Re^4: perllocale weirdness, bug, or...?

[Krambambuli]

The problem is that I want to _reproduce_ under Perl the same order relation that the system level sort uses, as I want to process under Perl a file sorted outside Perl with sort and I rely on the strict ordering in the file to be conform with what Perl uses.

[ikegami]

The problem is that I want to _reproduce_ under Perl the same order relation that the system level sort uses

What problem? You should have checked what order the system's sort uses.

$ cat >data
[e-mail]
[e-mail]
aaa2000
aaa_2000

$ export LC_COLLATE=en_US.UTF-8

$ sort data
aaa2000
aaa_2000
[e-mail]
[e-mail]

$ perl -le'use locale; chomp(@a=<>); print for sort @a;' data
aaa2000
aaa_2000
[e-mail]
[e-mail]

$ export LC_COLLATE=C

$ sort data
aaa2000
[e-mail]
aaa_2000
[e-mail]

$ perl -le'use locale; chomp(@a=<>); print for sort @a;' data
aaa2000
[e-mail]
aaa_2000
[e-mail]

Whether the order makes sense or not, it's doing exactly what you want.

[Krambambuli]

Thank you; indeed, Perl has no guilt.

As I was trying to track down a slightly more complicated issue, I just was on the wrong track for a while.

Looks like _my_ real problem originates from something I now believe to be an en_US.UTF-8 locale/glibc bug or problem which I can reproduce at system level.

Given a simple file having two records with two TAB separated fields each, like

a_2    2
a2     1

a command like 'sort -k 1 | cut -f 1 ' on that file displays

a2
a_2

(as if the field delimiter wouldn't work and the second field determines the ordering), whereas on the same file _without_ the second field the sort order is reversed.

This looks like a bug to me, tough I'm not sure yet and don't really know so far whom I should or could report this.

Thanks!
---

[ikegami]

Maybe trailing numbers are special. Add one character to every item you want to sort, sort the items, then chop them.

Update: I guess that only works if the character is the one with the lowest sort order, and which one that is isn't known.

[choroba]

You should use sort -k1,1 to make sort ignore the rest of the line.