`rawBody` http property should not be a UTF-8 string for byte data

[mhoeger]

Ideally, the body that we pass to language workers should always be the raw form (ex: if the body is application/json, then it should be the string version and not the parse'd object). If the language worker wants to add a method on top of that to get a "pretty" object, that is ok, but it should be language-dependent and as an additional method (not the default behavior).

One solution to passing only raw HTTP bodies is to use the rawBody property.

However - rawBody is not always the true raw body. In cases where the body is encoded as bytes, rawBody is actually the UTF-8 string version of the body (code here).

This issue tracks making rawBody consistent, with the ultimate goal of passing the "rawest" form of data possible.

cc: @pragnagopa and @maiqbal11 to add any details

[Lytigas]

I was running into this issue locally the other day in Node. Since the update docs issue is closed, I'm wondering if this behavior is still being looked at. I rewrote my code in another language to get around it. (Content-Type headers can fix it, but don't help in my situation).

$ func
Azure Functions Core Tools (3.0.2245 Commit hash: 1d094e2f3ef79b9a478a1621ea7ec3f93ac1910d)
Function Runtime Version: 3.0.13139.0

I never tested it on Azure itself.

[mhoeger]

This is a pretty significant breaking change and we didn't take it for v3.x. It's still a big consideration for future versions.

What other language did you choose to do this? And if you need content-type to be added to all requests (say you don't necessarily have control over that), you can use functions proxies

[brettsam]

@pragnagopa -- do you have context on this? Is it something we should/could take for v4?