home.social

#azuresdk — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #azuresdk, aggregated by home.social.

fetched live
  1. For anyone interested, apparently #dotnet #azureSDK based on Microsoft.Rest does have a default exponential back off retry policy that is not obvious. It's part of ServiceClient and you can call `SetRetryPolicy(null)` to turn off retries.

  2. Look. I get it. You don't want to make an `enum` for your constrained string values collection so you make it a `struct` instead.

    But for the love of got, then, make it implicitly convertible to a string. (or, did you know there's an `EnumMember` attribute? even better!)

    I'm looking at you #AzureSDK folks.

    #azure #azureAI

  3. Prototyping test recording for for : github.com/heaths/recorded-tes

    Attribute sync or async tests using `#[recorded]` and can accept optional `TestContext` parameter. Full implementation should automatically set up HTTP transport to record or play back.

    Thoughts? Still early.

  4. TIL that the azure-mgmt-core library (#AzureSDK for #Python) actually contains some useful helper functions:
    - parse_resource_id(), which returns the parts of the id
    - resource_id(), which builds a resource id based on the parameters passed into the function
    - is_valid_resource_id(), to validate a resource id
    - is_valid_resource_name(), to validate a resource name against ARM guidelines
    - get_arm_endpoints(), to return ARM endpoint and ARM credential scopes

    github.com/Azure/azure-sdk-for

    #azure

  5. @vbfox @khalidabuhakmeh an RFC is already approved and partly implemented that accomplishes some of this. We’re hoping it’s stabilized before we GA our first for Rust crates.

  6. This may be enough for us to ditch our plans of using builders over ugly nested `..Default::default()` calls for nested client options in the for rust. It wasn’t the only reason, but a significant one.
    hachyderm.io/@ekuber/113369453

  7. We've been discussing internally and leaning toward transport-level mocking - fakes - like we do in our for : github.com/Azure/azure-sdk-for

    Basically, a struct which implements our `Policy` to be used as our transport in an already-pluggable pipeline and declares distinct `Arc<dyn Fn(..) -> impl Future<..>` fields that are mapped to endpoint calls.

    Seems the most flexible and requires no change to the client API since we already support replacing the transport.

  8. Would love feedback on mocking service clients for for : play.rust-lang.org/?version=st

    Our current design requires pulling a trait into scope, and while rust-analyzer or other LSPs makes that easy not everything uses them: azure.github.io/azure-sdk/rust

    A colleague came up with a better way I prototyped above to better demo our current public API. The only caveat seems no worse than what we have now and should affect few developers, we believe.

  9. @ekis @[email protected] I often work with monorepos like the for . That's when git worktrees and sparse checkouts can help reduce checkout time and drive space exhaustion. Check out heaths.dev/tips/2022/06/18/red for some tips. I should probably update that to mention worktrees too, though that really just helps reduce cloning time and minimizes disk space.

    One advantage is having different copies of dependencies in different states, though I wonder how often that's a problem.

  10. Great article with basic concepts to use Azure SDK 👉 devblogs.microsoft.com/azure-s Since Azure SDK follow the same guidelines, using them is much easier and nicer. Unfortunately, some SDK (Cosmos, Azure IoT) still go their own way, but globally the dev experience has improved.

    Yet, I still sometimes come across code that wraps Azure SDK. People, please stop that, it's useless and it prevents other developers to learn how to correctly use Azure SDK.

    #azure #azuresdk

  11. The `azd` has been available for a short while as a feature, which I recently used in github.com/heaths/azcrypto: github.com/heaths/azcrypto/pul

    It will not only install `azd`, but the dev extension for . It makes deploying resources and applications a breeze.

    I use it in azcrypto in lieu of the resource provisioning scripts I originally wrote for all of to deploy resources uniformly. This is a publicly available (near) equivalent.

  12. devblogs.microsoft.com/azure-s

    Azure Core for all our language SDKs plays a core part of our HTTP pipeline (and other protocols in some cases, like AMQP), serialization, and more. My colleague, Anne, does a great job explaining a lot of it here.

  13. Now that the for has released v1.0.0, I have updated to it and released v1 of pkg.go.dev/github.com/heaths/a : a cryptography client for Key Vault and that not only makes it easier to call crypto operations but tries to first cache the public key and do public key operations locally to improve performance and help mitigate throttling.

    We have this in our other languages' SDKs but doesn't fit our design goals for , so I wrote it as a separate module.

  14. Will today be the day I figure out how to migrate from Microsoft.Azure.Management.Fluent to Azure.ResourceManager? 🤞 Only time will tell. #azure #dotnet #azuresdk

  15. @larryosterman to clarify for others interested in or , we - including others from different departments at - have established guidelines for both service APIs (typically REST) as well as client libraries we write and generate. Our code generators follow those same guidelines.

    is designed to bridge those concepts. With the same expressive grammar, service teams have more central control (with guard rails) over both service and client code.