Do not override user's settings from config

[andreaTP]

Fixes #3378 for Java and C# as a start, if we agree that this is the desired encoding I'll review and propagate the changes to the rest of the languages.

This is a sample result for C#:

#if NETSTANDARD2_1_OR_GREATER || NETCOREAPP3_1_OR_GREATER
#nullable enable
public async Task<double?> MethodName(string b, string c, Action<RequestConfig>? config = default) {
#nullable restore
#else
public async Task<double?> MethodName(string b, string c, Action<RequestConfig> config = default) {
#endif
    if(string.IsNullOrEmpty(b)) throw new ArgumentNullException(nameof(b));
    if(string.IsNullOrEmpty(c)) throw new ArgumentNullException(nameof(c));
    var requestInfo = new RequestInformation {
        HttpMethod = Method.GET,
        UrlTemplate = UrlTemplate,
        PathParameters = PathParameters,
    };
    if (config != null) {
        var requestConfig = new RequestConfig();
        config.Invoke(requestConfig);
        requestInfo.AddQueryParameters(requestConfig.QueryParameters);
        requestInfo.AddRequestOptions(requestConfig.Options);
        requestInfo.AddHeaders(requestConfig.Headers);
    }
    if (!requestInfo.Headers.ContainsKey("Accept")) {
        requestInfo.Headers.Add("Accept", "application/json");
    }
    if (!requestInfo.Headers.ContainsKey("Content-Type")) {
        requestInfo.SetContentFromScalar(RequestAdapter, "", b);
    }
    return requestInfo;
}

[baywet]

Thanks for starting this. To reduce the amount of code we're projecting, I'm suggesting to add TryAdd methods in the headers classes in abstractions. Thoughts?

[andreaTP]

To reduce the amount of code we're projecting, I'm suggesting to add TryAdd methods in the headers classes in abstractions.

This change is going to be not retro compatible and would require, at minimum a minor version bump for the GA languages.

Thoughts?

If we decide to go down this path I'd propose we go full-in and refactor the entire logic into abstractions, something like:

generated:

public async Task<double?> MethodName(string b, string c, Action<RequestConfig> config = default) {
#endif
    if(string.IsNullOrEmpty(b)) throw new ArgumentNullException(nameof(b));
    if(string.IsNullOrEmpty(c)) throw new ArgumentNullException(nameof(c));
    var requestInfo = new RequestInformation {
        HttpMethod = Method.GET,
        UrlTemplate = UrlTemplate,
        PathParameters = PathParameters,
    };
    applyUserConfig(requestInfo, config);
    return requestInfo;
}

library (in abstractions):

public async void applyUserConfig(RequestInfo requestInfo, Action<RequestConfig> config) {
    if (config != null) {
        var requestConfig = new RequestConfig();
        config.Invoke(requestConfig);
        requestInfo.AddQueryParameters(requestConfig.QueryParameters);
        requestInfo.AddRequestOptions(requestConfig.Options);
        requestInfo.AddHeaders(requestConfig.Headers);
    }
    if (!requestInfo.Headers.ContainsKey("Accept")) {
        requestInfo.Headers.Add("Accept", "application/json");
    }
    if (!requestInfo.Headers.ContainsKey("Content-Type")) {
        requestInfo.SetContentFromScalar(RequestAdapter, "", b);
    }
}

wdyt?

[andreaTP]

oh, yes, if we go and refactor the entire thing, adding a TryAdd method is fine 👍

[baywet]

Adding a try add method: We expect users to update dependencies to what the kiota info command indicates as they refresh clients. So if we ship a generation change that requires additional API surface in kiota abstractions, it's not considered a breaking change.

Apply user configuration method: that's a nice idea. The only issues with that:

• The type of the request config varies (so calling new will be impossible in some languages)
• The type of config (the callback) also varies accordingly.
• Some request configuration classes do not have a query parameter property (no documented query parameters)
• Not all methods send a request body (content type header)

For all those reasons, we had originally left this as generated code instead of wrapping everything in the abstractions.

[andreaTP]

Thanks for the context @baywet , make sense.
I'll add the tryAdd method to all of the abstractions and make the changes for this PR.

[andreaTP]

I need to get microsoft/kiota-abstractions-php#91 merged and released to finish this one 😓

[baywet]

I need to get microsoft/kiota-abstractions-php#91 merged and released to finish this one 😓

I just did release 0.8.4 a few minutes ago, you should be good to go. Let me know if you need anything else.

[andreaTP]

This is now ready for review (IT tests are passing on my fork).
I think we have an unrelated regression though: #3401 (comment)

[baywet]

@andreaTP I think we're almost there. Can you add a changelog entry please?

[andreaTP]

sure done 👍