UNIX flavours and double evaluation

When I write shell scripts (ksh) I do not like to rely on a PATH so I tend to hardcode all UNIX executables. This means that they can generally run as is using cron.

Example

CAT=/usr/bin/cat
$CAT milk

However this is not very useful when mixing UNIX variants so I have come up with this way of coding so that we can catch as many UNIX flavours as possible

# Commands
typeset -u Command
for Name in cat date grep hostname mailx mv perl ps rm
do
       Command=$Name

       case $Name in
               grep) PathNames=/usr/xpg4/bin:/usr/bin:/bin ;;
               *)    PathNames=/usr/bin:/bin ;;
       esac

       while [[ -n $PathNames ]]
       do
               PathName=${PathNames%%:*}
               PathNames=${PathNames#*:}

               if [[ $PathName = $PathNames ]]
               then
                       PathNames=""
               fi

               if [[ -x $PathName/$Name ]]
               then
                       eval "$Command=$PathName/$Name"
                       break
               fi
       done

       if [[ -z $(eval print `print \\$${Command}`) ]]
       then
               print -u2 "Unable to find a $Name executable"
               exit 1
       fi
done

The bit I like the best is that last if statement which checks that the executable was found.


The eval command tries to print out, say, $CAT to check that the CAT variable has something in it, but as CAT is the value of the variable Command we have to do some jiggery pokery to get it to parse properly.