dvt-core 0.52.2__cp310-cp310-macosx_10_9_x86_64.whl

dbt/compute/jdbc_utils.py

@@ -0,0 +1,678 @@
+"""
+JDBC Utilities for Spark Engine
+
+Provides utilities for converting dbt adapter credentials to JDBC configurations
+and helpers for optimizing parallel reads via partitioning.
+
+This module enables DVT to bypass memory bottlenecks by using Spark JDBC connectors
+to read data directly from source databases into Spark workers (distributed).
+
+Architecture:
+- Maps adapter credentials → JDBC URL + properties
+- Auto-detects optimal partition columns for parallel reads
+- Estimates partition bounds for efficient data distribution
+"""
+
+from typing import Dict, Optional, Tuple
+
+from dbt.adapters.base import BaseAdapter
+from dbt.adapters.contracts.connection import Credentials
+from dbt_common.exceptions import DbtRuntimeError
+
+
+# JDBC driver class mapping for database types
+# DVT v0.5.9: Complete support for all dbt adapters with JDBC connectivity
+JDBC_DRIVER_MAPPING = {
+    # ============================================================
+    # Cloud Data Warehouses
+    # ============================================================
+    "postgres": "org.postgresql.Driver",
+    "postgresql": "org.postgresql.Driver",
+    "snowflake": "net.snowflake.client.jdbc.SnowflakeDriver",
+    "bigquery": "com.simba.googlebigquery.jdbc.Driver",
+    "redshift": "com.amazon.redshift.jdbc.Driver",
+    "databricks": "com.databricks.client.jdbc.Driver",
+    "firebolt": "com.firebolt.FireboltDriver",
+
+    # ============================================================
+    # Microsoft Ecosystem (all use same JDBC driver)
+    # ============================================================
+    "sqlserver": "com.microsoft.sqlserver.jdbc.SQLServerDriver",
+    "mssql": "com.microsoft.sqlserver.jdbc.SQLServerDriver",
+    "fabric": "com.microsoft.sqlserver.jdbc.SQLServerDriver",
+    "synapse": "com.microsoft.sqlserver.jdbc.SQLServerDriver",
+
+    # ============================================================
+    # Enterprise Data Warehouses
+    # ============================================================
+    "oracle": "oracle.jdbc.OracleDriver",
+    "db2": "com.ibm.db2.jcc.DB2Driver",
+    "teradata": "com.teradata.jdbc.TeraDriver",
+    "exasol": "com.exasol.jdbc.EXADriver",
+    "vertica": "com.vertica.jdbc.Driver",
+
+    # ============================================================
+    # SQL Engines & Query Platforms
+    # ============================================================
+    "spark": "org.apache.hive.jdbc.HiveDriver",
+    "trino": "io.trino.jdbc.TrinoDriver",
+    "presto": "io.prestosql.jdbc.PrestoDriver",
+    "athena": "com.simba.athena.jdbc.Driver",
+    "hive": "org.apache.hive.jdbc.HiveDriver",
+    "impala": "com.cloudera.impala.jdbc.Driver",
+    "dremio": "com.dremio.jdbc.Driver",
+    "glue": "com.amazonaws.glue.sql.jdbc.Driver",
+
+    # ============================================================
+    # Open Source Databases
+    # ============================================================
+    "mysql": "com.mysql.cj.jdbc.Driver",
+    "mariadb": "org.mariadb.jdbc.Driver",
+    "sqlite": "org.sqlite.JDBC",
+    "duckdb": "org.duckdb.DuckDBDriver",
+    "cratedb": "io.crate.client.jdbc.CrateDriver",
+
+    # ============================================================
+    # OLAP & Analytics Databases
+    # ============================================================
+    "clickhouse": "com.clickhouse.jdbc.ClickHouseDriver",
+    "singlestore": "com.singlestore.jdbc.Driver",
+    "starrocks": "com.mysql.cj.jdbc.Driver",  # StarRocks uses MySQL protocol
+    "doris": "com.mysql.cj.jdbc.Driver",  # Apache Doris uses MySQL protocol
+    "greenplum": "org.postgresql.Driver",  # Greenplum uses PostgreSQL protocol
+    "monetdb": "org.monetdb.jdbc.MonetDriver",
+
+    # ============================================================
+    # Time-Series & Streaming
+    # ============================================================
+    "timescaledb": "org.postgresql.Driver",  # TimescaleDB uses PostgreSQL
+    "questdb": "org.postgresql.Driver",  # QuestDB supports PostgreSQL wire protocol
+    "materialize": "org.postgresql.Driver",  # Materialize uses PostgreSQL wire protocol
+    "rockset": "com.rockset.jdbc.RocksetDriver",
+
+    # ============================================================
+    # Graph & Multi-Model
+    # ============================================================
+    "neo4j": "org.neo4j.Driver",
+
+    # ============================================================
+    # Data Lake Formats (via Spark connectors)
+    # ============================================================
+    "delta": "org.apache.hive.jdbc.HiveDriver",  # Delta Lake via Spark
+    "iceberg": "org.apache.hive.jdbc.HiveDriver",  # Apache Iceberg via Spark
+    "hudi": "org.apache.hive.jdbc.HiveDriver",  # Apache Hudi via Spark
+
+    # ============================================================
+    # AlloyDB (Google - PostgreSQL compatible)
+    # ============================================================
+    "alloydb": "org.postgresql.Driver",  # AlloyDB is PostgreSQL-compatible
+}
+
+
+def _rewrite_localhost_for_docker(jdbc_url: str) -> str:
+    """
+    Rewrite localhost/127.0.0.1 to host.docker.internal for Docker Spark clusters.
+
+    DVT v0.51.8: When using Docker-based Spark clusters, workers inside containers
+    need host.docker.internal to reach the host machine. With host.docker.internal
+    also added to the host's /etc/hosts (pointing to 127.0.0.1), the same JDBC URL
+    works for both driver (on host) and workers (in containers).
+
+    :param jdbc_url: Original JDBC URL
+    :returns: JDBC URL with localhost replaced by host.docker.internal
+    """
+    import re
+    # Replace localhost or 127.0.0.1 with host.docker.internal
+    url = re.sub(r'//localhost([:/?])', r'//host.docker.internal\1', jdbc_url)
+    url = re.sub(r'//127\.0\.0\.1([:/?])', r'//host.docker.internal\1', url)
+    return url
+
+
+# Global flag to enable Docker JDBC URL rewriting
+_docker_mode_enabled = False
+
+
+def set_docker_mode(enabled: bool) -> None:
+    """Enable or disable Docker mode for JDBC URL rewriting."""
+    global _docker_mode_enabled
+    _docker_mode_enabled = enabled
+
+
+def build_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """
+    Build JDBC configuration from dbt adapter credentials.
+
+    Converts adapter-specific credentials to JDBC URL and connection properties
+    that can be used by Spark JDBC connectors.
+
+    :param credentials: dbt adapter credentials object
+    :returns: Tuple of (jdbc_url, jdbc_properties)
+    :raises DbtRuntimeError: If adapter type is not supported or credentials are invalid
+
+    Example:
+        >>> from dbt.adapters.postgres import PostgresCredentials
+        >>> creds = PostgresCredentials(
+        ...     host="localhost",
+        ...     port=5432,
+        ...     user="analytics",
+        ...     password="secret",
+        ...     database="warehouse",
+        ...     schema="public"
+        ... )
+        >>> url, props = build_jdbc_config(creds)
+        >>> print(url)
+        jdbc:postgresql://localhost:5432/warehouse
+        >>> print(props)
+        {'user': 'analytics', 'password': 'secret', 'driver': 'org.postgresql.Driver'}
+    """
+    adapter_type = credentials.type.lower()
+
+    # Check if adapter type is supported
+    if adapter_type not in JDBC_DRIVER_MAPPING:
+        raise DbtRuntimeError(
+            f"JDBC connectivity not supported for adapter type '{adapter_type}'. "
+            f"Supported types: {', '.join(JDBC_DRIVER_MAPPING.keys())}"
+        )
+
+    # Build JDBC URL and properties based on adapter type
+    if adapter_type in ("postgres", "postgresql"):
+        jdbc_url, jdbc_props = _build_postgres_jdbc_config(credentials)
+    elif adapter_type == "mysql":
+        jdbc_url, jdbc_props = _build_mysql_jdbc_config(credentials)
+    elif adapter_type == "snowflake":
+        jdbc_url, jdbc_props = _build_snowflake_jdbc_config(credentials)
+    elif adapter_type == "redshift":
+        jdbc_url, jdbc_props = _build_redshift_jdbc_config(credentials)
+    elif adapter_type == "bigquery":
+        jdbc_url, jdbc_props = _build_bigquery_jdbc_config(credentials)
+    elif adapter_type in ("sqlserver", "mssql"):
+        jdbc_url, jdbc_props = _build_sqlserver_jdbc_config(credentials)
+    elif adapter_type == "oracle":
+        jdbc_url, jdbc_props = _build_oracle_jdbc_config(credentials)
+    elif adapter_type == "databricks":
+        jdbc_url, jdbc_props = _build_databricks_jdbc_config(credentials)
+    else:
+        raise DbtRuntimeError(
+            f"JDBC configuration builder not implemented for adapter type '{adapter_type}'"
+        )
+
+    # DVT v0.51.8: Rewrite localhost URLs for Docker Spark clusters
+    if _docker_mode_enabled:
+        jdbc_url = _rewrite_localhost_for_docker(jdbc_url)
+
+    return jdbc_url, jdbc_props
+
+
+def _build_postgres_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for PostgreSQL."""
+    creds_dict = credentials.to_dict()
+
+    host = creds_dict.get("host", "localhost")
+    port = creds_dict.get("port", 5432)
+    database = creds_dict.get("database")
+    user = creds_dict.get("user")
+    password = creds_dict.get("password", "")
+
+    if not database:
+        raise DbtRuntimeError("PostgreSQL credentials missing required field: database")
+    if not user:
+        raise DbtRuntimeError("PostgreSQL credentials missing required field: user")
+
+    jdbc_url = f"jdbc:postgresql://{host}:{port}/{database}"
+
+    jdbc_properties = {
+        "user": user,
+        "password": password,
+        "driver": JDBC_DRIVER_MAPPING["postgres"],
+    }
+
+    # Optional: Add SSL configuration if present
+    if creds_dict.get("sslmode"):
+        jdbc_properties["ssl"] = "true" if creds_dict["sslmode"] != "disable" else "false"
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_mysql_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for MySQL."""
+    creds_dict = credentials.to_dict()
+
+    host = creds_dict.get("host", "localhost")
+    port = creds_dict.get("port", 3306)
+    database = creds_dict.get("database")
+    user = creds_dict.get("user")
+    password = creds_dict.get("password", "")
+
+    if not database:
+        raise DbtRuntimeError("MySQL credentials missing required field: database")
+    if not user:
+        raise DbtRuntimeError("MySQL credentials missing required field: user")
+
+    jdbc_url = f"jdbc:mysql://{host}:{port}/{database}"
+
+    jdbc_properties = {
+        "user": user,
+        "password": password,
+        "driver": JDBC_DRIVER_MAPPING["mysql"],
+    }
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_snowflake_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for Snowflake."""
+    creds_dict = credentials.to_dict()
+
+    account = creds_dict.get("account")
+    user = creds_dict.get("user")
+    password = creds_dict.get("password", "")
+    database = creds_dict.get("database")
+    warehouse = creds_dict.get("warehouse")
+    schema = creds_dict.get("schema", "public")
+
+    if not account:
+        raise DbtRuntimeError("Snowflake credentials missing required field: account")
+    if not user:
+        raise DbtRuntimeError("Snowflake credentials missing required field: user")
+
+    # Snowflake JDBC URL format with Arrow disabled via URL parameter
+    # This is more reliable than JDBC properties for Snowflake driver
+    jdbc_url = f"jdbc:snowflake://{account}.snowflakecomputing.com/?JDBC_QUERY_RESULT_FORMAT=JSON"
+
+    jdbc_properties = {
+        "user": user,
+        "password": password,
+        "driver": JDBC_DRIVER_MAPPING["snowflake"],
+        # CRITICAL FIX v0.4.4: Disable Arrow format to avoid Java 21 module access errors
+        # Property must be uppercase and set in BOTH URL and properties for reliability
+        "JDBC_QUERY_RESULT_FORMAT": "JSON",
+        "jdbc_query_result_format": "json",  # Lowercase variant for compatibility
+        # Additional Snowflake-specific optimizations
+        "JDBC_USE_SESSION_TIMEZONE": "false",  # Use UTC for consistency
+    }
+
+    # Add optional properties
+    if database:
+        jdbc_properties["db"] = database
+    if warehouse:
+        jdbc_properties["warehouse"] = warehouse
+    if schema:
+        jdbc_properties["schema"] = schema
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_redshift_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for Amazon Redshift."""
+    creds_dict = credentials.to_dict()
+
+    host = creds_dict.get("host")
+    port = creds_dict.get("port", 5439)
+    database = creds_dict.get("database")
+    user = creds_dict.get("user")
+    password = creds_dict.get("password", "")
+
+    if not host:
+        raise DbtRuntimeError("Redshift credentials missing required field: host")
+    if not database:
+        raise DbtRuntimeError("Redshift credentials missing required field: database")
+    if not user:
+        raise DbtRuntimeError("Redshift credentials missing required field: user")
+
+    jdbc_url = f"jdbc:redshift://{host}:{port}/{database}"
+
+    jdbc_properties = {
+        "user": user,
+        "password": password,
+        "driver": JDBC_DRIVER_MAPPING["redshift"],
+    }
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_bigquery_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for Google BigQuery."""
+    creds_dict = credentials.to_dict()
+
+    project = creds_dict.get("project")
+    dataset = creds_dict.get("dataset") or creds_dict.get("schema")
+
+    if not project:
+        raise DbtRuntimeError("BigQuery credentials missing required field: project")
+
+    # BigQuery JDBC URL format
+    jdbc_url = "jdbc:bigquery://https://www.googleapis.com/bigquery/v2:443"
+
+    jdbc_properties = {
+        "ProjectId": project,
+        "driver": JDBC_DRIVER_MAPPING["bigquery"],
+    }
+
+    if dataset:
+        jdbc_properties["DefaultDataset"] = dataset
+
+    # Handle authentication
+    # BigQuery typically uses service account JSON or OAuth
+    if creds_dict.get("keyfile"):
+        jdbc_properties["OAuthType"] = "0"  # Service account
+        jdbc_properties["OAuthServiceAcctEmail"] = creds_dict.get("client_email", "")
+        jdbc_properties["OAuthPvtKeyPath"] = creds_dict["keyfile"]
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_sqlserver_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for Microsoft SQL Server."""
+    creds_dict = credentials.to_dict()
+
+    host = creds_dict.get("host", "localhost")
+    port = creds_dict.get("port", 1433)
+    database = creds_dict.get("database")
+    user = creds_dict.get("user")
+    password = creds_dict.get("password", "")
+
+    if not database:
+        raise DbtRuntimeError("SQL Server credentials missing required field: database")
+    if not user:
+        raise DbtRuntimeError("SQL Server credentials missing required field: user")
+
+    jdbc_url = f"jdbc:sqlserver://{host}:{port};databaseName={database}"
+
+    jdbc_properties = {
+        "user": user,
+        "password": password,
+        "driver": JDBC_DRIVER_MAPPING["sqlserver"],
+    }
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_oracle_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """Build JDBC config for Oracle Database."""
+    creds_dict = credentials.to_dict()
+
+    host = creds_dict.get("host", "localhost")
+    port = creds_dict.get("port", 1521)
+    database = creds_dict.get("database") or creds_dict.get("service_name")
+    user = creds_dict.get("user")
+    password = creds_dict.get("password", "")
+
+    if not database:
+        raise DbtRuntimeError("Oracle credentials missing required field: database/service_name")
+    if not user:
+        raise DbtRuntimeError("Oracle credentials missing required field: user")
+
+    # Oracle thin driver format
+    jdbc_url = f"jdbc:oracle:thin:@{host}:{port}:{database}"
+
+    jdbc_properties = {
+        "user": user,
+        "password": password,
+        "driver": JDBC_DRIVER_MAPPING["oracle"],
+    }
+
+    return jdbc_url, jdbc_properties
+
+
+def _build_databricks_jdbc_config(credentials: Credentials) -> Tuple[str, Dict[str, str]]:
+    """
+    Build JDBC config for Databricks SQL Warehouse or Cluster.
+
+    DVT v0.51.5: Added support for Databricks JDBC connectivity.
+
+    Databricks JDBC URL format:
+    jdbc:databricks://<host>:443/default;transportMode=http;ssl=1;httpPath=<http_path>;AuthMech=3;
+
+    The dbt-databricks adapter credentials include:
+    - host: Databricks workspace URL (e.g., dbc-xxxxx.cloud.databricks.com)
+    - http_path: SQL warehouse or cluster HTTP path
+    - token: Personal access token for authentication
+    - catalog: Unity Catalog name (optional)
+    - schema: Default schema (optional)
+    """
+    creds_dict = credentials.to_dict()
+
+    host = creds_dict.get("host")
+    http_path = creds_dict.get("http_path")
+    token = creds_dict.get("token")
+    catalog = creds_dict.get("catalog", "hive_metastore")
+    schema = creds_dict.get("schema", "default")
+
+    if not host:
+        raise DbtRuntimeError("Databricks credentials missing required field: host")
+    if not http_path:
+        raise DbtRuntimeError("Databricks credentials missing required field: http_path")
+    if not token:
+        raise DbtRuntimeError("Databricks credentials missing required field: token")
+
+    # Build Databricks JDBC URL
+    # Format: jdbc:databricks://<host>:443/<catalog>;transportMode=http;ssl=1;httpPath=<http_path>;AuthMech=3;
+    jdbc_url = (
+        f"jdbc:databricks://{host}:443/{catalog};"
+        f"transportMode=http;ssl=1;httpPath={http_path};AuthMech=3"
+    )
+
+    jdbc_properties = {
+        "UID": "token",  # Databricks uses "token" as username for PAT auth
+        "PWD": token,
+        "driver": JDBC_DRIVER_MAPPING["databricks"],
+    }
+
+    return jdbc_url, jdbc_properties
+
+
+def auto_detect_partition_column(adapter: BaseAdapter, schema: str, table: str) -> Optional[str]:
+    """
+    Auto-detect the best column for partitioning parallel JDBC reads.
+
+    Queries table metadata to find a suitable partition column. Prioritizes:
+    1. Primary key columns (single column PKs only)
+    2. Columns named 'id' or ending with '_id'
+    3. Timestamp/date columns
+    4. Integer columns
+
+    :param adapter: dbt adapter to use for querying metadata
+    :param schema: Schema/dataset name
+    :param table: Table name
+    :returns: Column name suitable for partitioning, or None if not found
+
+    Example:
+        >>> column = auto_detect_partition_column(adapter, "public", "users")
+        >>> if column:
+        ...     print(f"Using {column} for partitioning")
+        ... else:
+        ...     print("No suitable partition column found")
+    """
+    try:
+        # Strategy 1: Check for primary key
+        pk_column = _get_primary_key_column(adapter, schema, table)
+        if pk_column:
+            return pk_column
+
+        # Strategy 2: Get all columns and look for ID-like columns
+        columns = _get_table_columns(adapter, schema, table)
+
+        # Look for ID columns (exact match or suffix)
+        for col_name, col_type in columns:
+            col_name_lower = col_name.lower()
+            if col_name_lower == "id" or col_name_lower.endswith("_id"):
+                # Check if it's an integer type
+                if _is_integer_type(col_type):
+                    return col_name
+
+        # Strategy 3: Look for timestamp/date columns
+        for col_name, col_type in columns:
+            if _is_timestamp_type(col_type):
+                return col_name
+
+        # Strategy 4: Look for any integer column
+        for col_name, col_type in columns:
+            if _is_integer_type(col_type):
+                return col_name
+
+        # No suitable column found
+        return None
+
+    except Exception:
+        # If metadata query fails, return None (caller can decide to read without partitioning)
+        return None
+
+
+def estimate_partition_bounds(
+    adapter: BaseAdapter, schema: str, table: str, column: str
+) -> Tuple[int, int]:
+    """
+    Estimate partition bounds (min/max) for a numeric partition column.
+
+    Queries the table to get MIN and MAX values of the partition column,
+    which are used by Spark JDBC to distribute reads across workers.
+
+    :param adapter: dbt adapter to use for querying
+    :param schema: Schema/dataset name
+    :param table: Table name
+    :param column: Partition column name
+    :returns: Tuple of (lower_bound, upper_bound)
+    :raises DbtRuntimeError: If query fails or column is not numeric
+
+    Example:
+        >>> lower, upper = estimate_partition_bounds(adapter, "public", "orders", "order_id")
+        >>> print(f"Partition range: {lower} to {upper}")
+        Partition range: 1 to 1000000
+    """
+    try:
+        # Build qualified table name
+        qualified_table = f"{schema}.{table}"
+
+        # Query for min/max
+        sql = f"SELECT MIN({column}) as min_val, MAX({column}) as max_val FROM {qualified_table}"
+
+        # Execute via adapter
+        response, result_table = adapter.execute(sql, auto_begin=False, fetch=True)
+
+        if not result_table or len(result_table.rows) == 0:
+            raise DbtRuntimeError(
+                f"Failed to estimate partition bounds for {qualified_table}.{column}: "
+                "Query returned no results"
+            )
+
+        row = result_table.rows[0]
+        min_val = row[0]
+        max_val = row[1]
+
+        if min_val is None or max_val is None:
+            raise DbtRuntimeError(
+                f"Failed to estimate partition bounds for {qualified_table}.{column}: "
+                "Column contains only NULL values"
+            )
+
+        # Convert to integers
+        lower_bound = int(min_val)
+        upper_bound = int(max_val)
+
+        return lower_bound, upper_bound
+
+    except Exception as e:
+        raise DbtRuntimeError(
+            f"Failed to estimate partition bounds for {schema}.{table}.{column}: {str(e)}"
+        ) from e
+
+
+# Helper functions for metadata queries
+
+
+def _get_primary_key_column(adapter: BaseAdapter, schema: str, table: str) -> Optional[str]:
+    """
+    Get primary key column name (if single-column PK exists).
+
+    Implementation is adapter-specific. Returns None if not implemented
+    or if PK is composite.
+    """
+    adapter_type = adapter.type().lower()
+
+    try:
+        if adapter_type in ("postgres", "postgresql", "redshift"):
+            # PostgreSQL/Redshift: Query information_schema
+            sql = f"""
+            SELECT a.attname
+            FROM pg_index i
+            JOIN pg_attribute a ON a.attrelid = i.indrelid AND a.attnum = ANY(i.indkey)
+            WHERE i.indrelid = '{schema}.{table}'::regclass
+            AND i.indisprimary
+            """
+            response, result = adapter.execute(sql, auto_begin=False, fetch=True)
+            if result and len(result.rows) == 1:
+                return result.rows[0][0]
+
+        elif adapter_type == "mysql":
+            # MySQL: Query information_schema
+            sql = f"""
+            SELECT COLUMN_NAME
+            FROM information_schema.KEY_COLUMN_USAGE
+            WHERE TABLE_SCHEMA = '{schema}'
+            AND TABLE_NAME = '{table}'
+            AND CONSTRAINT_NAME = 'PRIMARY'
+            """
+            response, result = adapter.execute(sql, auto_begin=False, fetch=True)
+            if result and len(result.rows) == 1:
+                return result.rows[0][0]
+
+        # For other adapters or if query fails, return None
+        return None
+
+    except Exception:
+        return None
+
+
+def _get_table_columns(adapter: BaseAdapter, schema: str, table: str) -> list[Tuple[str, str]]:
+    """
+    Get list of (column_name, column_type) for a table.
+    """
+    adapter_type = adapter.type().lower()
+
+    try:
+        if adapter_type in ("postgres", "postgresql", "redshift"):
+            sql = f"""
+            SELECT column_name, data_type
+            FROM information_schema.columns
+            WHERE table_schema = '{schema}'
+            AND table_name = '{table}'
+            ORDER BY ordinal_position
+            """
+            response, result = adapter.execute(sql, auto_begin=False, fetch=True)
+            return [(row[0], row[1]) for row in result.rows]
+
+        elif adapter_type == "mysql":
+            sql = f"""
+            SELECT COLUMN_NAME, DATA_TYPE
+            FROM information_schema.COLUMNS
+            WHERE TABLE_SCHEMA = '{schema}'
+            AND TABLE_NAME = '{table}'
+            ORDER BY ORDINAL_POSITION
+            """
+            response, result = adapter.execute(sql, auto_begin=False, fetch=True)
+            return [(row[0], row[1]) for row in result.rows]
+
+        else:
+            # Fallback: Use LIMIT 0 query to get columns
+            sql = f"SELECT * FROM {schema}.{table} LIMIT 0"
+            response, result = adapter.execute(sql, auto_begin=False, fetch=True)
+            # Return column names with unknown types
+            return [(col, "unknown") for col in result.column_names]
+
+    except Exception:
+        return []
+
+
+def _is_integer_type(sql_type: str) -> bool:
+    """Check if SQL type is an integer type."""
+    sql_type_upper = sql_type.upper()
+    return any(
+        int_type in sql_type_upper
+        for int_type in ["INT", "INTEGER", "BIGINT", "SMALLINT", "SERIAL"]
+    )
+
+
+def _is_timestamp_type(sql_type: str) -> bool:
+    """Check if SQL type is a timestamp/date type."""
+    sql_type_upper = sql_type.upper()
+    return any(time_type in sql_type_upper for time_type in ["TIMESTAMP", "DATETIME", "DATE"])