Validate argument with regex

[catleeball]

Describe your use case

It would be nice to have a clap feature for argument validation via a regex. This would be handy and flexible for users, allowing them custom validations without writing a custom validator to do so, and hopefully without requiring too large of overhead for implementation and testing in clap itself.

Describe the solution you'd like

For example, using clap yaml format, if we want an argument whose value wants only digit values 1 through 100:

args:
  - jpg-quality:
    about: Quality of jpg image to write. Must be a number 1-100.
    takes_value: true
    regex: "^[1-9][0-9]?$|^100$"

This could expand to complex regex use cases, like matching if an arg value is a URL:

args:
  - url:
    about: URL to load.
    index: 1
    required: true
    # Regex stolen from: https://stackoverflow.com/a/3809435
    regex: "[-a-zA-Z0-9@:%._\+~#=]{1,256}\.[a-zA-Z0-9()]{1,6}\b([-a-zA-Z0-9()@:%_\+.~#?&//=]*)"

Thanks for this very nice arg parsing library!

[CreepySkeleton]

My first thought was: "This is pretty trivial with normal validators!". And indeed:

lazy_static! {
    static ref RE: Regex = Regex::new("\\d+").unwrap();
}

app.validator(|val| {
    if RE.is_match(val)) {
        Ok(())
    } else {
        Err(format!("value '{}' doesn't match the `\\d+` regex", val))
    }
})

But on the other side, validators cannot be deserialized from YAML, and "matches a regex" does sound common enough to warrant a special case.

I say we might gate it behind regex feature.

Unresolved questions: what about error messages? I have doubts that "value doesn't match regex" is good user-facing message. Perhaps the API should look like:

fn regex_validator("-?\\d+", "expected an integer");

// rendered as
"Invalid value `foo`: expected an integer"

@pksunkara @kbknapp Any thoughts?

[catleeball]

Agreed on your point about needing a more helpful error, @CreepySkeleton . I personally like the example function you provided where the user supplies an error string.

[loewenheim]

How would the error message be written in a YAML file?

[CreepySkeleton]

Just as any other tuple would

regex_validator: 
  - r"[-a-zA-Z0-9@:%._\+~#=]{1,256}\.[a-zA-Z0-9()]{1,6}\b([-a-zA-Z0-9()@:%_\+.~#?&//=]*)"
  - error message

# or if regex and error are short enough
regex_validator: ["-?\\d+", "expected an integer"]

[tmankita]

Hi,
I would like to pick up this issue if no one is working on it yet.

[epage]

FYI I am considering removing validator_regex, see #3743