Buitlin commands such as ls, cp do not respect --flag=false

[crabdancing]

Describe the bug

Awhile ago, --flag=true / --flag=false was added to fix the combinatorial explosion when making wrappers. This works great on user-defined scripts. Although the parser appears to also understand this syntax with respect to builtin commands, it doesn't actually seem to respect what the syntax is saying.

How to reproduce

1. ls --long=false

2. *ls is long* :c

name	type	target	readonly	mode	num_links	inode	user	group	size	created	accessed	modified
bar	file		false	rw-r--r--	1	24217229	root	root	0 B	2023-12-31 08:45:31 (Sun, Dec) -0700	2023-12-31 08:45:31 (Sun, Dec) -0700	2023-12-31 08:45:31 (Sun, Dec) -0700
foo	file		false	rw-r--r--	1	24217228	root	root	0 B	2023-12-31 08:45:31 (Sun, Dec) -0700	2023-12-31 08:45:31 (Sun, Dec) -0700	2023-12-31 08:45:31 (Sun, Dec) -0700

Expected behavior

--flag=false should disable flag for the command.

Screenshots

No response

Configuration

key	value
version	0.88.1
branch
commit_hash
build_os	linux-x86_64
build_target	x86_64-unknown-linux-gnu
rust_version	rustc 1.74.0 (79e9716c9 2023-11-13) (built from a source tarball)
cargo_version	cargo 1.74.0
build_time	1980-01-01 00:00:00 +00:00
build_rust_channel	release
allocator	mimalloc
features	dataframe, default, extra, sqlite, trash, which, zip
installed_plugins	from eml, from ics, from ini, from vcf, gstat, net, query, query json, query web, query xml, regex

Additional context

No response

[NotLebedev]

The issue is the has_flag method. It is used in commands to check if flag is set, but in reality it checks if a named argument is present (without evaluating its argument if it exists)

I think I can fix it. I will rename existing has_flag method to has_named and replace usages in parser with has_named. I will move the has_flag command to CallExt and change body to something like this:

    fn has_flag(
        &self,
        engine_state: &EngineState,
        stack: &mut Stack,
        flag_name: &str,
    ) -> Result<bool, ShellError> {
        for name in self.named_iter() {
            if flag_name == name.0.item {
                return if let Some(expr) = &name.2 {
                    // Check --flag=false
                    let result = eval_expression(engine_state, stack, expr)?;
                    match result {
                        Value::Bool { val, .. } => Ok(val),
                        _ => Err(ShellError::CantConvert {
                            to_type: "bool".into(),
                            from_type: result.get_type().to_string(),
                            span: result.span(),
                            help: Some("".into()),
                        }),
                    }
                } else {
                    Ok(true)
                };
            }
        }

        Ok(false)
    }

which will additionally evaluate expression of flag if it exists and check if it is true. Unlike get_flag it will return Result<bool, ShellError> and produce an error if result of expression evaluation is not a bool.
I tried it in ls only and it seems to work so I'll try replacing it in all commands and see if it works

[WindSoilder]

Is it resolved? @NotLebedev

[NotLebedev]

Is it resolved? @NotLebedev

Yes

[WindSoilder]

Thanks for resolving it! It highly improves user experience when using builtin commands :)