Help usage and shellcheck fixes

[jasonkarns]

shellcheck fixes

fixed some things that shellcheck static analysis found

• SC2164: Use cd ... || exit in case cd fails.
• SC2035: Use ./* so names with dashes won't become options.
• SC2086: Double quote to prevent globbing and word splitting.
• SC2166: Prefer [ p ] && [ q ] as [ p -a q ] is not well defined.
• SC2155: Declare and assign separately to avoid masking return values.
• SC2035: Use ./..- so names with dashes won't become options.
• SC2035: Use ./-.. so names with dashes won't become options.

only print --usage help, not entire help output when given unknown args

remove extraneous 'exec'

Tests are failing on master and failing for identical reasons on this branch.

[tpope]

The --usage addition makes sense, and the replacement of sed is a nice touch. I'm not sure about the rest.

• exec means one less process running. Does it hurt anything?
• * to ./* doesn't matter because echo_lines doesn't accept options. Plus it muddies up the echo with ./
• I know of portability problems with -a, but I don't think they apply when running under bash.
• The extra quotes don't hurt anything, but I don't see what they help, given that the referenced variables are guaranteed not to contain spaces best I can tell.

[jasonkarns]

Most the changes (except exec), were caught with the shellcheck tool.

• I removed exec because it seemed inconsistent with other invocations. The only rationale I have against it is that it makes the exit implicit. Indeed, this PR itself is broken because I forgot to add exits when I removed exec :/ . And the exit status is carried from the exec'd command; which may or may not be desired. I'm assuming it's not desired in this case, because it's only emitting help usage. (An error while emitting help usage probably shouldn't cascade?) I'll update the PR per your direction, but I would suggest it be a bit more consistent with exec usage (using it in more places where early exit is desired).
• echo_lines doesn't accept options, but that doesn't mean that whatever it may be piped to (in the future) does not. echo_lines_without_symlinks doesn't accept options, either, but its output is filtered (unmodified) and immediately piped to sed, which does take options.
• I'm generally against -a for stylistic reasons in addition to portability concerns, but that's subjective, naturally. Up to you.
• Again, my personal preference is to excessively double quote since in the process of maintenance and future changes, what was once unnecessary can quickly become necessary (and buggy). The quotes don't hurt, yet help prevent future bugs in such a case as they become necessary.

[tpope]

• If help is invoked due to invalid arguments, the exit status should be a failure, actually. So we should drop the exec and add an exit 1 afterwards (which means expanding the || case to a proper if).
• sed doesn't read arguments from stdin so I don't understand your concern about piping here.
• I don't think stylistic preference is reason enough to change existing, functioning code in either of these cases.

[jasonkarns]

If help is invoked due to invalid arguments, the exit status should be a failure, actually. So we should drop the exec and add an exit 1 afterwards (which means expanding the || case to a proper if).

good call

sed doesn't read arguments from stdin so I don't understand your concern about piping here.

yep, brain fart

I don't think stylistic preference is reason enough to change existing, functioning code in either of these cases.

legit

Changes forthcoming

[jasonkarns]

• reverted shellcheck fixes
• unknown args emit usage help as error (for both alias and unalias)
• adds --help to print full help (not just usage) with clean exit
• switch unalias help output to unalias. Unalias help was actually printing alias help. Was this intentional? Assuming that unalias should be printing its own help/usage output, considering it has help/usage doc at top of file.

[jasonkarns]

also created and refactored to a separate fail helper function to handle printing an error message to stderr and exiting with 1. pushed in separate branch. can merge to this pr if desired or open separate pr.

https://github.com/jasonkarns/rbenv-aliases/compare/help-usage...jasonkarns:fail-helper

[tpope]

Please clean up the history. One commit would be fine or you can try to split it into a small number.

[jasonkarns]

Do you want the fail-helper refactor included or no?

[tpope]

Uh I guess leave it out for now.

[jasonkarns]

squashed to 3 commits