why can't I operate on a file and replace it

atjurhs · 01-08-2013, 11:50 AM

that worked!

thanks,

Tabby

Shadow_7 · 01-08-2013, 07:24 PM

Quote:

Originally Posted by ntubski

Hmm, maybe I'm missing something obvious, but I think it's actually quite a good method (provided you create a unique filename with mktemp(1)). This is the same method used by sed -i actually:

Take for instance a counter stored in a file, and accessed constantly by a thousand plus users. If you rm and then move, then the file may not exist at the time that some other user makes a grab for it. Not that such a scenario is at play in this case. But even if the timing never works out to fail that condition, it is still posible for the timing to work out where two or more users perform the same update because they grabbed the same old data before performing an update. Which can be an issue when the counter is used to generate unique IDs. Lame example, but sort of my point. It's not always a simple single threaded, single user world these days.

ntubski · 01-09-2013, 07:36 AM

Quote:

Originally Posted by Shadow_7

Take for instance a counter stored in a file, and accessed constantly by a thousand plus users.

Okay, in that scenario temp+mv is not sufficient, but adding a rm in the middle just makes the bug harder to uncover. A solution for that would be flock(1) around temp+mv.

jpollard · 01-09-2013, 07:47 AM

Quote:

Originally Posted by Shadow_7

Take for instance a counter stored in a file, and accessed constantly by a thousand plus users. If you rm and then move, then the file may not exist at the time that some other user makes a grab for it. Not that such a scenario is at play in this case. But even if the timing never works out to fail that condition, it is still posible for the timing to work out where two or more users perform the same update because they grabbed the same old data before performing an update. Which can be an issue when the counter is used to generate unique IDs. Lame example, but sort of my point. It's not always a simple single threaded, single user world these days.

That calls for a shared memory segment instead.

And the use of semaphores to serialize access. It is not an application for file level locking, though you can do that to. Open the file read/write, then lock for exclusive access.

And no, don't use fopen. Use open - that way you don't get the conflict/race with the buffer usage.

atjurhs · 01-09-2013, 10:43 AM

theNbomr

so this to me is really goofy, but sometimes the values in the 3rd column of my data are positive and sometimes they're negative idk??? but doesn't really matter why, and the same is true for the 4th column, again idk.

so i need to have an "if" statement in the script for each of column. so no matter how the data is given to me, i need all the values in the 3rd column to always be negative and all the values in the 4th column to always be positive.

so how do i include an if statement in there. i tried....

Code:

#! /usr/bin/perl -w
#
#  LQatjurhs.pl
#
#  Usage: LQatjuhrs.pl -i file.dat
#

use strict;

    while(<>){ 
        my @z=split; 
        print "$z[0], $z[1], $z[2], ",$z[3]*-1 if($z[3]>0 ,"$z[4]*-1 if($z[4]<0)\n";
    } 
    exit 0;

and of course that's wrong

thanks for your help, Tabby

theNbomr · 01-09-2013, 12:26 PM

It is not clear whether you want to print only the values matching your criteria, or if you want to modify all of the values so that they do meet the citeria: "i need all the values in the 3rd column to always be negative and all the values in the 4th column to always be positive."

This prints only the ones that already match.

Code:

#! /usr/bin/perl -w
#
#  LQatjurhs.pl
#
#  Usage: perl -i LQatjuhrs.pl file.dat
#

use strict;

    while(<>){ 
        my @z=split;
        if( $z[2] gt 0 && $z[3] lt 0 ){
            print "$z[0], $z[1], $z[2], ",$z[47]*-1,"$z[4]\n";
        }
    } 
    exit 0;

--- rod.

atjurhs · 01-09-2013, 12:53 PM

sorry, i need "or if you want to modify all of the values so that they do meet the citeria" then print them

theNbomr · 01-09-2013, 01:08 PM

Code:

#! /usr/bin/perl -w
#
#  LQatjurhs.pl
#
#  Usage: perl -i LQatjuhrs.pl file.dat

use strict;

    while(<>){ 
        my @z=split;
        $z[2]  = abs( $z[2] ) * -1;
        $z[47] = abs( $z[47] );
        print "$z[0], $z[1], $z[2], $z[47], $z[4]\n";
    } 
    exit 0;

Where you refer to columns, this assumes you are talking about the columns in the output. The attempt in the previous post got that wrong.

--- rod.

atjurhs · 01-09-2013, 02:40 PM

thanks!

just one other simple question, how come you represent the third column as

Code:

 $z[47]  instead of  $z[3]

is there a coding reason? i've just been using a 3

theNbomr · 01-09-2013, 06:52 PM

Your original Awk spec

Code:

awk -F " " '{print $1, $2, $3, $48*(-1), $5}'

used $48, which is the 48th field in the input. In Perl, when you split on whitespace, like Awk does by default, the fields are indexed starting at zero instead of one. That was the basis for my comment about distinguishing between the input columns and the output columns.

--- rod.

Shadow_7 · 01-10-2013, 03:27 AM

Bear in mind that for things like awk the delimiter is a space. So if your filenames contain spaces, and you're trying to grab the whole filename, that could prove difficult. With various other quirks with quotes and commas and other things depending on what thing(s) your messing with at any given moment.