Non-destructive request header access

[kellytk]

I encountered an issue when reordering a handler's extractors that caused form submission data to cease posting as expected.

The offending extractor was HeaderMap and moving it to the end of the extractors solves the problem as the docs advise. I dislike having bugs emerge due to argument order changes and am seeking a permanent solution.

What options are available for accessing request header data without consuming it and causing other extractors to fail?

[davidpdrsn]

The design of extractors doesn't allow for this unfortunately. Extractors that consume the request body have the same problems, such as String, Bytes, and Json. So even if we made HeaderMap clone all the headers (which I don't think we should) body extractors would still have the issue.

I recommend you write own HeaderMapClone extractor that clones the headers and leaves the original ones in place for other extractors.

[davidpdrsn]

There isn't a better way than looking at the implementation unfortunately.

All extractors in axum also have rejections that aim to be clear about what the error was. For example Form's rejection is FormRejection which has a variant called HeadersAlreadyExtracted. That tells you Form wont work if all headers have been extracted first. It needs the headers to look at the Content-Type. Of course this wont always work for extractors provided by other libraries so it should be documented or otherwise clear from the types.

In general I would also recommend extracting only the headers you need via TypedHeader without removing all headers with HeaderMap.

[kellytk]

Thanks once again. Would you accept a PR adding a note to https://docs.rs/axum/0.3.0/axum/extract/struct.TypedHeader.html that it is recommended over HeaderMap wherever possible?

[davidpdrsn]

Definitely!

[kellytk]

Done.

In support of your eventual ideal solution, and as a learning resource and reference to users, what's your opinion of adding a table to https://docs.rs/axum/0.3.0/axum/extract/index.html that details the requirements and mutations for each extractor? For example, the Form extractor's requirement is the Content-Type header, and the HeaderMap's mutation takes the headers.

With the search query https://docs.rs/axum/0.3.0/axum/extract/index.html?search=headers I've already found how I can ferret out much of the information despite being new to axum.

[davidpdrsn]

I feel that might be overkill and tedious to maintain 🤷‍♂️ It's only relevant to consider when using HeaderMap and body extractors and those are already documented.