Version 1.15.0

Features:

• Support for sharded transactions on MongoDB sharded clusters 4.2+.
• Add convenient transaction runner (mongoc_client_session_with_transaction), which accepts a callback and performs appropriate retry logic.
• Add a new transaction option to specify maximum time to wait for a commit, mongoc_transaction_opts_set_max_commit_time_ms.
• Add URI option "retryReads=true" safely and automatically retries certain read operations if the server is a MongoDB 3.6+.
• Poll SRV records to mongos servers periodically.
• Keep connections alive after a primary stepdown detected.
• Standardizes URI options supported across all spec-compliant MongoDB drivers.
• "retryWrites" URI option now defaults to true (requires crypto for session support).
• Send any aggregate with $out or $merge stage to a primary.
• Add the ability to specify an aggregate pipeline as an update document.
• Add a database aggregate helper, mongoc_database_aggregate.
• Add option for change streams, "startAfter".
• Add mongoc_change_stream_get_resume_token, which returns the resume token which should be used to resume a change stream.
• Add support for zstd compression.

Bug fixes:

• Correctly report an error in mongoc_change_stream_next if the resume token (_id) is not a document. Previously, an error was only reported if the field was missing.
• Fix mongoc_collection_update with MONGOC_UPDATE_MULTI_UPDATE, mongoc_collection_remove, and mongoc_collection_delete when retryWrites was enabled. They would fail previously.
• Command options are now correctly taken into account when batching bulk writes for OP_QUERY. It was possible to exceed the maximum document size before.
• Fix a crash if a multi-batch bulk write with OP_MSG errored on a batch.