I declare a `my_die() { echo "$" 1>&2; exit 1; }` on top of each file. Makes life easier by knowing why the script failed instead of having only exit code or having to turn `set -x` on and rerun.
Only if I could somehow mix `if` & `set -e`in a readable way... I wanted it to only capture errors of explicit `return 1` from bash functions, not from commands within those bash functions. But I guess I'm doing too much* of the job in bash now and it's getting messy.
#!/usr/bin/env bash
set -eEuo pipefail # pipefail bash 4.2+, which is basically everything. There's a longer version for backwards compatibility, but it's rare.
die() {
echo >&2 "ERROR: $*"
exit 1
}
# e= & exit preserves the original exit code
# trap - ... prevents multiple cleanup() calls
# To only run on error instead of always, replace both EXITs with ERR
trap 'e=$?; trap - EXIT; cleanup; exit $e' EXIT
cleanup() {
: # Delete this line and place cleanup code here.
}
# Minimal script here.
# Use semantic-named functions.
# DRY.
# Don't use many levels of indentation.
# All ordinary variables must be defined before use, so use checks like [[ -n "${X-}" ]]
# All pipeline and bare command non-zero exit codes are errors.
I believe that Bash scripts should be trivially short and simple. As soon as any complexity is introduced they should be written in another language.
I agree, the moment bash script needs "if" statement, you are using wrong language.
Is that so? Sounds like this commandment has a lot of authority, I'd better start following it.
You’re free to program in language with only one data type all you want!
Tcl is great
> We can assign the value of $? to an environment variable
exit_code is not an environment variable?
https://www.gnu.org/software/bash/manual/html_node/Shell-Par...
Oops, pretty sure I meant a regular variable. Will modify the post, thanks for pointing this out.
I declare a `my_die() { echo "$" 1>&2; exit 1; }` on top of each file. Makes life easier by knowing why the script failed instead of having only exit code or having to turn `set -x` on and rerun.
Only if I could somehow mix `if` & `set -e`in a readable way... I wanted it to only capture errors of explicit `return 1` from bash functions, not from commands within those bash functions. But I guess I'm doing too much* of the job in bash now and it's getting messy.