SDB:Changes in Posix.2

Version: 9.1

Symptom

You use a shell script or a command for editing classifications ([A-Z]), making calculations with numbers with a decimal separator (1.5), or sorting. However, the results produced recently by this command or script differ from previous ones.

Examples

Sorting:

user@earth:~> export LC_COLLATE=de_DE
user@earth:~> ls
aaa  AAA  bbb  BBB  ccc  CCC
user@earth:~> export LC_COLLATE=sk_SK
user@earth:~> ls
AAA  aaa  BBB  bbb  CCC  ccc

Ranges:

user@earth:~> export LC_COLLATE=de_DE
user@earth:~> echo [a-c]*
aaa  AAA  bbb  BBB  ccc
user@earth:~> export LC_COLLATE=sk_SK
user@earth:~> echo [a-c]*
aaa BBB bbb CCC ccc

Calculations with decimal separators:

user@earth:~> LC_NUMERIC=de_DE
user@earth:~> awk 'BEGIN{x=0} {x+=$1} END{print x}' <<END
> 1.5
> 2
> END
3
user@earth:~> LC_NUMERIC=en_US
user@earth:~> awk 'BEGIN{x=0} {x+=$1} END{print x}' <<END
> 1.5
> 2
> END
3.5

Cause

Many programs, including Bash, take language-specific differences into account when sorting or calculating. For example, sorting operations in German are case-insensitive, but in Slovak, uppercase always precedes lowercase letters.

Solution

If you expect certain tasks and build subsequent script components on these expectations, make sure that the script runs in a defined environment, that is, the language (locale) must be set. This can be easily done by way of the variable LC_ALL. For this purpose, insert export LC_ALL=C at the beginning of the script to ensure that the output suits your needs. <keyword>bash,LANG,locale,posix,ranges,calc,float,number,format</keyword>