feat(core): experimental runtime model interpreter 🚧🚧

[kuhe]

🚧🚧🚧

Prototyping serde alternative that accepts a Smithy json model at runtime.

AwsRestJson1 implementation is in aws/aws-sdk-js-v3#5935

Goals:

• performance and code size parity or improvement vs. current code gen serde
• allow runtime provision of models such as a regular Smithy JSON model file or compressed form
• swap protocols at runtime in the same client?? 🤔

[kuhe]

testing of the AwsRestJson1 protocol implementation shows the following:

Sizes are in kb
stat (input files total size) / parsed (output bundle file size) / gzip (bundle plus gzip compression)

// Lambda bare-bones client from prod
/**
 * 366/129/33
 */

// Lambda aggregated client from prod
/**
 * 591/223/46
 */

// creating the aggregated client using the bare-bones client and the runtime model interpreter
// using the default lambda.json smithy model file.
/**
 * 764/524/88
 */

// creating the aggregated client using the bare-bones client and the runtime model interpreter
// using a compressed smithy model file (using pattern recognition algorithm)
// this would be the type used in production, hypothetically
/**
 * 475/215/51
 */

The conclusion is that runtime model interpretation, at least for JSON, is approximately the same as the existing aggregated client. It doesn't present meaningful performance/size benefits other than enabling modification of the model/protocol at runtime.

[kuhe]

sample usage

// import bare-bones client with no commands on it
import { LambdaClient } from "@aws-sdk/client-lambda";
// import protocol implementation
import { AwsRestJson1 } from "@aws-sdk/core";
// Smithy json model
import model from "../../../scripts/smithy-model/temp/lambda.min";

const awsRestJson1 = new AwsRestJson1(model);

// create aggregated client from model in runtime
const LambdaAggregatedClient = awsRestJson1.create(LambdaClient);

// aggregated client is able to call methods
new LambdaAggregatedClient({}).getFunction({ FunctionName: "" });
new LambdaAggregatedClient({}).invoke({ FunctionName: "", Payload: "" });