Azure-cosmos-dotnet-v2: BulkImportAsync does not respect JsonSerializerSettings

[cgelon]

Description:
When using BulkImportAsync, the JsonSerializerSettings passed into DocumentClient is not respected.

Expected:
Objects added to the database with BulkImportAsync should have the same TypeNameHandling as the JsonSerializerSettings in the DocumentClient.

Actual:
No types are ever written in the JSON of the document.

Simplified Repro:

interface IInterface {
    Property { get; set; }
}

class ImplementedInterface : IInterface {
    Property { get; set; }
}

class Document {
    IInterface DocumentProperty { get; set; }
}

var documentClient = new DocumentClient(serializerSettings: new JsonSerializerSettings { TypeNameHandling = TypeNameHandling.All });
var bulkExecutor = new BulkExecutor(documentClient);
await bulkExecutor.BulkImportAsync(new List<Document> { new Document { DocumentProperty = new ImplementedInterface() } });

The Document JSON will not have any types.

My Current Workaround:

await bulkExecutor.BulkImportAsync(documents.Select(document => JsonConvert.SerializeObject(document, BulkImportJsonSerializerSettings)));

Serializing every object before passing it to BulkImportAsync solves the issue of typing, but adds an increase in memory, which builds up when trying to import 100,000s of documents.

[ealsur]

@cgelon Which version of the BulkExecutor library are you using?

[cgelon]

I've used both 1.0.2 and 1.1.0, and both of them had this issue.
I see that the package got updated today to 1.1.1; I will try that and report back.

[ealsur]

@cgelon Please let me know if it repros with 1.1.1 so we can investigate.

[cgelon]

It still repros with 1.1.1.

[danielcrenna]

It still repros with 2.1.0-preview as well.

[aaron-0]

Still an issue in 2.3.0-preview2.

Here is my workaround:

public static async Task<BulkImportResponse> BulkImportAsync(this IBulkExecutor bulkExecutor, JsonSerializerSettings serializerSettings, IEnumerable<object> documents, bool enableUpsert, CancellationToken cancellationToken )
{
    var serializer = JsonSerializer.Create(serializerSettings);
    var serializedDocuments = documents.Select(x => JObject.FromObject(x, serializer));
    var importResponse = await bulkExecutor.BulkImportAsync(serializedDocuments, enableUpsert, cancellationToken: cancellationToken);
    return importResponse;
}

It looks like internally the libary checks first to see if the objects passed in are of type JObject and if so will save them without trying to do serialization.

[danielcrenna]

This appears to be fixed on the latest version stream.