How to print error and exit if command fails?

Shell Programming and Scripting
1

Guys any tips on printing a certain error message to stderr and exiting should a command fail within a ksh script? I'm trying to null some output files.

Touch isn't suitable as i need to null them.
print "" > file isn't suitable as i need to check elsehere for if they are 0bytes or not.

I've tried these below examples and none work correctly. I don't want to have to put a check after each command as :-

if [[ $? -ne 0 ]];then
   print "error blah blah" >&2
   exit 2
fi

Below tests don't work correctly. I'm guessing its spawning something.

OUTFILE=/tmp/out

# Null outfiles. Security already checked

> ${OUTFILE} || print "ERROR: blah blah \n" >&2 ; exit 2   #doesnt work

> ${OUTFILE} || (print "ERROR: blah blah \n" >&2 ; exit 2 )  #doesn't work

> ${OUTFILE} || (print "ERROR: blah blah \n" >&2 && exit 2) #doesnt work

if [[ -n "$(> ${OUTFILE} 2>&1)" ]];then
     print "ERROR: blah blah \n" >&2
     exit 2
fi    # doesnt work

Any ideas or alternatives?

2

Specifically about reporting and exiting, this is a slightly more condensed idiom:

test $condition || { print "Crap blew up; exiting"; exit 2; }

You could also directly test for the result of your command:

command || { print "Crap blew up; exiting"; exit 2; }

I enjoy reading that out loud to myself as "either you do this or DIE!", as if threathening the script.

3

Thanks dude thats worked like a charm.

Seems i was almost there but didn't use the correct { }

Could you explain the difference between { } and ( ) in the command grouping? Also i notied the ; at the end before } is vitally important otherwise the next command doesn't work.

e.g

This works and exits if cant null but echos got to here if can.

> ${OUTFILE} || { print "ERROR: cannot null output file. Exiting\n" >&2; exit 2; }
> ${TMPFILE} || { print "ERROR: cannot null tmp file. Exiting\n" >&2; exit 2; }

echo "got to here"

This doesnt work and never gets to echo even if null is successful

> ${OUTFILE} || { print "ERROR: cannot null output file. Exiting\n" >&2; exit 2 }
> ${TMPFILE} || { print "ERROR: cannot null tmp file. Exiting\n" >&2; exit 2 }

echo "got to here"
4

() executes the statements in a new subshell, with separate state. {} executes stuff in the current shell.

Example:

(aVariable="hello"); echo $aVariable

this doesn't print "hello" because the variable was assigned inside a new shell, whose state was discarded when the () expression ended. Think of variable scoping in C; variables live and die inside the block where they're declared.

{aVariable="hello"; }; echo $aVariable

this prints "hello" because the variable was assigned in the same shell as the next statement.

It's the same difference as:

sh myScript.sh

and

source myScript.sh

Also, yes, you need a semicolon to end the last statement inside {}; that's how the shell's grammar is defined.
Kind of how you can either do this:

if $condition; then $statements; fi

or this, using newlines instead of semicolons to separate the syntax parts:

if $condition
then
    $statements
fi

According to the bash man page, it's different from () because { and } are reserved words instead of metacharacters, which means they don't automatically cause word breaks. Presumably the same applies in the rest of the shells.

5

Dude thats real great and an informative reply. Thanks a lot. Its certainly giving me a better understanding of the shell rather than just trying things until they work :slight_smile:

6

Don't mention it; I love yapping about bash.
Its man page is long and daunting, but flick through it every once in a while; you're bound to learn great tidbits every time.
The parts about history and readline are particularly cool, and they're handy and fun to use.