::Temp practices RAII safety by putting the clean-up code in its destructor so that even if the program terminates through an exception the destructor will remove the temp file. Parallel processing can wreck havoc on safe resource acquisition, but File::Temp has you covered (though you should read the POD section on forking). Security, community refinement, active maintenance, and extensive testing are all attributes that contribute to making this a good choice. Dave	[reply]
Re: unique file ID by Corion (Patriarch) on Apr 30, 2013 at 11:01 UTC
As long as your filesystem is mounted on a single machine, the easiest approach is to use the process ID (PID) in the filename. The PID is guaranteed to be unique per process, per machine.	[reply]
Re^2: unique file ID by choroba (Cardinal) on Apr 30, 2013 at 11:03 UTC
I usually use username + date/time + machine + PID on multiuser shared file systems. لսႽ† ᥲᥒ⚪⟊Ⴙᘓᖇ Ꮅᘓᖇ⎱ Ⴙᥲ𝇋ƙᘓᖇ	[reply]
Re: unique file ID by karlgoethebier (Abbot) on Apr 30, 2013 at 12:02 UTC
Perhaps using GUIDs? Linux etc.: `myMac:~ karl: uuidgen EBB88FE7-1262-43BD-9017-7CAEEA2E8627` [download] Windows, using Powershell: `PS C:\Documents and settings\karl> [System.Guid]::NewGuid().ToString() 995187c6-993a-41e3-b249-e2c013d71c73` [download] Should be unique enough. There are also some Modules for this on but i didn't use them. Update: I knew i saw something interesting about this some time ago: Generate GUID from a string by GrandFather. Update 2: Btw...i posted already some other variations on this "unique" stuff... #!/usr/bin/perl + + use strict; use warnings; my ( @l, @u, @n, @p, $i, $j ); open( URANDOM, "</dev/urandom" ) \|\| die $!; read( URANDOM, $_, 4 ); close URANDOM; srand( unpack( "L", $_ ) ); @l = ( "a" .. "z" ); @u = ( "A" .. "Z" ); @n = ( 0 .. 9 ); push( @p, splice( @l, int( rand(@l) ), 1 ) ) for ( 1 .. 10 ); push( @p, splice( @u, int( rand(@u) ), 1 ) ) for ( 1 .. 10 ); push( @p, splice( @n, int( rand(@n) ), 1 ) ) for ( 1 .. 2 ); for ( $i = @p ; --$i ; ) { $j = int( rand( $i + 1 ) ); next if $i == $j; @p[ $i, $j ] = @p[ $j, $i ]; } print join( "", @p ) . qq(\n); __END__ Karls-Mac-mini:monks karl$ ./random_filename.pl gRjyL1ivolIPHsJx0MCSeQ [download] ...and i still think that this is not a bad ("easy") approach - if you don't plan to deploy world wide ;-) Regards, Karl «The Crux of the Biscuit is the Apostrophe»	[reply] [d/l] [select]
Re^2: unique file ID by tobyink (Canon) on Apr 30, 2013 at 12:56 UTC
I've used both Data::UUID and UUID::Tiny in the past and had good results with both. Although I've not tried it yet, Data::GUID::Any also looks good. `package Cow { use Moo; has name => (is => 'lazy', default => sub { 'Mooington' }) } say Cow->new->name`	[reply]
Re^3: unique file ID by karlgoethebier (Abbot) on Apr 30, 2013 at 18:31 UTC
Thanks tobyink, next time i need something i will use Data::GUID::Any. Looks really good... Best regards, Karl «The Crux of the Biscuit is the Apostrophe»	[reply]
Re: unique file ID by MidLifeXis (Monsignor) on Apr 30, 2013 at 13:17 UTC
File::Temp, specifically tempdir? Update: Missed Re: unique file ID in my original thread scan, where marto recommended the same solution. --MidLifeXis	[reply]
Re: unique file ID by ruzam (Curate) on Apr 30, 2013 at 14:39 UTC
`my $file_name = time() . "_$$";` [download] Time will be unique to the second and the $$ (the PID) will be unique to each executing script.	[reply] [d/l]


Perl: the Markov chain saw
	PerlMonks