Variations across unices

[guy at sun.UUCP]

> > eval tr '"[A-Z][a-z]" "[N-Z][A-M][n-z][a-m]"' ${1:+'<$1'}
>                                                    ^
> What good is this colon?  We were speculating that it might serve to
> terminate a numeric expression a la ${foo+bar-1:+defalt} but were not
> able to get this shell script to work until the colon be removed.

See previous postings concerning alleged "incompatibility" between 4.2BSD
and Sun UNIX 3.0.  "${1+'<$1'}" means if "$1" (first parameter to the script)
is set (i.e., a parameter was supplied to the script), substitute "<$1",
otherwise substitute nothing.  Thus, if an argument is supplied, "tr" has
its input redirected to the argument supplied, otherwise it reads from the
standard input.  "${1:+'<$1'}" means if "$1" is set and is non-null,
substitute "<$1", otherwise substitute nothing.  The ":" is a feature of the
S3 and later shells, and isn't in the 4.2BSD shell, which is why it didn't
work until you removed it.  It's not really necessary.

> We also had to remove the [square brackets] for it seems that "tr" on this
> system not only doesn't need them but treats them as normal characters!
> ...
> Are there in fact variant unices that require the [square brackets] 
> on "tr" commands and allow integer expressions in ${parm+defalt} lists?

Yes to the first (for some reason, the S5 and, I think, S3 versions of "tr"
require the square brackets) and no to the second (the "1" refers to the
first argument, and is not a numeric value).

> Or is this just somebody's idea of a bad joke?

Unless there was a good reason to incompatibly change the syntax of "tr"
from what AT&T's own V7 supported, one could argue that the USDL perpetrated
a bad joke on us all by doing so....
-- 
	Guy Harris
	{ihnp4, decvax, seismo, decwrl, ...}!sun!guy
	guy at sun.com (or guy at sun.arpa)