Shell Programming and Scripting

It might be just me, but the fact that the "print"- and "read"-builtins are I/O-stream-aware (in the sense that they have the "-u"-option to direct their output) feels cleaner in my eyes.

You named it: if there is a command option one can use variables contents to feed arguments to it. It would be complicated to build the following ksh-code in bsh/bash:

Code:

exec 3>file.1
exec 4>file.2
exec 5>file.3

typeset -L1 chDescriptor="4"     # replace with "3" or "5" to switch streams

print -u${chDescriptor} - "this goes to a I/O-stream selected by chDescriptor"

exec 3>&-
exec 4>&-
exec 5>&-

bakunin

I guess I don't see why:

Code:

#!/bin/ksh
exec 3>file.1
exec 4>file.2
exec 5>file.3

chDescriptor="4"     # replace with "3" or "5" to switch streams

echo "this goes to a I/O-stream selected by chDescriptor" >&"$chDescriptor"

exec 3>&-
exec 4>&-
exec 5>&-

is so much more complicated and works just fine with with ksh, bash, or any other shell that performs standard I/O redirections.

I wasn't aware that parameter substitution at this place was possible. Because i use Korn shell almost exclusively i never had to resort to this. My impression was that command line parsing would have made some "eval" necessary for this to work. In this case my point is moot.

bakunin

I found a thousand ways to do that which didn't work Never stumbled across that exact way. Is it a "special" expansion like "$@" which needs to be used that exact way?

fd1>&fd2 (no spaces allowed) where fd1 and fd2 are numeric strings specifying file descriptor numbers is the standard way of redirecting the output found on file descriptor fd1 to the file connected to file descriptor fd2. If fd1 is not given, it defaults to 1 (i.e. stdout).

You frequently see things like:

Code:

printf "Diagnostic message\n" >&2

to write diagnostics to stderr instead of stdout in shell scripts. Using a variable to specify a file descriptor instead of a constant is seen much less often, but the shell is perfectly happy in either case.