ArgGroup nesting? (or complex requires)

[zendril]

I would like to have either param --a OR (param --b AND --c)
Stated differently, either a needs to be present, or both b and c should be present.

This almost feels like I should have 2 groups. An arg group with b and c nested inside an outer arg group that has arg a and the bc group.

Is there a way to achieve this in clap, specifically using the derive syntax, but I'll take any options TBH.

[epage]

We currently have a bug around nested groups (#4697).

However, you don't have to have a parent group but could just make --a conflict with the group for --b / --c.

You can use any builder method as an attribute. If you want some of this done for you at the type level, we do implicitly make all structs as groups, so you could put --b and --c into a struct that derives Args and then have --a conflicts_with = "BandCStruct". We are tracking enum support for this in #2621.

[zendril]

Thanks for the quick reply.

I think I'm following what you are saying.
Do you happen to have an example of this anywhere?

And if I did mark a conflicts_with = "BandCStruct", how would I still ensure that at least a or b/c are required?

[epage]

Do you happen to have an example of this anywhere?

Not quite the same as your exact scenario (requires rather than conflicts_with): https://docs.rs/clap/latest/clap/_derive/_tutorial/chapter_3/index.html#argument-relations

And if I did mark a conflicts_with = "BandCStruct", how would I still ensure that at least a or b/c are required?

On BandCStruct, set #[group(required = true)] and wrap --as type in an Option and add a #[arg(required = true)].

[zendril]

Not quite the same as your exact scenario (requires rather than conflicts_with): https://docs.rs/clap/latest/clap/_derive/_tutorial/chapter_3/index.html#argument-relations

Yeah.. that is where I'm at currently..
I have it so that if b requires c, and c requires b and a,b,c are all Option..
But that means I could pass none of them and clap is happy.

On BandCStruct, set #[group(required = true)] and wrap --as type in an Option and add a #[arg(required = true)].

I'll play with this.. but I think this means that b&c are required all the time.. That isn't the intention.. either a OR b+c are required.
Altho I admit, I guess I need to read up on how required=true mixes with Option<T>.. not sure what behavior that is indicating.

[epage]

I'll play with this.. but I think this means that b&c are required all the time.. That isn't the intention.. either a OR b+c are required.

Conflicts override required.

Altho I admit, I guess I need to read up on how required=true mixes with Option.. not sure what behavior that is indicating.

Whether a field's type is wrapped in Option conveys two things

• Whether the field is required = true or required = false in clap
• Whether the field is required to instantiate the struct

By default, clap couples these together. If you make a field Option but set required = true manually, it allows clap to override the required = true and be able to instantiate the struct.

[zendril]

ahh.. thx