awk's asort sort order according to locale

BerzinTehvs · 07-30-2010, 01:05 AM

Hi!

How could it be possible to make awk asort function to be locale-aware?

Systems locale is set to lv_LV, but asort ignores it and sorts as if LC_COLLATE is en_US. Text to be sorted is ISO8859-13 encoded.

output of locale:

LANG=lv_LV
LC_CTYPE="lv_LV"
LC_NUMERIC="lv_LV"
LC_TIME="lv_LV"
LC_COLLATE="lv_LV"
LC_MONETARY="lv_LV"
LC_MESSAGES=en_GB
LC_PAPER="lv_LV"
LC_NAME="lv_LV"
LC_ADDRESS="lv_LV"
LC_TELEPHONE="lv_LV"
LC_MEASUREMENT="lv_LV"
LC_IDENTIFICATION="lv_LV"
LC_ALL=

malfunctioning part of script:

Code:

TMPO=`echo $F4 | awk -F ", " '{RS=ORS=", "; n=split($0,arr); asort(arr); for (i = 1; i <= n; i++) print arr[i] }'`

Tinkster · 07-30-2010, 01:38 AM

How about sample data and sample output?

BerzinTehvs · 07-30-2010, 02:49 AM

I'll better attach the script and the input file.

it should be used so: sinon.sh -s tabula.txt

the last line of current output looks so:

Code:

uzvinnēt:1:(darb.v.):uzvarēt, āpmakt

but it has to look so:

Code:

uzvinnēt:1:(darb.v.):āpmakt, uzvarēt

besause "ā" is the second letter in Latvian alphabet. Sorting within regular a-z works ok.

(sorry, comments in script are in Latvian)

BerzinTehvs · 08-03-2010, 02:12 AM

As far as it seems to me after studying of the source - gawk's asort() function is not locale-aware. At least - I did not notice any code comparable to sort's from coreutils.

Tinkster · 08-03-2010, 02:01 PM

Quote:

Originally Posted by BerzinTehvs

As far as it seems to me after studying of the source - gawk's asort() function is not locale-aware. At least - I did not notice any code comparable to sort's from coreutils.

That's a real shame. Maybe you should submit a bug, or
a feature request?

Cheers,
Tink