tr A-Z a-z

[Dan Nelson]

In the last episode (May 07), Christoph P. Kukulies said:
> On Fri, May 07, 2004 at 01:59:01AM -0700, Kris Kennaway wrote:
> > On Fri, May 07, 2004 at 10:53:01AM +0200, Christoph Kukulies wrote:
> > > Strange: I was used to do upper case lower case conversion always like this
> > > and it suddenly doesn't work anymore:
> > > 
> > > $ echo Z | tr "[A-Z]" "[a-z]"
> > > ÿ
> > 
> > Something locale-related?
> 
> locale
> LANG=en_US.ISO_8859-1
> LC_COLLATE="en_US.ISO_8859-1"
>
> echo Z | tr "A-Z" "a-z" | od -x
> 0000000      0aff
> 0000002

>From the tr manpage:

     c-c        For non-octal range endpoints represents the range of
                characters between the range endpoints, inclusive, in
                ascending order, as defined by the collation sequence.

Note that 8859-1 has uppercase and lowercase accented characters, which
collate alongside the unaccented characters.
/usr/src/share/colldef/la_LN.ISO8859-1.src holds the collation sequence
for en_US.ISO_8859-1.  There are two lowercase y, but three uppercase
Y's.  This means that your ranges are different sizes, and Z maps to
<y:>, which happens to be 0xff in the 8859-1 charset.

> It must be something too obvious but I don't see it at the moment.
> 
> I found that it depends on my special environment settings.
> A different user doesn't have this problem.

-- 
	Dan Nelson
	dnelson at allantgroup.com