Skip to content

v5.7.0

Compare
Choose a tag to compare
@github-actions github-actions released this 06 Jul 20:07
e16246d

5.7.0 (2023-07-06)

The MongoDB Node.js team is pleased to announce version 5.7.0 of the mongodb package!

Release Notes

Write Concern legacy options deprecated

wtimeout, j, and fsync options have been deprecated, please use wtimeoutMS and journal instead.

Legacy SSL options deprecated

In an effort to simplify TLS setup and use with the driver we're paring down the number of custom options to the ones that are common to all drivers. This should reduce inadvertent misconfiguration due to conflicting options.

The legacy "ssl-" options have been deprecated, each has a corresponding "tls-" option listed in the table below (except for sslCRL, you may directly use the Node.js crl option instead). tlsCertificateFile has also been deprecated, please use tlsCertificateKeyFile or pass the cert directly to the MongoClient constructor.

In addition to the common driver options, the Node.js driver also passes through Node.js TLS options provided on the MongoClient to Node.js' tls.connect API, which may be convenient to reuse with other Node.js APIs.

Node.js native option MongoDB driver option name legacy option name driver option type
ca tlsCAFile sslCA string
crl N/A sslCRL string
cert tlsCertificateKeyFile sslCert string
key tlsCertificateKeyFile sslKey string
passphrase tlsCertificateKeyFilePassword sslPass string
rejectUnauthorized tlsAllowInvalidCertificates sslValidate boolean

New includeResultMetadata option for findOneAnd... family of methods.

This option defaults to true, which will return a ModifyResult type. When set to false, which will
become the default in the next major release, it will return the modified document or null if nothing matched.
This applies to findOneAndDelete, findOneAndUpdate, findOneAndReplace.

// With a document { _id: 1, a: 1 } in the collection
await collection.findOneAndDelete({ a: 1 }, { includeResultMetadata: false }); // returns { _id: 1, a: 1 }
await collection.findOneAndDelete({ a: 2 }, { includeResultMetadata: false }); // returns null
await collection.findOneAndDelete({ a: 1 }, { includeResultMetadata: true }); // returns { ok: 1, lastErrorObject: { n: 1 }, value: { _id: 1, a: 1 }}

Support for change stream split events

When change stream documents exceed the max BSON size limit of 16MB, they can be split into multiple fragments in order to not error when sending events over the wire. In order to enable this functionality, the collection must be created with changeStreamPreAndPostImages enabled and the change stream itself must include an $changeStreamSplitLargeEvent aggregation stage. This feature requires a minimum server version of 7.0.0.

Example:

await db.createCollection('test', { changeStreamPreAndPostImages: { enabled: true }});
const collection = db.collection('test');
const changeStream = collection.watch([{ $changeStreamSplitLargeEvent: {} ], {
  fullDocumentBeforeChange: 'required'
});

for await (const change of changeStream) {
  console.log(change.splitEvent); // If changes over 16MB: { fragment: n, of: n }
}

Programmatic management of search indexes

This PR adds support for managing search indexes (creating, updating, deleting and listing indexes). The new methods are available on the Collection class.

const indexes = await collection.listSearchIndexes().toArray(); // produces an array of search indexes
await collection.createSearchIndex({ name: 'my-index', definition: < index definition > } ); 
await collection.updateSearchIndex('my-index', < new definition >);
await collection.dropSearchIndex('my-index');

BSON updated to v5.4.0

Take a look at the bson package's release notes!

Snappy loaded lazily

Unlike our other compression mechanisms snappy was loaded at the module level, meaning it would be optionally imported whether or not the driver was configured to use snappy compression. Snappy is now aligned with our other optional peer dependencies and is only loaded when enabled.

Zstd and Kerberos are now optional peer dependencies

This allows users who do not use these features to not have them installed. Users who do use these feature will now have them lazy loaded upon first use.

listDatabases nameOnly option bug fix

The listDatabases API exposes the nameOnly option which allows you to limit its output to only the names of the databases on a given mongoDB deployment:

db.admin().listDatabases({ nameOnly: true });
// [
//   { name: 'local' },
//   { name: 'movies' },
//   ...
// ]

Prior to this fix, the option was not being set properly on the command, so the output was always given in full.

Thanks to @redixhumayun for submitting this fix!

saslprep "is not a function" fix for bundled deployments

saslprep is an optional dependency used to perform Stringprep Profile for User Names and Passwords for SCRAM-SHA-256 authentication. The saslprep library breaks when it is bundled, causing the driver to throw TypeErrors.

This release includes a fix that prevents the driver throwing TypeErrors when attempting to use saslprep in bundled environments.

Cursor.map transform bug fixes

The cursor API provides the ability to apply a map function to each document in the cursor:

const cursor = collection.find({ name: 'john doe' }).map(({ name }) => name);
for await (const document of cursor) {
  console.error(document); // only prints the `name` field from each document
}

Cursor iteration properly catches errors thrown from the transform applied in Cursor.map

Starting in version 4.0 of the driver, if the transform function throws an error, there are certain scenarios where the driver does not correctly catch this error and an uncaught exception is thrown:

const cursor = collection.find({ name: 'john doe' }).map(() => {
  throw new Error('oh no! error here'); // 
});
await cursor.next(); // process crashes with uncaught error

This release adds logic to ensure that whenever we transform a cursor document, we handle any errors properly. Any errors thrown from a transform function are caught and returned to the user.

Cursor.hasNext no longer transforms documents if a transform has been applied to the cursor

Version 4.0 introduced a bug that would apply a transform function to documents in the cursor when the cursor was iterated using Cursor.hasNext(). When combined with Cursor.next(), this would result in transforming documents multiple times.

const cursor = collection.find({ name: 'john doe' }).map((document) => document.name);
while (await cursor.hasNext()) { // this transforms the first document in the cursor once
	const doc = await cursor.next(); // the second document in the cursor is transformed again
}

This release removes the transform logic from Cursor.hasNext, preventing cursor documents from being transformed twice when iterated using hasNext.

Features

Bug Fixes

Documentation

We invite you to try the mongodb library immediately, and report any issues to the NODE project.