Options that require --long should either imply it or be silently ignored

[elyscape]

Currently, if you try to execute exa with a long view option but without specifying --long/-l, it gives an error:

scapeless:exa elyscape$ exa -@
Option --extended is useless without option --long.
scapeless:exa elyscape$ exa -l@
.rw-r--r--   22k elyscape 30 Mar 21:18 Cargo.lock
.rw-r--r--   724 elyscape 30 Mar 21:18 Cargo.toml
.rw-r--r--  1.1k elyscape 30 Mar 21:18 LICENCE
.rw-r--r--   759 elyscape 30 Mar 21:18 Makefile
.rw-r--r--@ 5.5k elyscape 30 Mar 21:18 README.md
                                       └── someattr (len 9)
.rw-r--r--  7.3k elyscape 30 Mar 21:18 Vagrantfile
drwxr-xr-x     - elyscape 30 Mar 21:18 contrib
.rw-r--r--  454k elyscape 30 Mar 21:18 screenshots.png
drwxr-xr-x     - elyscape 30 Mar 21:18 src
drwxr-xr-x     - elyscape 30 Mar 21:18 xtests

For comparison, here's what happens on macOS with ls:

scapeless:exa elyscape$ ls -@
Cargo.lock	LICENCE		README.md	contrib		src
Cargo.toml	Makefile	Vagrantfile	screenshots.png	xtests
scapeless:exa elyscape$ ls -l@
total 1000
-rw-r--r--   1 elyscape  staff   22074 Mar 30 21:18 Cargo.lock
-rw-r--r--   1 elyscape  staff     724 Mar 30 21:18 Cargo.toml
-rw-r--r--   1 elyscape  staff    1080 Mar 30 21:18 LICENCE
-rw-r--r--   1 elyscape  staff     759 Mar 30 21:18 Makefile
-rw-r--r--@  1 elyscape  staff    5511 Mar 30 21:18 README.md
	someattr	     9 
-rw-r--r--   1 elyscape  staff    7275 Mar 30 21:18 Vagrantfile
drwxr-xr-x   4 elyscape  staff     136 Mar 30 21:18 contrib
-rw-r--r--   1 elyscape  staff  454908 Mar 30 21:18 screenshots.png
drwxr-xr-x   9 elyscape  staff     306 Mar 30 21:18 src
drwxr-xr-x  29 elyscape  staff     986 Mar 30 21:18 xtests

That is to say, if -@ is specified without -l, it is silently ignored. Rather than error out, this still gives some useful input. That being said, it makes a certain amount of sense to assume that, in the event a user has specified, say, --git, the user probably wants long output and that it should be automatically turned on.

[lilyball]

I agree that it should go ahead and output anyway. I don't think it should imply --long though. Perhaps the best approach is to print the current warning about the option being useless, followed up by the normal output (as if the useless option wasn't specified).

[Mange]

Perhaps the best approach is to print the current warning about the option being useless

Then one cannot have an alias with the common options they want and only specify -l when they want long output.
Common example: alias ls='ls -h' for those that want human sized by default when giving -l.

Things like --git could also support the short format in the future (like a colored marker after filenames), which means it should probably not imply --long automatically, or else a --short option has to be added later.

[lorenzos]

+1 mainly because what @Mange said about aliases. I'm using aliases to "replace" ls with exa --git every time.

[ogham]

It's not ready for release yet, but in the option-pars-ng branch (the hardest part of development is coming up with the branch name), there's been a lot of work to support this case.

Basically, exa kept on running into limits of the default getopts crate: it didn't accept invalid UTF-8, it threw an error on repeated options, and doesn't support flags later in the list overriding ones from earlier, like how ls does it. I looked at the options, decided that the world needed one more, and wrote a custom option parser that works the "right" way. With this in place, exa will start to ignore options that require --long unless that option is passed in to.

I wrote some more stuff on it in one of the files if you're interested in the details.

[lilyball]

Nice. I would suggest the usage of https://clap.rs instead of doing something custom, except apparently clap-rs doesn't have any way to allow later arguments to override earlier ones. There's already an issue open about this (clap-rs/clap#976), so I'm going to go comment on that now.

[lorenzos]

I updated the repo, but looks the fatal warning is still there:

# git log --oneline 
97d1472 Merge branch 'strict-mode-properly'
b286676 Add actual error messages for the error messages
adaa36e Integrate strict mode, use it to test file sizes
dbebd60 Extract var_os and use the mock to test
532ebbc Only complain about long options in strict mode

# make clean
cargo clean

# make
cargo build --release --no-default-features --features "default"
   Compiling [...]
   Compiling exa v0.7.0 (file:///home/lorenzo/Dev/exa)
    Finished release [optimized] target(s) in 93.23 secs

# target/release/exa --git
Option --git is useless without option --long (-l).

[ogham]

Argh I forgot to change it for the --git flag