chore: add Into conversions for Blob to [u8]

[ajewellamz]

Motivation and Context

Often I have a &[u8] or a Vec, but the sdk function takes a Blob.
It would be convenient to pass the my_vec instead of aws_smithy_types::Blob::new(my_vec)

Description

In blob.rs I added
Into<Vec> for Blob
Into for Vec
Into for &[u8]

Testing

Added a test to blob.rs to test the various scenarios.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[ysaito1001]

Thanks for contributing. I have two questions.
Q1.

but the sdk function takes a Blob.

Is it possible that you could point us to an example function, just to be on the same page?

Q2. What's the reason for also adding

impl From<Blob> for Vec<u8> {
    fn from(value: Blob) -> Self {
        value.into_inner()
    }
}

when Blob already has .into_inner? I see the other two impl From make it more ergonomic when passing Vec<u8> and &[u8] to a function taking impl Into<Blob>. Do you also have a use case where there is a function that takes impl Into<Vec<u8>> and you need to pass a Blob?

[ajewellamz]

Thanks for contributing. I have two questions. Q1.

but the sdk function takes a Blob.

Is it possible that you could point us to an example function, just to be on the same page?

How about aws_sdk_dynamodb::types::AttributeValue::B?

Better yet, look at the docs for aws-sdk-kms and search for Blob as a parameter.
You'll find many many things that take blobs as input and return blobs as output.

Q2. What's the reason for also adding impl From<Blob> for Vec<u8>

I dunno. Completeness I guess, or symmetry . I'll delete it, if it's a problem.
But why provide only a non-standard way, when a standard way exists?
Many constructors, including Blob, look like this :
pub fn new<T: Into<Vec<u8>>>(input: T) -> Self
and so into_inner does no good. I has to be Into

[ysaito1001]

How about aws_sdk_dynamodb::types::AttributeValue::B?

Better yet, look at the docs for aws-sdk-kms and search for Blob as a parameter. You'll find many many things that take blobs as input and return blobs as output.

Hmm in aws-sdk-kms, I don't see methods that take specifically impl Into<Blob> for some reason.

➜  kms git:(main) pwd
/Users/awsaito/src/aws-sdk-rust/sdk/kms
➜  kms git:(main) rg Blob | wc -l  
     329
➜  kms git:(main) rg Blob | tail -n 10
src/operation/generate_mac/_generate_mac_input.rs:    pub fn get_message(&self) -> &::std::option::Option<::aws_smithy_types::Blob> {
src/operation/generate_mac/builders.rs:    pub fn message(mut self, input: ::aws_smithy_types::Blob) -> Self {
src/operation/generate_mac/builders.rs:    pub fn set_message(mut self, input: ::std::option::Option<::aws_smithy_types::Blob>) -> Self {
src/operation/generate_mac/builders.rs:    pub fn get_message(&self) -> &::std::option::Option<::aws_smithy_types::Blob> {
src/types/_recipient_info.rs:    pub attestation_document: ::std::option::Option<::aws_smithy_types::Blob>,
src/types/_recipient_info.rs:    pub fn attestation_document(&self) -> ::std::option::Option<&::aws_smithy_types::Blob> {
src/types/_recipient_info.rs:    pub(crate) attestation_document: ::std::option::Option<::aws_smithy_types::Blob>,
src/types/_recipient_info.rs:    pub fn attestation_document(mut self, input: ::aws_smithy_types::Blob) -> Self {
src/types/_recipient_info.rs:    pub fn set_attestation_document(mut self, input: ::std::option::Option<::aws_smithy_types::Blob>) -> Self {
src/types/_recipient_info.rs:    pub fn get_attestation_document(&self) -> &::std::option::Option<::aws_smithy_types::Blob> {
➜  kms git:(main) rg Blob | rg Into | wc -l
       0
➜  kms git:(main) rg Blob | rg impl | wc -l
       0

(I don't either in /Users/awsaito/src/aws-sdk-rust/sdk/kms/dynamodb)

Can you send us a URL to the doc for a function/method that takes impl Into<Blob>, otherwise even after providing impl Into<Vec> for Blob or impl From<&[u8]> for Blob, callers still have to call .into() at the call sites before calling the method in question (which seems as ergonomic as calling Blob::new, IMO).

[ajewellamz]

Can you send us a URL to the doc for a function/method that takes impl Into<Blob>, otherwise even after providing impl Into<Vec> for Blob or impl From<&[u8]> for Blob, callers still have to call .into() at the call sites before calling the method in question (which seems as ergonomic as calling Blob::new, IMO).

I was only referring to functions that that a Blob, not Into

There's a chicken and egg problem. Why support an Into parameter type, if there aren't any types that support Into yet.

Personally, I think .into() is much preferable to Blob::new, because it's more idiomatic and more discoverable. The compiler will tell you "use .into()" whereas without it, one must dig through the documentation to find Blob::new.

[ysaito1001]

I was only referring to functions that that a Blob, not Into

There's a chicken and egg problem. Why support an Into parameter type, if there aren't any types that support Into yet.

Personally, I think .into() is much preferable to Blob::new, because it's more idiomatic and more discoverable. The compiler will tell you "use .into()" whereas without it, one must dig through the documentation to find Blob::new.

Ah, I see where you're coming from. These methods in kms are code generated, so even after this PR has been merged, call sites still have to call .into() to pass a Blob to the methods. We'll create a backlog task to update the methods signatures to take impl Into<Blob> (no estimated delivery time at this time).

Thanks again for contributing. Could you bump the version of aws-smithy-types (as shown in CI error) to maybe 1.2.8 and push the change? We'll merge it in.

Canary passed