Ripgrep: "was provided more than once" when combining RIPGREP_CONFIG_PATH with command-lines using condensed short option syntax

[ssokolow]

What version of ripgrep are you using?

ripgrep 12.1.1
+SIMD -AVX (compiled)
-SIMD -AVX (runtime)

How did you install ripgrep?

RUSTFLAGS="-C target-cpu=native" cargo +nightly install ripgrep --features 'simd-accel'

What operating system are you using ripgrep on?

$ lsb_release -a                                                                             
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 16.04.7 LTS
Release:        16.04
Codename:       xenial

Describe your bug.

ripgrep exits with the following error...

error: The argument '--pretty' was provided more than once, but cannot be used multiple times

...if -p or --pretty is in the configuration file and condensed short option syntax is used on the command line.

What are the steps to reproduce the behavior?

1. Set up RIPGREP_CONFIG_PATH and a configuration file containing -p or --pretty
2. Run ripgrep with -p -z or --pretty -z at the command line.
3. Notice that it works as expected, as it should for use in scripts.
4. Run ripgrep with -p -z -p at the command line
5. Notice that it works in an alias-compatible way as it should.
6. Run ripgrep with -pz at the command line.
7. Get an error message.

What is the actual behavior?

$ rg -pz foo /usr/share/doc 
error: The argument '--pretty' was provided more than once, but cannot be used multiple times

USAGE:

    rg [OPTIONS] PATTERN [PATH ...]
    rg [OPTIONS] [-e PATTERN ...] [-f PATTERNFILE ...] [PATH ...]
    rg [OPTIONS] --files [PATH ...]
    rg [OPTIONS] --type-list
    command | rg [OPTIONS] PATTERN

For more information try --help

What is the expected behavior?

It shouldn't matter what form it takes.

What do you think ripgrep should have done?

-pz should behave identically to -p -z.

[okdana]

I think the config file is a red herring. I get the same issue with just the command line. The option-stacking thing does seem relevant though

These fail:

% RIPGREP_CONFIG_PATH= \rg -p -pz x <<< xyz
% RIPGREP_CONFIG_PATH= \rg -ppz x <<< xyz

But these are OK:

% RIPGREP_CONFIG_PATH= \rg -pz -p x <<< xyz
% RIPGREP_CONFIG_PATH= \rg -pzp x <<< xyz
% RIPGREP_CONFIG_PATH= \rg -zpp x <<< xyz
% RIPGREP_CONFIG_PATH= \rg -pp -z x <<< xyz
% RIPGREP_CONFIG_PATH= \rg -p -p -z x <<< xyz

It does it with all other options i tried, too, not just --pretty. I assume it's a Clap bug...?

[okdana]

Yeah, i think so, since fd also does it:

% fd -1 -pp -L README
README.md

% fd -1 -ppL README
error: The argument '--full-path' was provided more than once, but cannot be used multiple times ...

I couldn't find a related issue in the clap repo, i guess it needs reported there

[ssokolow]

OK. I'll try to find time to put together a minimal reproducer tomorrow.

EDIT: The day didn't go well. Tomorrow maybe.

[ssokolow]

I've tried to put together a reproducer, but I can't remember what the proper way is to ask clap for that alias-compatible behaviour where you can specify an argument multiple times but the additional copies have no semantic significance within the program.

[BurntSushi]

@ssokolow Thanks for looking into this! I believe the setting you're looking for is clap::AppSettings::AllArgsOverrideSelf. ripgrep sets it here.

[ssokolow]

Thanks. I've now got a reproducer:

use clap::{Arg, App, AppSettings};

fn main() {
    let schema = App::new("ripgrep#1701 reproducer")
        .setting(AppSettings::AllArgsOverrideSelf)
        .arg(Arg::with_name("pretty")
            .short("p")
            .long("pretty"))
        .arg(Arg::with_name("search_zip")
            .short("z")
            .long("search-zip"));

    let test_args = &[
       vec!["reproducer", "-pz", "-p"],
       vec!["reproducer", "-pzp"],
       vec!["reproducer", "-zpp"],
       vec!["reproducer", "-pp", "-z"],
       vec!["reproducer", "-p", "-p", "-z"],

       vec!["reproducer", "-p", "-pz"],
       vec!["reproducer", "-ppz"],
    ];


    for argv in test_args {
        let matches = schema.clone().get_matches_from_safe(argv);
        match matches {
            Ok(_) => println!(" OK: {:?}", argv),
            Err(e) => println!("ERR: {:?} ({:?})", argv, e.kind),
        }
    }
}

 OK: ["reproducer", "-pz", "-p"]
 OK: ["reproducer", "-pzp"]
 OK: ["reproducer", "-zpp"]
 OK: ["reproducer", "-pp", "-z"]
 OK: ["reproducer", "-p", "-p", "-z"]
ERR: ["reproducer", "-p", "-pz"] (UnexpectedMultipleUsage)
ERR: ["reproducer", "-ppz"] (UnexpectedMultipleUsage)

I'm not feeling motivated to open a bug right now but I'll do it later if someone else doesn't beat me to it. (It's near the end of my day and anything which introduces a new social component feels tiring.)

[ssokolow]

Reported as clap-rs/clap#2171