Allow --flag, --flag=true and --flag=false with any default.

[iago-lito]

I wish my boolean setting to be input either as:

• --flag=true or --flag : meaning true.
• --flag=false : meaning false.
• (nothing) : meaning <default value> (either true or false).

What I have come up with is:

    #[arg(
        long,
        num_args(0..=1),
        default_value_t = true, // (or false)
        default_missing_value("true"),
    )]
    flag: std::primitive::bool,

It works the way I expect, but it looks very complicated. Is there a better way?
Or should I need to ask for some new magic like:

#[arg(long, default_value_t = true, explicit_value_allowed)]
flag: bool,

or something?

[epage]

Another way of writing this is

    #[arg(
        long,
        num_args(0..=1),
        action = clap::builder::ArgAction::Set,
        default_value = "false",
        default_missing_value = "true",
    )]
    flag: bool,

At the time, there is not a better way. I would not want to provide a way of doing this without an explicit num_args so people are more likely to notice all of the cautions and decide whether they want to add require_equals = true.

As an alternative, I think if we relax the debug_assert on SetTrue from only allowing num_args(0) to anything with a cap of 1, then most of this could probably simplified down to

    #[arg(
        long,
        num_args(0..=1),
    )]
    flag: bool,

[iago-lito]

This would sound nice :) I agree with your concerns and I agree that they seem correctly addressed with just lifting the assertion. Notice that default_value_t = false would still be needed to match exactly.

As a motivation : I have found this pattern useful in a situation with numerous (4-6) similar-looking boolean flags, not all of which are false by default. The repetition of this large attribute block for every flag is awkward and not self-explanatory, but the only alternative I could think of was to duplicate all flags into --flag and --no-flag, and then checking/constraining that neither both or neither are set for each pair. I thought that that duplication was more explanatory but then even more awkward, thus my suggestion here.

Thank you for feedback (and for clap <3).

[epage]

#5912 was released in 4.5.30