Unable to access the xmlns values

[estk]

I'm writing a code generator for xml-schemas and I'm in the process of trying to sort out the namespace that an element or type may be associated with. Unfortunately without access to the xml namespace information from an xml document I'm unable to do this. I could implement a custom deserializer, but I figured it would be better to fold the work in here.

I was thinking a nice way to do this would be with a Serde field attribute, maybe something like the following.

use xml::namespace::NamespaceStack;
...
    #[serde(rename = "$namespaces")]
    nss: HashMap<String, NamespaceStack>

We could even re-export a new-type wrapper around the hash map.

The code generator is a major work in process, but its here if you're curious.

[punkstarman]

@estk, regarding the feature request, I think that this is better kept for a custom deserializer. What you effectively want is a model of XML (or maybe just XML Schema). If you look at serde_json and serde_yaml, they have custom ser/de for mapping to JSON and YAML respectively (see https://github.com/serde-rs/json/tree/master/src/value and https://github.com/dtolnay/serde-yaml/tree/master/src/value). I would rather implement this feature in the same manner.

Regarding your code generator project, it is highly dependent on design choices made in this project. Just saying.

[estk]

@punkstarman To clarify, I am building that model which is a pre-requisite of the code generation project. Ideally I would be able to derive the ser/de code necessary with this project. It would be an awful lot more work to write all of that custom deserialization code, but perhaps you're right about the tight coupling. The reason I was under the impression that it would belong here is that somehow xml seems incomplete if the user is unable to access some of the attributes (the ns ones as an example).

Anyway it strikes me now that I should be able to have the best of both worlds by slapping a #[serde(deserialize_with = "path")] attribute on the namespace fields so I think this feature request is unnecessary. Thanks for talking it through and feel free to close the pr.