UNIX for Advanced & Expert Users

Hi,

I'm trying to sort a text file "test":

Code:

S12
S_S12
S_S1_12
S15
S_N15
S_N1_15

By "sort test", I get:

Code:

S12
S15
S_N1_15
S_N15
S_S1_12
S_S12

It seems weird:
Comparing Line 2 and Line 3, it must be that '-' is bigger than '1'; however, comparing Line 3 and Line 4, it seems that '5' is bigger than '-'.

And the result of comparing one line to a previous line using strcmp function is:

Code:

S12
S15                        3
S_N1_15                46
S_N15                    -42
S_S1_12                 5
S_S12                     -45.

It seems that "S_N15" and "S_S12" are not in the correct position.

The result (according to strcmp) should be:

Code:

S12
S15
S_N15
S_N1_15
S_S12
S_S1_12

First of all you have underscores in your strings and not dashes.
By default 'sort' sort alphabetically based on your local.

The octal ASCII value of the underscore is 137
The octal ASCII value of the 1 is 061
And upper case letters start at 101

This is what I get based on your sample after sorting with not switches:

Code:

S12
S15
S_N15
S_N1_15
S_S12
S_S1_12

which looks correct based on the previous ASCII value explanation.
I'm in the US. What local are you at?

Or better yet, how would you like to sort?

The kernel I'm using is:
Linux ccm18 2.6.32-31-generic #61-Ubuntu SMP Fri Apr 8 18:25:51 UTC 2011 x86_64 GNU/Linux

Can't figure out why the "sort" behaves differently from yours.

Most likely your locale isnt "C" so use the locale command to find out what it is set to...

Code:

locale

Output of above command lists a bunch of LC_* variables and if LC_COLLATE isnt set to "C" then do so or just set LC_ALL to "C" which overrides all settings and sets all LC_* variables to "C"...

Code:

export LC_COLLATE="C"
or
export LC_ALL="C"

Now if you rerun the sort you will get the desired result...