feat(codegen): general data mapping function

[kuhe]

Problem:

Object field names are often repeated multiple times per line in code that is only transferring fields from one object to another (mapping), with repetitive checks.

Examples

Request headers

const headers = {
    "content-type": "application/xml",
    ...(isSerializableHeaderValue(input.ChecksumCRC32) && { "x-amz-checksum-crc32": input.ChecksumCRC32 }),
    ...(isSerializableHeaderValue(input.ChecksumCRC32C) && { "x-amz-checksum-crc32c": input.ChecksumCRC32C }),
    ...(isSerializableHeaderValue(input.ChecksumSHA1) && { "x-amz-checksum-sha1": input.ChecksumSHA1 }),
    ...(isSerializableHeaderValue(input.ChecksumSHA256) && { "x-amz-checksum-sha256": input.ChecksumSHA256 }),
    ...(isSerializableHeaderValue(input.RequestPayer) && { "x-amz-request-payer": input.RequestPayer }),
    ...(isSerializableHeaderValue(input.ExpectedBucketOwner) && {
        "x-amz-expected-bucket-owner": input.ExpectedBucketOwner,
    }),
    ...(isSerializableHeaderValue(input.SSECustomerAlgorithm) && {
        "x-amz-server-side-encryption-customer-algorithm": input.SSECustomerAlgorithm,
    }),
    ...(isSerializableHeaderValue(input.SSECustomerKey) && {
        "x-amz-server-side-encryption-customer-key": input.SSECustomerKey,
    }),
    ...(isSerializableHeaderValue(input.SSECustomerKeyMD5) && {
        "x-amz-server-side-encryption-customer-key-md5": input.SSECustomerKeyMD5,
    }),
};

Response headers

if (output.headers["x-amz-expiration"] !== undefined) {
    contents.Expiration = output.headers["x-amz-expiration"];
}
if (output.headers["etag"] !== undefined) {
    contents.ETag = output.headers["etag"];
}
if (output.headers["x-amz-checksum-crc32"] !== undefined) {
    contents.ChecksumCRC32 = output.headers["x-amz-checksum-crc32"];
}
if (output.headers["x-amz-checksum-crc32c"] !== undefined) {
    contents.ChecksumCRC32C = output.headers["x-amz-checksum-crc32c"];
}
if (output.headers["x-amz-checksum-sha1"] !== undefined) {
    contents.ChecksumSHA1 = output.headers["x-amz-checksum-sha1"];
}

Body deserialization

if (data["Bucket"] !== undefined) {
    contents.Bucket = __expectString(data["Bucket"]);
  }
  if (data["ChecksumCRC32"] !== undefined) {
    contents.ChecksumCRC32 = __expectString(data["ChecksumCRC32"]);
  }
  if (data["ChecksumCRC32C"] !== undefined) {
    contents.ChecksumCRC32C = __expectString(data["ChecksumCRC32C"]);
  }
  if (data["ChecksumSHA1"] !== undefined) {
    contents.ChecksumSHA1 = __expectString(data["ChecksumSHA1"]);
  }
  if (data["ChecksumSHA256"] !== undefined) {
    contents.ChecksumSHA256 = __expectString(data["ChecksumSHA256"]);
  }
  if (data["ETag"] !== undefined) {
    contents.ETag = __expectString(data["ETag"]);
  }
  if (data["Key"] !== undefined) {
    contents.Key = __expectString(data["Key"]);
  }
  if (data["Location"] !== undefined) {
    contents.Location = __expectString(data["Location"]);
  }

... and more.

Proposed solution:

A general purpose data mapping function. It accepts transfer instructions in a brevity form:

{
    key: value, // unfiltered form
    key: [filter, value], // immediate filtered form
    key: [(value) => boolean, () => value] // lazy form
}

Where the key's value in the target object is created with a combination of the filter and the value, both of which can be value expressions or lazy functional providers.

source: https://github.com/aws/aws-sdk-js-v3/blob/800e8e97ea1f184111e2a85208b07d68fb7297b2/packages/smithy-client/src/object-mapping.ts

Also targeted were the following repeated patterns:

Resolved path construction, XML node creation, and default error deserialization. These were turned into helper functions.

Effects

The change for S3 in for example dist-cjs and dist-es:
dist-cjs/protocols -> 568kb to 440kb
dist-es/protocols -> 692kb to 556kb

overall reduction is approximately 0.1mb out of 1.2 to 1.3mb.

[kuhe]

aws/aws-sdk-js-v3#3830