This seems more like a proposal. I'm adding the text/template owners in order to begin the discussion.

/cc @robpike @mvdan

@cagedmantis @robphoenix @mvdan please any update on this task? It would be super helpful if we can use this style of property resolver.

@bilak @cagedmantis @robphoenix @mvdan do you have some opinions on this?

There are dozens of open text/template issues. Please don't ping proposals every few weeks; we have limited time.

Personally, I think we should implement the already-accepted proposals (#20531 and #31103) before we consider new ones. Any help with those is welcome.

It was 5 months ago this was written, and I find it common courtesy to give a short comment that it has been seen. But each repo to his own.

The way to make progress on this is to turn it into a proposal with a precise definition of the suggested change and the new semantics.

For example, if I pass

var s struct {
    p *struct {
        f int
    }
}

to a template and write s.p.f then assuming that s.p is not nil I will get an integer. What should happen for s.p.?f if p is nil? Should I get 0? If I get nil that may cause a different error.

The initial comment on the issue says "feel free to add more precise info on how this translates to what needs to be done in text/template" but that is hard for us to do. There are many many many more people suggesting changes than there are people implementing those changes. This can only scale if the people interested in the change can describe exactly what how it should work.

Thanks.

@ianlancetaylor I do not have the knowledge in Go to give specifics, but regarding semantics one can look at https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Optional_chaining which can serve as the spec for how it should work.

But regarding s.p?.f it should return nil (since undefined does not exist in Go?).

Anyone more comfortable in Go; feel free to write up a proper proposal. I can link it in the original issue.

There is no value in nil pointer panics and maybe not much code that depends on it to work properly. Because it just breaks things and doesn't fix anything, the nicest solution would be to make optional chaining the default.

Now I know we can't have nice things because of the remote chance that some code depends on the panics, backward compatibility and all, so the next best thing would be to make it a configuration option on template level.