All notable customer-facing changes to the Google Cloud BigQuery JDBC driver (google-cloud-bigquery-jdbc) are documented below.

• Client-side usage telemetry (opt-out): Starting in version 1.7.0, the driver collects anonymous client-side usage and diagnostic telemetry by default to help improve driver reliability.
  • No identifiable information collected: Telemetry never collects personally identifiable information (PII), credentials, project IDs, SQL query text, query parameter values, dataset or table names, database schema information, or query result data.
  • What is collected: Anonymous environment metadata (driver version, JVM major version, OS platform/version, and a locally generated anonymous correlation ID), connection outcome and authentication method type, high-level operation categories (such as SELECT, INSERT, CREATE TABLE), execution substrate used (REST vs. Storage APIs), latency distributions, feature usage counters, and internal error codes. Collection runs asynchronously in the background and fails silently if network access is restricted.
  • How to disable telemetry: You can opt out and disable telemetry for the entire JVM lifecycle using any of the following methods:
    • Connection string property: Append EnableDiagnosticTelemetry=0; (or EnableDiagnosticTelemetry=false;) to your JDBC URL:

      jdbc:bigquery://https://bigquery.googleapis.com:443;ProjectId=<PROJECT_ID>;OAuthType=<AUTH_TYPE>;EnableDiagnosticTelemetry=0;
      

    • Environment variable:

      export BIGQUERY_JDBC_TELEMETRY_ENABLED=false

    • JVM system property:

      -DBIGQUERY_JDBC_TELEMETRY_ENABLED=false

• Public JDBC methods (execute*, next, commit, rollback, close, and getters) now consistently throw checked SQLException instead of unchecked RuntimeException or IllegalStateException to comply with the JDBC specification.
• ResultSet.getString() and getObject(..., String.class) now format TIMESTAMP values (including nested values in ARRAY and STRUCT) in UTC rather than the local JVM timezone.

• Enabled Post-Quantum Cryptography (PQC) TLS cipher suites (X25519MLKEM768 and MLKEM1024) by default across direct and proxied connections.
• Exposed QueryStatistics on BigQueryResultSet for fast-path (jobless) queries.

• Preserved exact UTC instants when binding Timestamp, Instant, OffsetDateTime, and ZonedDateTime parameters (including Calendar overloads), and bound LocalDateTime as STRING for compatibility with DATETIME columns.
• Honored Calendar timezones and fixed ClassCastException errors in ResultSet temporal getters (getDate, getTime, getTimestamp).
• Corrected the timezone offset direction in PreparedStatement.setTime(..., Calendar).
• Preserved microsecond precision for TIME values (including inside ARRAY and STRUCT) when calling getString() or converting to LocalTime.
• Formatted nested ARRAY fields inside STRUCT values as valid JSON strings in Struct.toString() and ResultSet.getString().
• Stopped background producer threads, cancelled active gRPC streams, and released Arrow buffer memory immediately when ResultSet.close() is called.
• Fixed memory leaks in ResultSet GC finalizer queues and per-connection logger registrations during long-running sessions.
• Preserved default BigQuery SDK HTTP connect and read timeouts when custom timeout properties are not specified.
• Capped the Storage Read API Arrow batch queue at 32 batches per ResultSet to bound heap memory usage on large result sets.

• Added pre-execution parameter type and count inference via dry-run in PreparedStatement, populating getParameterMetaData() and getMetaData() prior to execution while ignoring ? characters inside string literals.

• Automatically terminated active BigQuery sessions when closing the JDBC Connection.

(Includes changes from interim version 1.4.1)

• Added end-to-end support for picosecond-precision timestamps (TIMESTAMP(12)) via the EnableTimestampPicos connection property across REST JSON, Storage Read API (Arrow), nested ARRAY/STRUCT/RANGE types, PreparedStatement parameters, and metadata APIs.
• Added support for colon-delimited project qualifiers (project:dataset and project:warehouse.namespace) in the DefaultDataset connection property for Lakehouse and multi-catalog environments.

• Fixed SQL NULL handling in ResultSet primitive getters (getInt, getLong, getBoolean, etc.) and getBlob/getClob to return JDBC-compliant defaults (0, false, null) instead of throwing NullPointerException.
• Handled null parameter bindings properly during Storage Write API bulk inserts.
• Fixed automatic fallback from the Storage Read API to the REST API when permission errors occur.

• Preserved local wall-clock time without unintended UTC shifts when reading DATETIME values inside ARRAY and RANGE columns.
• Enforced strict numeric overflow and underflow bounds checking and rejected invalid cross-temporal coercions (Date to Time and vice versa) in ResultSet getters.
• Resolved statement types via getJob fallback to avoid unnecessary post-execution dry-run queries for DDL statements.

• Eliminated redundant synchronous jobs.get RPC calls on the fast query execution path.

• Unified SQL and Java type mapping and coercion across read, write, and metadata paths through a centralized type registry.

• Fixed session ID propagation across sequential statements when EnableSession=1 is enabled.
• Propagated connection proxy settings (ProxyHost, ProxyPort) to OpenTelemetry span exporters.
• Prevented OpenTelemetry log handlers from being detached when driver logger levels or handlers are reconfigured.
• Fixed exact schema name matching in DatabaseMetaData.getExportedKeys() catalog fallback queries.

• Eliminated pre-execution dry-run RPCs previously used to classify statement types and DML row counts.
• Accelerated DatabaseMetaData queries and prevented thread-pool starvation deadlocks using a two-tier parallel execution model.

• Added a comprehensive USER_GUIDE.md detailing connection URL properties, authentication methods, and custom endpoint configuration.

• DatabaseMetaData methods now throw a SQLException on 403 Forbidden access errors and project-level 404 Not Found errors instead of silently returning empty results.

• Added OpenTelemetry distributed tracing, metrics, and Cloud Logging support, including automatic trace and span ID injection into local log files.
• Added SSLTrustStoreType and SSLTrustStoreProvider connection properties for custom TLS trustStore configurations.
• Implemented ParameterMetaData with dynamic IN, OUT, and INOUT parameter mode support and completed PreparedStatement parameter setter methods.

• Fixed Storage Write API bulk inserts in custom endpoint and VPC-SC environments by routing JsonStreamWriter through the configured client settings.
• Fixed timezone coercion, Calendar defensive copying, and wasNull() state tracking across PreparedStatement and CallableStatement.

• Accelerated DatabaseMetaData constraint queries (getPrimaryKeys, getImportedKeys, getExportedKeys, getCrossReference) using native BigQuery API and hybrid metadata lookups.
• Reduced memory footprint during JSON result set streaming.

• Added the EnableProjectDiscovery connection property to allow DatabaseMetaData (getCatalogs, getSchemas, getTables) to discover datasets across all accessible Google Cloud projects.
• Respected standard JVM trustStore system properties (javax.net.ssl.trustStore and related properties) by default for TLS connections.
• Implemented ResultSet fetch size, fetch direction, and SQL warning methods (getFetchSize, setFetchSize, getFetchDirection, setFetchDirection, getWarnings, clearWarnings).

• Prevented unintended transaction rollbacks when closing statements in manual commit mode (autoCommit=false).
• Propagated connection proxy settings (ProxyHost, ProxyPort, ProxyUid, ProxyPwd) to Google Auth Library requests for OAuth token retrieval.
• Fixed large result destination table handling (AllowLargeResults, LargeResultsDataset, LargeResultsTable) in PreparedStatement and automatically matched dataset location when creating temporary datasets.
• Corrected FilterTablesOnDefaultDataset fallback behavior when empty strings or wildcard patterns are supplied to DatabaseMetaData.
• Fixed BigQueryDriver to report accurate major and minor driver version numbers instead of 0.0.
• Shaded org.slf4j classes in the standalone uber JAR to prevent classpath conflicts with host applications.
• Preserved accurate page row counts across paginated REST TableResult responses.
• Propagated thread interruptions and exceptions in background metadata fetchers as SQLException instead of returning truncated results.

• Isolated query and metadata thread pools per connection using daemon threads to prevent cross-connection blocking and allow clean JVM shutdown.
• Optimized DatabaseMetaData catalog and schema enumeration with parallel project scanning, API field masking, and cached job statistics.
• Eliminated eager stack-trace walking overhead during result set type coercion when SLF4J logging bridges are active.

(Version 0.11.0 was promoted to 1.0.0 GA with identical functionality)

• Validated integer connection properties and DataSource setters at configuration time, rejecting negative values with a driver exception.

• Implemented JDBC 4.2 ResultSet.getObject(column, Class<T>) type conversion support.
• Added support for executing EXPORT DATA, EXPORT MODEL, and LOAD DATA SQL statements.
• Implemented JDBC Wrapper methods (unwrap and isWrapperFor) across Connection, Statement, ResultSet, DataSource, DatabaseMetaData, and ResultSetMetaData.
• Completed per-connection log isolation across background worker threads and result sets.

• Fixed manual commit mode (autoCommit=false) query failures when operating on datasets in non-US regions.
• Added automatic fallback to the REST API when the Storage Read API is blocked by IAM permission errors.
• Fixed timezone shifts when converting epoch values to java.sql.Time and java.sql.Timestamp in Arrow and JSON result sets.
• Standardized precision, scale, and SQL type mappings for numeric, temporal, STRING (NVARCHAR), GEOGRAPHY, JSON, and INTERVAL types in ResultSetMetaData and DatabaseMetaData.
• Added support for backslash (\) escape characters in DatabaseMetaData wildcard pattern filters.
• Isolated Statement instances per DatabaseMetaData query to prevent state interference across concurrent metadata calls.
• Initialized driver logging prior to JDBC URL parsing so connection string warnings and errors are captured in logs.
• Updated standalone uber JAR shading rules to prevent dependency conflicts.

• Tuned Storage Read API eligibility heuristics and increased the default HighThroughputMinTableSize threshold to 10,000 rows to avoid gRPC session overhead on small results.

• Added per-connection log files using unique UUID connection identifiers so concurrent connections can write isolated diagnostic logs.
• Bypassed dry-run job creation when authenticating with read-only OAuth tokens (bigquery.readonly scope).

• Added support for Service Account Impersonation when authenticating with Application Default Credentials (ADC).
• Provided default OAuth client ID and secret values to simplify user OAuth authentication flows.
• Improved DatabaseMetaData compliance and implemented Struct.getSQLTypeName() for nested STRUCT values.
• Enhanced driver exception messages and logs to include full causal chains, stack traces, and originating caller context.

• Added the Google Drive OAuth scope across all credential types to support querying Drive-backed external tables.
• Scoped DatabaseMetaData instances to their parent Connection and lazily initialized internal Statement resources.

• Refined dependency bundling and shading rules in the standalone uber JAR artifact.

• Fixed Statement.getMoreResults() to consistently advance past the current result cursor.
• Ignored unrecognized properties in the JDBC connection URL with a warning instead of failing the connection.

• Unshaded com.google.cloud.bigquery SDK classes in the uber JAR to maintain compatibility with applications using the BigQuery client library.

• Synchronized all supported connection properties across JDBC URLs, BigQueryConnection, and DataSource getters and setters.

• Reduced connection initialization latency by optimizing credential creation and connection setup.

• Enforced strict JDBC URL syntax validation, rejecting malformed connection strings instead of silently ignoring invalid segments.

• Added support for escaping special characters in JDBC URL property values.

• Added automatic stream retry handling for transient failures in the BigQuery Storage Read API (Arrow).

• Enabled gRPC keep-alive pings for the BigQuery Storage Read API to prevent idle stream disconnects on long-running reads.
• Added the RequestReason connection property to attach audit reasons to BigQuery API requests.
• Added HttpConnectTimeout and HttpReadTimeout connection properties to configure REST transport timeouts.

• Returned null from Driver.connect() when given a non-BigQuery JDBC URL, allowing DriverManager to delegate to other registered JDBC drivers.

• Avoided eager string formatting in disabled log statements to reduce CPU and allocation overhead.

• Initial preview release of the official Google Cloud BigQuery JDBC driver (google-cloud-bigquery-jdbc), supporting SQL query execution, PreparedStatement and CallableStatement, multi-statement transactions, Google Cloud authentication flows, the High-Throughput Storage Read API (Arrow), and the Storage Write API.