connection.py

"""
Copyright (c) Microsoft Corporation.
Licensed under the MIT license.
This module defines the Connection class, which is used to manage a connection to a database.
The class provides methods to establish a connection, create cursors, commit transactions,
roll back transactions, and close the connection.
Resource Management:
- All cursors created from this connection are tracked internally.
- When close() is called on the connection, all open cursors are automatically closed.
- Do not use any cursor after the connection is closed; doing so will raise an exception.
- Cursors are also cleaned up automatically when no longer referenced, to prevent memory leaks.
"""

import weakref
import re
import codecs
import warnings
import struct
from types import MappingProxyType
from typing import Any, Dict, Optional, Union, List, Tuple, Callable, Protocol, TYPE_CHECKING
import threading

import mssql_python
from mssql_python.cursor import Cursor
from mssql_python.helpers import (
    sanitize_user_input,
    validate_attribute_value,
)
from mssql_python.connection_string_parser import sanitize_connection_string
from mssql_python.logging import logger
from mssql_python import ddbc_bindings
from mssql_python.pooling import PoolingManager
from mssql_python.odbc_provider import ProviderManager
from mssql_python.exceptions import (
    Warning,  # pylint: disable=redefined-builtin
    Error,
    InterfaceError,
    DatabaseError,
    DataError,
    OperationalError,
    IntegrityError,
    InternalError,
    ProgrammingError,
    NotSupportedError,
    sqlstate_to_exception,
)
from mssql_python.auth import (
    extract_auth_type,
    process_auth_parameters,
    remove_sensitive_params,
    get_auth_token_info,
    compute_identity_key,
    compute_token_identity,
)
from mssql_python.constants import ConstantsDDBC, GetInfoConstants
from mssql_python.connection_string_parser import _ConnectionStringParser
from mssql_python.connection_string_builder import _ConnectionStringBuilder
from mssql_python.constants import (
    _RESERVED_PARAMETERS,
    _KEY_AUTHENTICATION,
    _KEY_UID,
    _KEY_PWD,
    _KEY_TRUSTED_CONNECTION,
    _AuthInternal,
)

if TYPE_CHECKING:
    from mssql_python.row import Row


class TokenProvider(Protocol):
    """Structural type for the ``token_provider`` parameter.

    Any object exposing a ``get_token(scope)`` method that returns an object
    with a ``.token`` attribute (the raw JWT string) satisfies this protocol.
    It is intentionally broader than ``azure.core.credentials.TokenCredential``
    -- whose ``get_token`` requires the ``(*scopes, **kwargs)`` shape -- so that
    a minimal ``get_token(scope)`` implementation (for example, a thin wrapper
    around a token obtained from Microsoft Fabric's ``mssparkutils``) is
    accepted by static type checkers, matching the documented contract. Every
    ``azure-identity`` credential (``DefaultAzureCredential``,
    ``AzureCliCredential``, ...) also conforms.
    """

    def get_token(self, scope: str) -> Any:
        """Return an object with a ``.token`` attribute for ``scope``."""
        ...


# Add SQL_WMETADATA constant for metadata decoding configuration
SQL_WMETADATA: int = -99  # Special flag for column name decoding
INFO_TYPE_STRING_THRESHOLD: int = 10000  # Legacy fallback for unlisted information types

# ODBC integers have fixed widths and native byte order; handles are pointer-sized.
_SQLUSMALLINT = struct.Struct("=H")
_SQLUINTEGER = struct.Struct("=I")
_SQLPOINTER = struct.Struct("P")

# SQLGetInfoW return types. Only type information is shared, never connection metadata.
_GETINFO_RETURN_TYPES = MappingProxyType(
    {
        GetInfoConstants.SQL_DATA_SOURCE_NAME.value: str,
        GetInfoConstants.SQL_DATABASE_NAME.value: str,
        GetInfoConstants.SQL_DRIVER_NAME.value: str,
        GetInfoConstants.SQL_DRIVER_VER.value: str,
        GetInfoConstants.SQL_SERVER_NAME.value: str,
        GetInfoConstants.SQL_USER_NAME.value: str,
        GetInfoConstants.SQL_DRIVER_ODBC_VER.value: str,
        GetInfoConstants.SQL_IDENTIFIER_QUOTE_CHAR.value: str,
        GetInfoConstants.SQL_CATALOG_NAME_SEPARATOR.value: str,
        GetInfoConstants.SQL_CATALOG_TERM.value: str,
        GetInfoConstants.SQL_SCHEMA_TERM.value: str,
        GetInfoConstants.SQL_TABLE_TERM.value: str,
        GetInfoConstants.SQL_KEYWORDS.value: str,
        GetInfoConstants.SQL_PROCEDURE_TERM.value: str,
        GetInfoConstants.SQL_SPECIAL_CHARACTERS.value: str,
        GetInfoConstants.SQL_SEARCH_PATTERN_ESCAPE.value: str,
        GetInfoConstants.SQL_ACCESSIBLE_PROCEDURES.value: str,
        GetInfoConstants.SQL_ACCESSIBLE_TABLES.value: str,
        GetInfoConstants.SQL_DATA_SOURCE_READ_ONLY.value: str,
        GetInfoConstants.SQL_EXPRESSIONS_IN_ORDERBY.value: str,
        GetInfoConstants.SQL_LIKE_ESCAPE_CLAUSE.value: str,
        GetInfoConstants.SQL_MULTIPLE_ACTIVE_TXN.value: str,
        GetInfoConstants.SQL_NEED_LONG_DATA_LEN.value: str,
        GetInfoConstants.SQL_PROCEDURES.value: str,
        GetInfoConstants.SQL_CATALOG_NAME.value: str,
        GetInfoConstants.SQL_COLUMN_ALIAS.value: str,
        GetInfoConstants.SQL_DESCRIBE_PARAMETER.value: str,
        GetInfoConstants.SQL_ORDER_BY_COLUMNS_IN_SELECT.value: str,
        GetInfoConstants.SQL_OUTER_JOINS.value: str,
        GetInfoConstants.SQL_MULT_RESULT_SETS.value: str,
        GetInfoConstants.SQL_DRIVER_HLIB.value: _SQLPOINTER,
        GetInfoConstants.SQL_DRIVER_HENV.value: _SQLPOINTER,
        GetInfoConstants.SQL_DRIVER_HDBC.value: _SQLPOINTER,
        GetInfoConstants.SQL_SQL_CONFORMANCE.value: _SQLUINTEGER,
        GetInfoConstants.SQL_IDENTIFIER_CASE.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_SUBQUERIES.value: _SQLUINTEGER,
        GetInfoConstants.SQL_CORRELATION_NAME.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_CATALOG_USAGE.value: _SQLUINTEGER,
        GetInfoConstants.SQL_SCHEMA_USAGE.value: _SQLUINTEGER,
        GetInfoConstants.SQL_TXN_CAPABLE.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_TXN_ISOLATION_OPTION.value: _SQLUINTEGER,
        GetInfoConstants.SQL_DEFAULT_TXN_ISOLATION.value: _SQLUINTEGER,
        GetInfoConstants.SQL_NUMERIC_FUNCTIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_STRING_FUNCTIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_TIMEDATE_FUNCTIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_SYSTEM_FUNCTIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_CONVERT_FUNCTIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_MAX_COLUMN_NAME_LEN.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_TABLE_NAME_LEN.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_SCHEMA_NAME_LEN.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_CATALOG_NAME_LEN.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_IDENTIFIER_LEN.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_STATEMENT_LEN.value: _SQLUINTEGER,
        GetInfoConstants.SQL_MAX_CHAR_LITERAL_LEN.value: _SQLUINTEGER,
        GetInfoConstants.SQL_MAX_BINARY_LITERAL_LEN.value: _SQLUINTEGER,
        GetInfoConstants.SQL_MAX_COLUMNS_IN_TABLE.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_COLUMNS_IN_SELECT.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_COLUMNS_IN_GROUP_BY.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_COLUMNS_IN_ORDER_BY.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_COLUMNS_IN_INDEX.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_TABLES_IN_SELECT.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_CONCURRENT_ACTIVITIES.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_DRIVER_CONNECTIONS.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_MAX_ROW_SIZE.value: _SQLUINTEGER,
        GetInfoConstants.SQL_MAX_USER_NAME_LEN.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_GETDATA_EXTENSIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_CURSOR_COMMIT_BEHAVIOR.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_CURSOR_ROLLBACK_BEHAVIOR.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_CURSOR_SENSITIVITY.value: _SQLUINTEGER,
        GetInfoConstants.SQL_BOOKMARK_PERSISTENCE.value: _SQLUINTEGER,
        GetInfoConstants.SQL_DYNAMIC_CURSOR_ATTRIBUTES1.value: _SQLUINTEGER,
        GetInfoConstants.SQL_DYNAMIC_CURSOR_ATTRIBUTES2.value: _SQLUINTEGER,
        GetInfoConstants.SQL_FORWARD_ONLY_CURSOR_ATTRIBUTES1.value: _SQLUINTEGER,
        GetInfoConstants.SQL_FORWARD_ONLY_CURSOR_ATTRIBUTES2.value: _SQLUINTEGER,
        GetInfoConstants.SQL_STATIC_CURSOR_ATTRIBUTES1.value: _SQLUINTEGER,
        GetInfoConstants.SQL_STATIC_CURSOR_ATTRIBUTES2.value: _SQLUINTEGER,
        GetInfoConstants.SQL_KEYSET_CURSOR_ATTRIBUTES1.value: _SQLUINTEGER,
        GetInfoConstants.SQL_KEYSET_CURSOR_ATTRIBUTES2.value: _SQLUINTEGER,
        GetInfoConstants.SQL_SCROLL_OPTIONS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_SCROLL_CONCURRENCY.value: _SQLUINTEGER,
        GetInfoConstants.SQL_FETCH_DIRECTION.value: _SQLUINTEGER,
        GetInfoConstants.SQL_STATIC_SENSITIVITY.value: _SQLUINTEGER,
        GetInfoConstants.SQL_BATCH_SUPPORT.value: _SQLUINTEGER,
        GetInfoConstants.SQL_BATCH_ROW_COUNT.value: _SQLUINTEGER,
        GetInfoConstants.SQL_PARAM_ARRAY_ROW_COUNTS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_PARAM_ARRAY_SELECTS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_POSITIONED_STATEMENTS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_GROUP_BY.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_OJ_CAPABILITIES.value: _SQLUINTEGER,
        GetInfoConstants.SQL_QUOTED_IDENTIFIER_CASE.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_CONCAT_NULL_BEHAVIOR.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_NULL_COLLATION.value: _SQLUSMALLINT,
        GetInfoConstants.SQL_ALTER_TABLE.value: _SQLUINTEGER,
        GetInfoConstants.SQL_UNION.value: _SQLUINTEGER,
        GetInfoConstants.SQL_DDL_INDEX.value: _SQLUINTEGER,
        GetInfoConstants.SQL_TIMEDATE_ADD_INTERVALS.value: _SQLUINTEGER,
        GetInfoConstants.SQL_TIMEDATE_DIFF_INTERVALS.value: _SQLUINTEGER,
        # Standard information types not currently exposed by GetInfoConstants.
        17: str,  # SQL_DBMS_NAME
        18: str,  # SQL_DBMS_VER
        10000: str,  # SQL_XOPEN_CLI_YEAR
        10021: _SQLUINTEGER,  # SQL_ASYNC_MODE
        127: _SQLUINTEGER,  # SQL_CREATE_ASSERTION
    }
)

# UTF-16 encoding variants that should use SQL_WCHAR by default
# Note: "utf-16" with BOM is NOT included as it's problematic for SQL_WCHAR
UTF16_ENCODINGS: frozenset[str] = frozenset(["utf-16le", "utf-16be"])

_SQLSTATE_RE = re.compile(r"^SQLSTATE:([A-Z0-9]{0,5}):(.*)", re.DOTALL)


def _raise_connection_error(e: RuntimeError) -> None:
    """Map a RuntimeError from the C++ pybind layer to the correct DB-API 2.0 exception.

    Connection::checkError() throws "SQLSTATE:XXXXX:<odbc_message>" so the SQLSTATE
    can be mapped via sqlstate_to_exception(), consistent with cursor-level error handling.
    """
    error_msg = str(e)
    match = _SQLSTATE_RE.match(error_msg)
    if match:
        sqlstate, ddbc_error = match.group(1), match.group(2)
        # Handle malformed SQLSTATE prefix (empty or invalid code)
        if not sqlstate or len(sqlstate) != 5:
            logger.error("Connection error (malformed SQLSTATE): %s", ddbc_error)
            raise OperationalError(
                driver_error="Connection operation failed",
                ddbc_error=ddbc_error,
            ) from None
        exc = sqlstate_to_exception(sqlstate, ddbc_error)
        if exc is None:
            logger.error("Unknown SQLSTATE %s, raising DatabaseError", sqlstate)
            raise DatabaseError(
                driver_error=f"An error occurred with SQLSTATE code: {sqlstate}",
                ddbc_error=ddbc_error,
            ) from None
        logger.error("Connection error (SQLSTATE %s): %s", sqlstate, ddbc_error)
        raise exc from None
    # Fallback: no SQLSTATE prefix — e.g. "Connection handle not allocated"
    logger.error("Connection error: %s", error_msg)
    raise OperationalError(
        driver_error="Connection operation failed",
        ddbc_error=error_msg,
    ) from None


def _validate_utf16_wchar_compatibility(
    encoding: str, wchar_type: int, context: str = "SQL_WCHAR"
) -> None:
    """
    Validates UTF-16 encoding compatibility with SQL_WCHAR.

    Centralizes the validation logic to eliminate duplication across setencoding/setdecoding.

    Args:
        encoding: The encoding string (already normalized to lowercase)
        wchar_type: The SQL_WCHAR constant value to check against
        context: Context string for error messages ('SQL_WCHAR', 'SQL_WCHAR ctype', etc.)

    Raises:
        ProgrammingError: If encoding is incompatible with SQL_WCHAR
    """
    if encoding == "utf-16":
        # UTF-16 with BOM is rejected due to byte order ambiguity
        logger.warning("utf-16 with BOM rejected for %s", context)
        raise ProgrammingError(
            driver_error="UTF-16 with Byte Order Mark not supported for SQL_WCHAR",
            ddbc_error=(
                "Cannot use 'utf-16' encoding with SQL_WCHAR due to Byte Order Mark ambiguity. "
                "Use 'utf-16le' or 'utf-16be' instead for explicit byte order."
            ),
        )
    elif encoding not in UTF16_ENCODINGS:
        # Non-UTF-16 encodings are not supported with SQL_WCHAR
        logger.warning(
            "Non-UTF-16 encoding %s attempted with %s", sanitize_user_input(encoding), context
        )

        # Generate context-appropriate error messages
        if "ctype" in context:
            driver_error = "SQL_WCHAR ctype only supports UTF-16 encodings"
            ddbc_context = "SQL_WCHAR ctype"
        else:
            driver_error = "SQL_WCHAR only supports UTF-16 encodings"
            ddbc_context = "SQL_WCHAR"

        raise ProgrammingError(
            driver_error=driver_error,
            ddbc_error=(
                f"Cannot use encoding '{encoding}' with {ddbc_context}. "
                f"SQL_WCHAR requires UTF-16 encodings (utf-16le, utf-16be)"
            ),
        )


def _validate_encoding(encoding: str) -> bool:
    """
    Cached encoding validation using codecs.lookup().

    Args:
        encoding (str): The encoding name to validate.

    Returns:
        bool: True if encoding is valid, False otherwise.

    Note:
        Uses LRU cache to avoid repeated expensive codecs.lookup() calls.
        Cache size is limited to 128 entries which should cover most use cases.
        Also validates that encoding name only contains safe characters.
    """
    # Basic security checks - prevent obvious attacks
    if not encoding or not isinstance(encoding, str):
        return False

    # Check length limit (prevent DOS)
    if len(encoding) > 100:
        return False

    # Prevent null bytes and control characters that could cause issues
    if "\x00" in encoding or any(ord(c) < 32 and c not in "\t\n\r" for c in encoding):
        return False

    # Then check if it's a valid Python codec
    try:
        codecs.lookup(encoding)
        return True
    except LookupError:
        return False


class Connection:
    """
    A class to manage a connection to a database, compliant with DB-API 2.0 specifications.

    This class provides methods to establish a connection to a database, create cursors,
    commit transactions, roll back transactions, and close the connection. It is designed
    to be used in a context where database operations are required, such as executing queries
    and fetching results.

    The Connection class supports the Python context manager protocol (with statement).
    When used as a context manager, it will automatically close the connection when
    exiting the context, ensuring proper resource cleanup.

    Example usage:
        with connect(connection_string) as conn:
            cursor = conn.cursor()
            cursor.execute("INSERT INTO table VALUES (?)", [value])
        # Connection is automatically closed when exiting the with block

    For long-lived connections, use without context manager:
        conn = connect(connection_string)
        try:
            # Multiple operations...
        finally:
            conn.close()

    Methods:
        __init__(database: str) -> None:
        connect_to_db() -> None:
        cursor() -> Cursor:
        commit() -> None:
        rollback() -> None:
        close() -> None:
        __enter__() -> Connection:
        __exit__() -> None:
        setencoding(encoding=None, ctype=None) -> None:
        setdecoding(sqltype, encoding=None, ctype=None) -> None:
        getdecoding(sqltype) -> dict:
        set_attr(attribute, value) -> None:
    """

    # DB-API 2.0 Exception attributes
    # These allow users to catch exceptions using connection.Error,
    # connection.ProgrammingError, etc.
    Warning = Warning
    Error = Error
    InterfaceError = InterfaceError
    DatabaseError = DatabaseError
    DataError = DataError
    OperationalError = OperationalError
    IntegrityError = IntegrityError
    InternalError = InternalError
    ProgrammingError = ProgrammingError
    NotSupportedError = NotSupportedError

    def __init__(
        self,
        connection_str: str = "",
        autocommit: bool = False,
        attrs_before: Optional[Dict[int, Union[int, str, bytes]]] = None,
        timeout: int = 0,
        native_uuid: Optional[bool] = None,
        token_provider: Optional["TokenProvider"] = None,
        **kwargs: Any,
    ) -> None:
        """
        Initialize the connection object with the specified connection string and parameters.

        Args:
            connection_str (str): The connection string to connect to.
            autocommit (bool): If True, causes a commit to be performed after
                              each SQL statement.
            attrs_before (dict, optional): Dictionary of connection attributes to set before
                                          connection establishment. Keys are SQL_ATTR_* constants,
                                          and values are their corresponding settings.
                                          Use this for attributes that must be set before
                                          connecting, such as SQL_ATTR_LOGIN_TIMEOUT,
                                          SQL_ATTR_ODBC_CURSORS, and SQL_ATTR_PACKET_SIZE.
            timeout (int): Login (connection-attempt) timeout in seconds. 0 (default)
                means the driver default is used. This sets SQL_ATTR_LOGIN_TIMEOUT
                before connecting and bounds the login/network connection attempt.
                (Microsoft Entra ID token acquisition happens before the ODBC
                connect and is not covered by this attribute.) It is distinct from
                the ``Connection.timeout`` property, which is the per-statement
                query timeout. An explicit ``attrs_before[SQL_ATTR_LOGIN_TIMEOUT]``
                takes precedence over this value. (pyodbc migration note: pyodbc
                lets a positive ``timeout=`` override ``attrs_before``; here the
                explicit ``attrs_before`` entry wins.)
            native_uuid (bool, optional): Controls whether UNIQUEIDENTIFIER columns return
                uuid.UUID objects (True) or str (False) for cursors created from this connection.
                None (default) defers to the module-level ``mssql_python.native_uuid`` setting (True).
            token_provider (object, optional): Advanced token provider for Microsoft Entra ID
                authentication. Must expose a callable ``.get_token(scope)`` method that returns
                an object with a ``.token`` attribute.

                This parameter is mutually exclusive with ``Authentication=`` in the connection
                string and with ``attrs_before[SQL_COPT_SS_ACCESS_TOKEN]``; supplying more than
                one token source raises ``InterfaceError`` at connect time.

                If ``UID``/``PWD``/``Trusted_Connection`` are also present in the connection
                string they are ignored (access-token auth wins) and a warning is emitted.

                .. note::
                    The token scope is fixed to the Azure **commercial** cloud
                    (``https://database.windows.net/.default``). Sovereign clouds (Azure US
                    Government, Azure China, Azure Germany) are **out of scope** for this
                    parameter — a token acquired for a different audience is rejected by SQL
                    Server at login. For sovereign clouds, acquire the token yourself and pass
                    it via ``attrs_before[SQL_COPT_SS_ACCESS_TOKEN]`` instead.

                .. note::
                    Connection pooling is enabled for access-token connections
                    (``token_provider=``, built-in ``Authentication=ActiveDirectory*``, or a raw
                    ``attrs_before[SQL_COPT_SS_ACCESS_TOKEN]``). The native pool key is
                    identity-aware — the sanitized connection string plus a per-identity suffix
                    (``msi:``/``acct:``/``tok:``) — so different principals sharing the same
                    server/database land in distinct pool buckets and are never handed each
                    other's authenticated connection, while same-identity reuse still benefits
                    from pooling.

                .. note::
                    Token lifecycle limitations: the access token is a *pre-connect* ODBC
                    attribute, so it cannot be refreshed on a live connection. Long-lived
                    connections must be recycled by the application once the token nears expiry,
                    and Continuous Access Evaluation (CAE) claims challenges are not handled.
                    These require native driver support and are tracked as follow-up work.
                    Interactive credentials (e.g. ``InteractiveBrowserCredential``) block
                    ``connect()`` until the user completes sign-in; prefer non-interactive
                    credentials in server contexts.
            **kwargs: Additional key/value pairs for the connection string.

        Returns:
            None

        Raises:
            InterfaceError: If ``token_provider`` is misused (combined with another token
                source, or lacking a valid ``.get_token`` method), or the credential returns
                no valid token.
            OperationalError: If acquiring a token from ``token_provider`` fails.
            ValueError: If the connection string is invalid or connection fails.

        This method sets up the initial state for the connection object,
        preparing it for further operations such as connecting to the
        database, executing queries, etc.

        Example:
            >>> # Setting login timeout using attrs_before
            >>> import mssql_python as ms
            >>> conn = ms.connect("Server=myserver;Database=mydb",
            ...                   attrs_before={ms.SQL_ATTR_LOGIN_TIMEOUT: 30})

            >>> # Return native uuid.UUID objects instead of strings
            >>> conn = ms.connect("Server=myserver;Database=mydb", native_uuid=True)
        """
        # Store per-connection native_uuid override.
        # None means "use module-level mssql_python.native_uuid".
        if native_uuid is not None and not isinstance(native_uuid, bool):
            raise ValueError("native_uuid must be a boolean value or None")
        self._native_uuid = native_uuid

        self.connection_str, parsed_params = self._construct_connection_string(
            connection_str, **kwargs
        )
        # Shallow-copy so we never mutate the caller's dict (e.g. when the
        # token_provider path injects SQL_COPT_SS_ACCESS_TOKEN). Mutating the
        # caller's object would leak the access token into user state and break
        # re-using the same attrs_before dict across multiple connections.
        self._attrs_before = dict(attrs_before) if attrs_before else {}

        # Validate and apply the LOGIN/connection-attempt timeout up front —
        # before any Entra/token acquisition below — so invalid input (negative,
        # non-int, bool) fails fast without triggering a network or interactive
        # token fetch. ``timeout`` matches pyodbc's login timeout: it sets
        # SQL_ATTR_LOGIN_TIMEOUT and bounds the login/network connection attempt.
        # It is distinct from Connection.timeout (the per-statement QUERY timeout,
        # tracked by self._timeout and applied to cursors), which defaults to
        # 0 = disabled. Only inject the login timeout when the caller asked for
        # one (> 0) and did not already set it explicitly via attrs_before (an
        # explicit attrs_before value wins).
        self._timeout = 0
        login_timeout = self._validate_timeout(timeout, kind="Login timeout")
        if login_timeout > 0:
            self._attrs_before.setdefault(ConstantsDDBC.SQL_ATTR_LOGIN_TIMEOUT.value, login_timeout)

        # Initialize encoding settings with defaults for Python 3
        # Python 3 only has str (which is Unicode), so we use utf-16le by default
        self._encoding_settings = {
            "encoding": "utf-16le",
            "ctype": ConstantsDDBC.SQL_WCHAR.value,
        }

        # Initialize decoding settings with Python 3 defaults
        # SQL_CHAR default uses SQL_WCHAR ctype so the ODBC driver returns
        # UTF-16 data for VARCHAR columns. This avoids encoding mismatches on
        # Windows where the driver returns raw bytes in the server's native
        # code page (e.g. CP-1252) that may fail to decode as UTF-8.
        self._decoding_settings = {
            ConstantsDDBC.SQL_CHAR.value: {
                "encoding": "utf-16le",
                "ctype": ConstantsDDBC.SQL_WCHAR.value,
            },
            ConstantsDDBC.SQL_WCHAR.value: {
                "encoding": "utf-16le",
                "ctype": ConstantsDDBC.SQL_WCHAR.value,
            },
            SQL_WMETADATA: {
                "encoding": "utf-16le",
                "ctype": ConstantsDDBC.SQL_WCHAR.value,
            },
        }
        self._decoding_generation = 0

        # Auth type for acquiring fresh tokens at bulk copy time.
        # We intentionally do NOT cache the token — a fresh one is acquired
        # each time bulkcopy() is called to avoid expired-token errors.
        self._auth_type = None
        # Credential constructor kwargs (e.g. user-assigned MSI client_id)
        # captured at __init__ time before remove_sensitive_params strips UID
        # from self.connection_str. bulkcopy() re-uses these when acquiring a
        # fresh token; re-parsing self.connection_str at that point would miss
        # them because UID is already gone.
        self._credential_kwargs: Optional[Dict[str, str]] = None
        # User-supplied token provider for custom Entra ID authentication.
        # Stored so bulk copy can call .get_token() for a fresh JWT later.
        self._token_provider: Optional["TokenProvider"] = None
        # POSIX timestamp (seconds) at which the current access token expires,
        # captured from the credential's AccessToken result. None when unknown.
        # The token is a pre-connect ODBC attribute and cannot be refreshed on
        # a live connection — this is exposed for diagnostics/logging only.
        # A custom token_provider may report a float POSIX timestamp, so the
        # hint is Optional[float] (int is accepted under the numeric tower).
        self._token_expires_on: Optional[float] = None

        # Composite, identity-aware pool key. Empty means "key the native pool
        # on the connection string" (legacy behavior, used for non-token auth).
        # For Entra access-token auth it is set to connStr + identity so that
        # distinct identities never share a pooled connection.
        self._pool_key: str = ""

        # Optional deferred-attrs callback handed to the native layer. When set,
        # native invokes it *only* when it actually opens a physical connection
        # (a pool miss or a non-pooled connect), so a same-identity pool hit
        # skips token acquisition entirely. None means "no deferred
        # token" — the token (if any) is already in self._attrs_before.
        #
        # NOTE: This is an internal, private callback and is intentionally NOT
        # the public ``token_provider=`` credential parameter. It returns the
        # full connect-attrs dict lazily; hence the distinct name
        # ``_token_factory``.
        self._token_factory = None

        # Custom token_provider= parameter — takes priority, mutually exclusive
        # with Authentication= in the connection string.
        if token_provider is not None:
            self._configure_token_provider(token_provider, parsed_params)

        # Handle Entra ID authentication if specified.
        # The parsed dict is used directly — no re-parsing of the connection string.
        elif _KEY_AUTHENTICATION in parsed_params:
            auth_type = process_auth_parameters(parsed_params)

            if auth_type:
                # Capture credential kwargs (e.g. user-assigned MSI client_id)
                # from the parsed dict *before* remove_sensitive_params strips UID.
                credential_kwargs: Optional[Dict[str, str]] = None
                if auth_type == _AuthInternal.MSI:
                    uid = (parsed_params.get(_KEY_UID) or "").strip()
                    if uid:
                        credential_kwargs = {"client_id": uid}

                # Strip sensitive params and rebuild the connection string.
                sanitized = remove_sensitive_params(parsed_params)
                self.connection_str = _ConnectionStringBuilder(sanitized).build()
                self._credential_kwargs = credential_kwargs

                # Make the pool key identity-aware so two callers using the
                # same server but different Entra identities never reuse each
                # other's authenticated connection. auth_type here
                # is the process_auth_parameters result, which is truthy only
                # for the token-bearing types (default/devicecode/msi/
                # interactive-non-Windows); ServicePrincipal and Windows
                # Interactive return None and keep their identity in the
                # connection string, so they stay on the legacy connStr key.
                #
                # First try to derive the identity *without* a token. For MSI
                # the client/object id comes straight from the params, so the
                # pool key is known before any token exists — which lets us
                # defer token acquisition to a provider callback that native
                # invokes only on a pool miss.
                #
                # The factory returns ``(attrs, expires_on)``: the connect-attrs
                # dict plus the token's POSIX-epoch expiry (or None). Native
                # stores the expiry so it can refresh/discard a pooled
                # connection whose token is near expiry on checkout.
                token_attr = ConstantsDDBC.SQL_COPT_SS_ACCESS_TOKEN.value
                base_attrs = self._attrs_before

                def _acquire_token_info():
                    # DB-API boundary: get_auth_token_info fails closed by
                    # letting the underlying Azure error propagate as a
                    # ValueError (unsupported auth type) or RuntimeError
                    # (credential/network/auth failure). Re-wrap those as a
                    # DB-API 2.0 InterfaceError so every auth failure surfaced
                    # from connect() / the token factory is a consistent,
                    # catchable driver exception rather than a bare RuntimeError.
                    try:
                        return get_auth_token_info(auth_type, credential_kwargs)
                    except (RuntimeError, ValueError) as e:
                        raise InterfaceError(
                            driver_error=(
                                "Failed to acquire an Entra ID access token for "
                                f"authentication type '{auth_type}': {e}"
                            ),
                            ddbc_error=str(e),
                        ) from e

                def _make_token_factory(expected_account: Optional[str] = None):
                    # Build the deferred connect-attrs provider handed to native.
                    # Native invokes the returned callable only when it actually
                    # opens a physical connection (a pool miss, non-pooled
                    # connect, or near-expiry refresh on checkout), so a
                    # same-identity pool hit never pays for a token.
                    #
                    # ``expected_account`` binds an account-keyed (``acct:``)
                    # pool to the exact account it was keyed on: the shared
                    # interactive/device-code credential can silently switch
                    # signed-in accounts between pooling and a later refresh, and
                    # opening a *different* principal's connection inside this
                    # pool would hand a caller a connection authenticated as the
                    # wrong account. MSI pools pass ``None`` (their identity is
                    # fixed by params, not by a mutable signed-in account).
                    def _token_factory():
                        attrs = dict(base_attrs)
                        info = _acquire_token_info()
                        if info and info.token_struct:
                            if (
                                expected_account is not None
                                and info.home_account_id != expected_account
                            ):
                                # Fail closed: the signed-in account changed out
                                # from under this account-keyed pool. Refuse
                                # rather than authenticate as the new principal.
                                raise InterfaceError(
                                    driver_error=(
                                        "The signed-in account changed between "
                                        "pooling and connect for authentication "
                                        f"type '{auth_type}'; refusing to open a "
                                        "connection for a different account in "
                                        "this pool."
                                    ),
                                    ddbc_error="Account mismatch in token factory.",
                                )
                            attrs[token_attr] = info.token_struct
                            return attrs, info.expires_on
                        # Fail closed: a token-backed pool must never open a
                        # physical connection without the token it is meant to
                        # carry. get_auth_token_info already raises when
                        # acquisition fails; this guards the empty-token edge so
                        # native never connects with Authentication= stripped.
                        raise InterfaceError(
                            driver_error=(
                                "Unable to acquire an Entra ID access token for "
                                f"authentication type '{auth_type}'."
                            ),
                            ddbc_error="Token factory produced no usable token.",
                        )

                    return _token_factory

                identity = compute_identity_key(auth_type, credential_kwargs)
                if identity:
                    # A real connection string can
                    # never contain \0, so the composite key can never collide
                    # with a bare connStr pool key. std::u16string map keys hold
                    # embedded NULs fine and the key is never used as a C string.
                    self._pool_key = self.connection_str + "\x00" + identity

                    # Lazy token acquisition: native invokes this only
                    # when it opens a physical connection, so same-identity pool
                    # hits never pay for a token. MSI identity is param-derived,
                    # so there is no signed-in account to bind.
                    self._token_factory = _make_token_factory()
                else:
                    # Token/account-dependent identity: acquire once to derive
                    # the key. Interactive / Device-code yield a stable
                    # home_account_id (key ``acct:``) so subsequent acquisitions
                    # can be deferred to the factory (silent refresh reuses the
                    # pool). DefaultAzureCredential / raw token key on the token
                    # hash (``tok:``); the pooled connection is bound to that
                    # exact token, so it is kept in attrs_before with no factory.
                    info = _acquire_token_info()
                    token = info.token_struct if info else None
                    home_account_id = info.home_account_id if info else None
                    identity = compute_identity_key(
                        auth_type,
                        credential_kwargs,
                        token_struct=token,
                        home_account_id=home_account_id,
                    )
                    if identity:
                        self._pool_key = self.connection_str + "\x00" + identity
                    if identity and identity.startswith("acct:"):
                        # Account-stable key: safe to defer to the factory so a
                        # same-account pool hit skips token acquisition and a
                        # near-expiry checkout can refresh silently. Bind the
                        # factory to this exact account so a concurrent sign-in
                        # that flips the shared credential's account can never
                        # open a different principal's connection in this pool.
                        self._token_factory = _make_token_factory(expected_account=home_account_id)
                    elif token:
                        # Token-hash key: bind the pooled connection to this
                        # exact token so its hash always matches the pool key.
                        # No token factory is attached, so this pool is NOT
                        # expiry-aware — the near-expiry refresh in the native
                        # layer only runs for factory-backed (msi:/acct:) pools.
                        # A driver-acquired DAC/raw token is reused until the
                        # connection dies; see the pooling notes in the README.
                        self._attrs_before[token_attr] = token
                    else:
                        # Fail closed: a token-backed auth type reached here but
                        # we could derive neither a deferred factory nor an
                        # eager token, so connecting now would strip
                        # Authentication= and open with no token (potentially
                        # authenticating as an unintended identity). Surface a
                        # clear error instead of silently falling through.
                        raise InterfaceError(
                            driver_error=(
                                "Unable to acquire an Entra ID access token for "
                                f"authentication type '{auth_type}'."
                            ),
                            ddbc_error="Token acquisition returned no usable token.",
                        )

            # Store auth type so bulkcopy() can acquire a fresh token later.
            # On Windows Interactive, process_auth_parameters returns None
            # (DDBC handles auth natively), so fall back to extract_auth_type.
            self._auth_type = auth_type or extract_auth_type(parsed_params)

        self._closed = False

        # Using WeakSet which automatically removes cursors when they are no
        # longer in use
        # It is a set that holds weak references to its elements.
        # When an object is only weakly referenced, it can be garbage
        # collected even if it's still in the set.
        # It prevents memory leaks by ensuring that cursors are cleaned up
        # when no longer in use without requiring explicit deletion.
        # TODO: Think and implement scenarios for multi-threaded access
        # to cursors
        self._cursors = weakref.WeakSet()

        # Initialize output converters dictionary and its lock for thread safety
        self._output_converters = {}
        self._converters_generation = 0
        self._converters_lock = threading.Lock()

        # Initialize encoding/decoding settings lock for thread safety
        # This lock protects both _encoding_settings and _decoding_settings dictionaries
        # from concurrent modification. We use a simple Lock (not RLock) because:
        # - Write operations (setencoding/setdecoding) replace the entire dict atomically
        # - Read operations (getencoding/getdecoding) return a copy, so they're safe
        # - No recursive locking is needed in our usage pattern
        # This is more performant than RLock for the multiple-readers-single-writer pattern
        self._encoding_lock = threading.Lock()

        # Initialize search escape character
        self._searchescape = None

        # Safety net for the raw access-token pattern: a caller may pass
        # SQL_COPT_SS_ACCESS_TOKEN directly in attrs_before with no
        # Authentication= keyword (the documented msodbcsql raw-token pattern),
        # or via the public token_provider= API. That path never runs the
        # identity-key logic above, so without this guard the pool key would
        # stay empty and two callers with different raw tokens against the same
        # server would share a pool — handing user B user A's authenticated
        # connection on a pool hit. Bind the pool key to the token hash so
        # distinct tokens never collide (upholds the "a token is present =>
        # key is never the bare connStr" invariant).
        if not self._pool_key:
            _raw_token = self._attrs_before.get(ConstantsDDBC.SQL_COPT_SS_ACCESS_TOKEN.value)
            if _raw_token is not None and not isinstance(_raw_token, (bytes, bytearray)):
                # Fail closed: an access token supplied as anything other than
                # raw bytes (e.g. a str) cannot be hashed into an identity-aware
                # pool key, so it would fall through with the bare connStr key
                # and two callers passing different str tokens against the same
                # server could share a pooled, authenticated connection. Reject
                # it up front with a clear DB-API error instead of relying on
                # ODBC to mangle/reject the byte-struct downstream. The native
                # setAttribute() enforces the same rule, so the invariant holds
                # at both boundaries.
                raise InterfaceError(
                    driver_error=(
                        "SQL_COPT_SS_ACCESS_TOKEN must be supplied as bytes (the "
                        "raw [length][UTF-16LE token] struct), not "
                        f"{type(_raw_token).__name__}."
                    ),
                    ddbc_error="Non-binary access token attribute rejected.",
                )
            if isinstance(_raw_token, (bytes, bytearray)):
                # Freeze a mutable bytearray token to immutable bytes ONCE and
                # store it back, so the pool-key hash below and the later native
                # connect both read the exact same value. Hashing the bytearray
                # and letting native read it separately would be a TOCTOU: a
                # caller mutating the bytearray in between would bind the pooled
                # connection to a key that no longer matches its token.
                _frozen_token = bytes(_raw_token)
                self._attrs_before[ConstantsDDBC.SQL_COPT_SS_ACCESS_TOKEN.value] = _frozen_token
                # This raw token has no param-derivable identity, so key it on
                # the token hash directly.
                _token_identity = compute_token_identity(_frozen_token)
                if _token_identity:
                    self._pool_key = self.connection_str + "\x00" + _token_identity

        # Auto-enable pooling if user never called
        if not PoolingManager.is_initialized():
            PoolingManager.enable()
        self._pooling = PoolingManager.is_enabled()

        # Resolve and freeze the ODBC provider, then hand the selection to the
        # native loader so it imports the matching provider package. Done here —
        # after every Python-side validation and token acquisition has succeeded
        # and immediately before the native driver loads — so a call that fails
        # earlier never freezes the selection as a side effect; a later, corrected
        # connection can then still choose a different provider without a process
        # restart.
        _provider = ProviderManager.ensure_available()
        ddbc_bindings._set_odbc_provider(_provider)

        try:
            self._conn = ddbc_bindings.Connection(
                self.connection_str,
                self._pooling,
                self._attrs_before,
                self._pool_key,
                self._token_factory,
            )
        except RuntimeError as e:
            _raise_connection_error(e)
        self.setautocommit(autocommit)

        # Register this connection for cleanup before Python shutdown
        # This ensures ODBC handles are freed in correct order, preventing leaks
        try:
            if hasattr(mssql_python, "_register_connection"):
                mssql_python._register_connection(self)
        except AttributeError as e:
            # If registration fails, continue - cleanup will still happen via __del__
            logger.warning(
                f"Failed to register connection for shutdown cleanup: {type(e).__name__}: {e}"
            )
        except Exception as e:
            # Catch any other unexpected errors during registration
            logger.error(
                f"Unexpected error during connection registration: {type(e).__name__}: {e}"
            )

    def _configure_token_provider(
        self, token_provider: "TokenProvider", parsed_params: Dict[str, str]
    ) -> None:
        """Validate a custom ``token_provider`` and wire it for pooled auth.

        Validates that ``token_provider`` is not combined with another token
        source and exposes a ``get_token()`` method, strips sensitive params
        from the connection string, and acquires a token once up front so an
        invalid credential fails fast at connect() and the expiry is captured.
        The acquired token is placed in ``attrs_before``; the pool is keyed on
        the token hash (``tok:``) by the safety net in ``__init__`` so distinct
        tokens never share a pooled connection. Mutually exclusive with
        ``Authentication=`` and a manual ``attrs_before`` access token.

        Raises:
            InterfaceError: If ``token_provider`` is combined with another token
                source, or lacks a ``get_token(scope)`` method.
            OperationalError: If acquiring a token from ``token_provider`` fails.
        """
        if _KEY_AUTHENTICATION in parsed_params:
            raise InterfaceError(
                driver_error=(
                    "Cannot specify both 'token_provider' parameter and "
                    "'Authentication' in the connection string. "
                    "Use one or the other."
                ),
                ddbc_error="",
            )
        if ConstantsDDBC.SQL_COPT_SS_ACCESS_TOKEN.value in self._attrs_before:
            raise InterfaceError(
                driver_error=(
                    "Cannot specify both 'token_provider' parameter and "
                    "attrs_before[SQL_COPT_SS_ACCESS_TOKEN]. "
                    "Use one token source."
                ),
                ddbc_error="",
            )
        get_token = getattr(token_provider, "get_token", None)
        if not callable(get_token):
            raise InterfaceError(
                driver_error=(
                    f"token_provider must have a .get_token() method. "
                    f"Got {type(token_provider).__name__}."
                ),
                ddbc_error="",
            )
        # The get_token() signature is NOT inspected here: inspect.signature()
        # is unreliable for partial/decorated/C-extension callables and would
        # produce false warnings on valid credentials. The actual call is the
        # source of truth — _get_token_from_credential turns a bad signature
        # (TypeError) into a clear InterfaceError.
        from mssql_python.auth import acquire_token_from_credential, _user_facing_stacklevel

        # access-token auth ignores UID/PWD/Trusted_Connection — warn so the
        # user is not surprised that those credentials are silently dropped.
        dropped = [
            key for key in (_KEY_UID, _KEY_PWD, _KEY_TRUSTED_CONNECTION) if key in parsed_params
        ]
        if dropped:
            warnings.warn(
                "token_provider is set, so the following connection-string "
                f"credential(s) are ignored: {', '.join(sorted(dropped))}. "
                "Remove them to silence this warning.",
                UserWarning,
                # Point the warning at the caller's own code rather than this
                # internal helper, regardless of call depth (connect() vs a
                # direct Connection()).
                stacklevel=_user_facing_stacklevel(),
            )
        self._token_provider = token_provider

        # Strip sensitive params (UID/PWD/Trusted_Connection) since access-token
        # auth is used — same as the Authentication= path. Do this BEFORE
        # building the pool key so the key is derived from the exact connection
        # string native will connect with.
        sanitized = remove_sensitive_params(parsed_params)
        self.connection_str = _ConnectionStringBuilder(sanitized).build()

        # Acquire the token once, up front. This validates the credential so an
        # invalid one fails fast at connect() (raising InterfaceError/
        # OperationalError here), captures the expiry for diagnostics, and fires
        # the already-expired-token warning. The token is placed directly in
        # attrs_before; the identity-aware safety net in __init__ then keys the
        # pool on the token hash (``tok:``).
        #
        # NOTE: a custom token_provider is keyed on the token it mints, NOT on
        # the provider object. Object-identity keying would be unsafe: a mutable
        # credential (e.g. AzureCliCredential, which follows whoever is logged
        # into the az CLI) can represent different principals over its lifetime,
        # yet a same-object pool hit skips re-acquisition — so a caller could be
        # handed a connection authenticated as a stale principal. Token-hash
        # keying re-derives identity from the actual token on every physical
        # connect, so a principal change always lands in a distinct pool. The
        # trade-off is weaker reuse (a rotated token opens a new bucket, which
        # the native idle sweep later evicts) and no expiry-aware refresh — the
        # pooled connection is reused until it dies. See the pooling notes in
        # the README.
        token, token_expires_on = acquire_token_from_credential(token_provider)
        self._token_expires_on = token_expires_on
        self._attrs_before[ConstantsDDBC.SQL_COPT_SS_ACCESS_TOKEN.value] = token

    def _construct_connection_string(
        self, connection_str: str = "", **kwargs: Any
    ) -> Tuple[str, Dict[str, str]]:
        """
        Construct the connection string by parsing, validating, and merging parameters.

        1. Parse and validate the base connection_str (validates against allowlist)
        2. Normalize parameter names (e.g., addr/address -> Server, uid -> UID)
        3. Merge kwargs (which override connection_str params after normalization)
        4. Add Driver and APP (always controlled by the driver)
        5. Build and return the final connection string + parameter dictionary

        Args:
            connection_str (str): The base connection string.
            **kwargs: Additional key/value pairs for the connection string.

        Returns:
            Tuple[str, Dict[str, str]]: The constructed connection string and
                the normalized parameter dictionary.
        """

        # Reject embedded NUL (\x00) up front, for both the base string and every
        # kwargs value. The ODBC layer terminates the connection string at the
        # first NUL (SQL_NTS), so anything after it is silently dropped; worse,
        # the identity-aware pool key joins the connection string and the
        # per-identity discriminator with a NUL separator, so a NUL smuggled into
        # a value could forge or collide pool keys. Fail closed at this Python
        # boundary rather than relying on downstream truncation.
        if isinstance(connection_str, str) and "\x00" in connection_str:
            raise InterfaceError(
                driver_error="Connection string must not contain a NUL (\\x00) character.",
                ddbc_error="Embedded NUL in connection string.",
            )
        for _key, _value in kwargs.items():
            if isinstance(_value, str) and "\x00" in _value:
                raise InterfaceError(
                    driver_error=(
                        f"Connection parameter '{_key}' must not contain a NUL "
                        "(\\x00) character."
                    ),
                    ddbc_error="Embedded NUL in connection parameter.",
                )

        # Step 1: Parse base connection string with allowlist validation
        # The parser validates everything: unknown params, reserved params, duplicates, syntax
        parser = _ConnectionStringParser(validate_keywords=True)
        parsed_params = parser._parse(connection_str)

        # Step 2: Normalize parameter names (e.g., addr/address -> Server, uid -> UID)
        # This handles synonym mapping and deduplication via normalized keys
        normalized_params = _ConnectionStringParser._normalize_params(
            parsed_params, warn_rejected=False
        )

        # Step 3: Process kwargs and merge with normalized_params
        # kwargs override connection string values (processed after, so they take precedence)
        for key, value in kwargs.items():
            normalized_key = _ConnectionStringParser.normalize_key(key)
            if normalized_key:
                # Driver and APP are reserved - raise error if user tries to set them
                if normalized_key in _RESERVED_PARAMETERS:
                    raise ValueError(
                        f"Connection parameter '{key}' is reserved and controlled by the driver. "
                        f"It cannot be set by the user."
                    )
                # kwargs override any existing values from connection string
                normalized_params[normalized_key] = str(value)
            else:
                logger.warning(f"Ignoring unknown connection parameter from kwargs: {key}")

        # Step 4: Add Driver and APP (always controlled by the driver).
        normalized_params["Driver"] = "ODBC Driver 18 for SQL Server"
        normalized_params["APP"] = "MSSQL-Python"

        # Step 5: Build final connection string
        conn_str = _ConnectionStringBuilder(normalized_params).build()

        logger.info("Final connection string: %s", sanitize_connection_string(conn_str))

        return conn_str, normalized_params

    @staticmethod
    def _validate_timeout(value: int, kind: str = "Timeout") -> int:
        """
        Validate a timeout value (login or query) and normalize it to ``int``.

        Shared by ``__init__`` (login timeout) and the ``timeout`` setter (query
        timeout) so both entry points reject invalid input identically and can
        never drift apart. ``kind`` only customizes the error-message wording
        ("Login timeout" vs "Query timeout"); the validation rules are identical.

        Args:
            value (int): Timeout in seconds. Must be a non-negative integer.
            kind (str): Human-readable label used in error messages.

        Returns:
            int: The validated timeout value.

        Raises:
            TypeError: If ``value`` is not an integer (``bool`` is rejected too).
            ValueError: If ``value`` is negative.
        """
        # ``bool`` is a subclass of ``int``; reject it explicitly so a stray
        # ``timeout=True`` cannot slip through as ``1`` (a classic int/bool footgun).
        if isinstance(value, bool) or not isinstance(value, int):
            raise TypeError(f"{kind} must be an integer")
        if value < 0:
            raise ValueError(f"{kind} cannot be negative")
        return value

    @property
    def timeout(self) -> int:
        """
        Get the current query timeout setting in seconds.

        Returns:
            int: The timeout value in seconds. Zero means no timeout (wait indefinitely).
        """
        return self._timeout

    @timeout.setter
    def timeout(self, value: int) -> None:
        """
        Set the query timeout for all operations performed by this connection.

        Args:
            value (int): The timeout value in seconds. Zero means no timeout.

        Returns:
            None

        Note:
            This timeout applies to all cursors created from this connection.
            It cannot be changed for individual cursors or SQL statements.
            If a query timeout occurs, an OperationalError exception will be raised.
        """
        self._timeout = self._validate_timeout(value, kind="Query timeout")
        logger.info(f"Query timeout set to {value} seconds")

    @property
    def autocommit(self) -> bool:
        """
        Return the current autocommit mode of the connection.
        Returns:
            bool: True if autocommit is enabled, False otherwise.
        """
        try:
            return self._conn.get_autocommit()
        except RuntimeError as e:
            _raise_connection_error(e)

    @autocommit.setter
    def autocommit(self, value: bool) -> None:
        """
        Set the autocommit mode of the connection.
        Args:
            value (bool): True to enable autocommit, False to disable it.
        Returns:
            None
        """
        self.setautocommit(value)
        logger.info("Autocommit mode set to %s.", value)

    @property
    def closed(self) -> bool:
        """
        Returns True if the connection is closed, False otherwise.

        This property indicates whether close() was explicitly called on
        the connection. Note that this does not indicate whether the
        connection is healthy/alive - if a timeout or network issue breaks
        the connection, closed would still be False until close() is
        explicitly called.

        Returns:
            bool: True if the connection is closed, False otherwise.
        """
        return self._closed

    def setautocommit(self, value: bool = False) -> None:
        """
        Set the autocommit mode of the connection.
        Args:
            value (bool): True to enable autocommit, False to disable it.
        Returns:
            None
        Raises:
            DatabaseError: If there is an error while setting the autocommit mode.
        """
        try:
            self._conn.set_autocommit(value)
        except RuntimeError as e:
            _raise_connection_error(e)

    def setencoding(self, encoding: Optional[str] = None, ctype: Optional[int] = None) -> None:
        """
        Records the requested text encoding settings for compatibility.

        SQL statements and str parameters are always sent as UTF-16LE; text
        parameters are bound as SQL_C_WCHAR on every platform. This applies to
        execute() and executemany(), with or without setinputsizes(). This method
        does not change that behavior or enforce the requested codec.

        Requests that pass validation but differ from UTF-16LE with SQL_WCHAR emit
        UserWarning. Invalid codec names, invalid ctypes, and incompatible
        combinations (such as UTF-8 with SQL_WCHAR) raise ProgrammingError before
        any warning is emitted or settings are stored. Accepted settings are still
        returned by getencoding(), not the effective binding. Use setdecoding()
        separately to configure how results are read.

        Args:
            encoding (str, optional): The requested encoding. This must be a valid Python
                encoding that converts text to bytes. If None, defaults to 'utf-16le'.
            ctype (int, optional): The requested C data type:
                SQL_CHAR or SQL_WCHAR. If not provided, SQL_WCHAR is used for
                UTF-16 variants (see UTF16_ENCODINGS constant). SQL_CHAR is used
                for all other encodings.

        Returns:
            None

        Raises:
            ProgrammingError: If the encoding or ctype is invalid, or their
                combination is incompatible.
            InterfaceError: If the connection is closed.

        Warns:
            UserWarning: If the request passes validation but its encoding or
                ctype cannot be honored.

        Example:
            # Restore the supported default.
            cnxn.setencoding()

            # Warns: parameters still use UTF-16LE / SQL_C_WCHAR.
            cnxn.setencoding(encoding='cp1252', ctype=mssql_python.SQL_CHAR)
        """
        logger.debug(
            "setencoding: Configuring encoding=%s, ctype=%s",
            str(encoding) if encoding else "default",
            str(ctype) if ctype else "auto",
        )
        if self._closed:
            logger.debug("setencoding: Connection is closed")
            raise InterfaceError(
                driver_error="Connection is closed",
                ddbc_error="Connection is closed",
            )

        # Set default encoding if not provided
        if encoding is None:
            encoding = "utf-16le"
            logger.debug("setencoding: Using default encoding=utf-16le")

        # Validate encoding using cached validation for better performance
        if not _validate_encoding(encoding):
            # Log the sanitized encoding for security
            logger.warning(
                "Invalid encoding attempted: %s",
                sanitize_user_input(str(encoding)),
            )
            raise ProgrammingError(
                driver_error=f"Unsupported encoding: {encoding}",
                ddbc_error=f"The encoding '{encoding}' is not supported by Python",
            )

        # Normalize encoding to casefold for more robust Unicode handling
        encoding = encoding.casefold()
        logger.debug("setencoding: Encoding normalized to %s", encoding)

        # Early validation if ctype is already specified as SQL_WCHAR
        if ctype == ConstantsDDBC.SQL_WCHAR.value:
            _validate_utf16_wchar_compatibility(encoding, ctype, "SQL_WCHAR")

        # Set default ctype based on encoding if not provided
        if ctype is None:
            if encoding in UTF16_ENCODINGS:
                ctype = ConstantsDDBC.SQL_WCHAR.value
                logger.debug("setencoding: Auto-selected SQL_WCHAR for UTF-16")
            else:
                ctype = ConstantsDDBC.SQL_CHAR.value
                logger.debug("setencoding: Auto-selected SQL_CHAR for non-UTF-16")

        # Validate ctype
        valid_ctypes = [ConstantsDDBC.SQL_CHAR.value, ConstantsDDBC.SQL_WCHAR.value]
        if ctype not in valid_ctypes:
            # Log the sanitized ctype for security
            logger.warning(
                "Invalid ctype attempted: %s",
                sanitize_user_input(str(ctype)),
            )
            raise ProgrammingError(
                driver_error=f"Invalid ctype: {ctype}",
                ddbc_error=(
                    f"ctype must be SQL_CHAR ({ConstantsDDBC.SQL_CHAR.value}) or "
                    f"SQL_WCHAR ({ConstantsDDBC.SQL_WCHAR.value})"
                ),
            )

        # Final validation: SQL_WCHAR ctype only supports UTF-16 encodings (without BOM)
        if ctype == ConstantsDDBC.SQL_WCHAR.value:
            _validate_utf16_wchar_compatibility(encoding, ctype, "SQL_WCHAR")

        if encoding != "utf-16le" or ctype != ConstantsDDBC.SQL_WCHAR.value:
            warnings.warn(
                "setencoding() does not change SQL statement encoding or text parameter binding: "
                "statements and str parameters always use UTF-16LE, and text parameters are "
                "bound as SQL_C_WCHAR. The requested settings are retained by getencoding() "
                "for compatibility but are not applied. Use setencoding() with no arguments "
                "to restore the supported defaults.",
                UserWarning,
                stacklevel=2,
            )

        # Store the encoding settings (thread-safe with lock)
        with self._encoding_lock:
            self._encoding_settings = {"encoding": encoding, "ctype": ctype}

        # Log with sanitized values for security
        logger.info(
            "Requested text encoding stored as %s with ctype %s",
            sanitize_user_input(encoding),
            sanitize_user_input(str(ctype)),
        )

    def getencoding(self) -> Dict[str, Union[str, int]]:
        """
        Gets the requested text encoding settings (thread-safe).

        These settings are retained for compatibility. They do not describe the
        effective binding: SQL statements and str parameters always use UTF-16LE,
        and text parameters are bound as SQL_C_WCHAR.

        Returns:
            dict: A dictionary containing 'encoding' and 'ctype' keys.

        Raises:
            InterfaceError: If the connection is closed.

        Example:
            settings = cnxn.getencoding()
            print(f"Current encoding: {settings['encoding']}")
            print(f"Current ctype: {settings['ctype']}")

        Note:
            This method is thread-safe and can be called from multiple threads concurrently.
            Returns a copy of the settings to prevent external modification.
        """
        if self._closed:
            raise InterfaceError(
                driver_error="Connection is closed",
                ddbc_error="Connection is closed",
            )

        # Thread-safe read with lock to prevent race conditions
        with self._encoding_lock:
            return self._encoding_settings.copy()

    def setdecoding(
        self, sqltype: int, encoding: Optional[str] = None, ctype: Optional[int] = None
    ) -> None:
        """
        Sets the text decoding used when reading SQL_CHAR and SQL_WCHAR from the database.

        Existing cursors refresh their cached SQL_CHAR/SQL_WCHAR decoding settings
        before their next fetch.

        This method configures how text data is decoded when reading from the database.
        In Python 3, all text is Unicode (str), so this primarily affects the encoding
        used to decode bytes from the database.

        Args:
            sqltype (int): The SQL type being configured: SQL_CHAR, SQL_WCHAR, or SQL_WMETADATA.
                SQL_WMETADATA is a special flag for configuring column name decoding.
            encoding (str, optional): The Python encoding to use when decoding the data.
                If None, defaults to ``'utf-16le'`` for all sqltypes (SQL_CHAR,
                SQL_WCHAR, and SQL_WMETADATA), matching the connection-level
                defaults set in ``Connection.__init__``. Passing ``encoding=None``
                therefore resets the sqltype to its initial default.
            ctype (int, optional): The C data type to request from SQLGetData:
                SQL_CHAR or SQL_WCHAR. If None, uses default based on encoding
                (SQL_WCHAR for UTF-16 variants, SQL_CHAR otherwise).

        Returns:
            None

        Raises:
            ProgrammingError: If the sqltype, encoding, or ctype is invalid.
            InterfaceError: If the connection is closed.

        Example:
            # Reset SQL_CHAR to the connection default (utf-16le + SQL_WCHAR ctype)
            cnxn.setdecoding(mssql_python.SQL_CHAR)

            # Configure SQL_CHAR to use UTF-8 decoding (opt-in, non-default)
            cnxn.setdecoding(mssql_python.SQL_CHAR, encoding='utf-8')

            # Configure column metadata decoding
            cnxn.setdecoding(mssql_python.SQL_WMETADATA, encoding='utf-16le')

            # Use explicit ctype
            cnxn.setdecoding(mssql_python.SQL_WCHAR, encoding='utf-16le',
                           ctype=mssql_python.SQL_WCHAR)
        """
        if self._closed:
            raise InterfaceError(
                driver_error="Connection is closed",
                ddbc_error="Connection is closed",
            )

        # Validate sqltype
        valid_sqltypes = [
            ConstantsDDBC.SQL_CHAR.value,
            ConstantsDDBC.SQL_WCHAR.value,
            SQL_WMETADATA,
        ]
        if sqltype not in valid_sqltypes:
            logger.warning(
                "Invalid sqltype attempted: %s",
                sanitize_user_input(str(sqltype)),
            )
            raise ProgrammingError(
                driver_error=f"Invalid sqltype: {sqltype}",
                ddbc_error=(
                    f"sqltype must be SQL_CHAR ({ConstantsDDBC.SQL_CHAR.value}), "
                    f"SQL_WCHAR ({ConstantsDDBC.SQL_WCHAR.value}), or "
                    f"SQL_WMETADATA ({SQL_WMETADATA})"
                ),
            )

        # Set default encoding based on sqltype if not provided.
        # All sqltypes default to UTF-16LE to match Connection.__init__ defaults.
        # SQL_CHAR uses utf-16le + SQL_WCHAR ctype so the ODBC driver returns
        # UTF-16 data for VARCHAR columns, avoiding encoding mismatches on
        # Windows where the driver may otherwise return raw bytes in the
        # server's native code page (e.g. CP-1252). This makes
        # ``setdecoding(SQL_CHAR)`` with no arguments a true reset-to-defaults.
        if encoding is None:
            encoding = "utf-16le"

        # Validate encoding using cached validation for better performance
        if not _validate_encoding(encoding):
            logger.warning(
                "Invalid encoding attempted: %s",
                sanitize_user_input(str(encoding)),
            )
            raise ProgrammingError(
                driver_error=f"Unsupported encoding: {encoding}",
                ddbc_error=f"The encoding '{encoding}' is not supported by Python",
            )

        # Normalize encoding to lowercase for consistency
        encoding = encoding.lower()

        # Validate SQL_WCHAR encoding compatibility
        if sqltype == ConstantsDDBC.SQL_WCHAR.value:
            _validate_utf16_wchar_compatibility(encoding, sqltype, "SQL_WCHAR sqltype")

        # SQL_WMETADATA can use any valid encoding (UTF-8, UTF-16, etc.)
        # No restriction needed here - let users configure as needed

        # Set default ctype based on encoding if not provided
        if ctype is None:
            if encoding in UTF16_ENCODINGS:
                ctype = ConstantsDDBC.SQL_WCHAR.value
            else:
                ctype = ConstantsDDBC.SQL_CHAR.value

        # Validate ctype
        valid_ctypes = [ConstantsDDBC.SQL_CHAR.value, ConstantsDDBC.SQL_WCHAR.value]
        if ctype not in valid_ctypes:
            logger.warning(
                "Invalid ctype attempted: %s",
                sanitize_user_input(str(ctype)),
            )
            raise ProgrammingError(
                driver_error=f"Invalid ctype: {ctype}",
                ddbc_error=(
                    f"ctype must be SQL_CHAR ({ConstantsDDBC.SQL_CHAR.value}) or "
                    f"SQL_WCHAR ({ConstantsDDBC.SQL_WCHAR.value})"
                ),
            )

        # Validate SQL_WCHAR ctype encoding compatibility
        if ctype == ConstantsDDBC.SQL_WCHAR.value:
            _validate_utf16_wchar_compatibility(encoding, ctype, "SQL_WCHAR ctype")

        # Store the decoding settings for the specified sqltype (thread-safe with lock)
        with self._encoding_lock:
            self._decoding_settings[sqltype] = {"encoding": encoding, "ctype": ctype}
            self._decoding_generation += 1

        # Log with sanitized values for security
        sqltype_name = {
            ConstantsDDBC.SQL_CHAR.value: "SQL_CHAR",
            ConstantsDDBC.SQL_WCHAR.value: "SQL_WCHAR",
            SQL_WMETADATA: "SQL_WMETADATA",
        }.get(sqltype, str(sqltype))

        logger.info(
            "Text decoding set for %s to %s with ctype %s",
            sqltype_name,
            sanitize_user_input(encoding),
            sanitize_user_input(str(ctype)),
        )

    def getdecoding(self, sqltype: int) -> Dict[str, Union[str, int]]:
        """
        Gets the current text decoding settings for the specified SQL type (thread-safe).

        Args:
            sqltype (int): The SQL type to get settings for: SQL_CHAR, SQL_WCHAR, or SQL_WMETADATA.

        Returns:
            dict: A dictionary containing 'encoding' and 'ctype' keys for the specified sqltype.

        Raises:
            ProgrammingError: If the sqltype is invalid.
            InterfaceError: If the connection is closed.

        Example:
            settings = cnxn.getdecoding(mssql_python.SQL_CHAR)
            print(f"SQL_CHAR encoding: {settings['encoding']}")
            print(f"SQL_CHAR ctype: {settings['ctype']}")

        Note:
            This method is thread-safe and can be called from multiple threads concurrently.
            Returns a copy of the settings to prevent external modification.
        """
        if self._closed:
            raise InterfaceError(
                driver_error="Connection is closed",
                ddbc_error="Connection is closed",
            )

        # Validate sqltype
        valid_sqltypes = [
            ConstantsDDBC.SQL_CHAR.value,
            ConstantsDDBC.SQL_WCHAR.value,
            SQL_WMETADATA,
        ]
        if sqltype not in valid_sqltypes:
            raise ProgrammingError(
                driver_error=f"Invalid sqltype: {sqltype}",
                ddbc_error=(
                    f"sqltype must be SQL_CHAR ({ConstantsDDBC.SQL_CHAR.value}), "
                    f"SQL_WCHAR ({ConstantsDDBC.SQL_WCHAR.value}), or "
                    f"SQL_WMETADATA ({SQL_WMETADATA})"
                ),
            )

        # Thread-safe read with lock to prevent race conditions
        with self._encoding_lock:
            return self._decoding_settings[sqltype].copy()

    def set_attr(self, attribute: int, value: Union[int, str, bytes, bytearray]) -> None:
        """
        Set a connection attribute.

        This method sets a connection attribute using SQLSetConnectAttr.
        It provides pyodbc-compatible functionality for configuring connection
        behavior such as autocommit mode, transaction isolation level, and
        connection timeouts.

        Args:
            attribute (int): The connection attribute to set. Should be one of the
                           SQL_ATTR_* constants (e.g., SQL_ATTR_AUTOCOMMIT,
                           SQL_ATTR_TXN_ISOLATION).
            value: The value to set for the attribute. Can be an integer, string,
                   bytes, or bytearray depending on the attribute type.

        Raises:
            InterfaceError: If the connection is closed or attribute is invalid.
            ProgrammingError: If the value type or range is invalid.
            ProgrammingError: If the attribute cannot be set after connection.

        Example:
            >>> conn.set_attr(SQL_ATTR_TXN_ISOLATION, SQL_TXN_READ_COMMITTED)

        Note:
            Some attributes (like SQL_ATTR_LOGIN_TIMEOUT, SQL_ATTR_ODBC_CURSORS, and
            SQL_ATTR_PACKET_SIZE) can only be set before connection establishment and
            must be provided in the attrs_before parameter when creating the connection.
            Attempting to set these attributes after connection will raise a ProgrammingError.
        """
        if self._closed:
            raise InterfaceError(
                "Cannot set attribute on closed connection", "Connection is closed"
            )

        # Use the integrated validation helper function with connection state
        is_valid, error_message, sanitized_attr, sanitized_val = validate_attribute_value(
            attribute, value, is_connected=True
        )

        if not is_valid:
            # Use the already sanitized values for logging
            logger.debug(
                "warning",
                f"Invalid attribute or value: {sanitized_attr}={sanitized_val}, {error_message}",
            )
            raise ProgrammingError(
                driver_error=f"Invalid attribute or value: {error_message}",
                ddbc_error=error_message,
            )

        # Log with sanitized values
        logger.debug(f"Setting connection attribute: {sanitized_attr}={sanitized_val}")

        try:
            # Call the underlying C++ method
            self._conn.set_attr(attribute, value)
            logger.info(f"Connection attribute {sanitized_attr} set successfully")

        except Exception as e:
            error_msg = f"Failed to set connection attribute {sanitized_attr}: {str(e)}"

            # Determine appropriate exception type based on error content
            error_str = str(e).lower()
            if "invalid" in error_str or "unsupported" in error_str or "cast" in error_str:
                logger.error(error_msg)
                raise InterfaceError(error_msg, str(e)) from e
            logger.error(error_msg)
            raise ProgrammingError(error_msg, str(e)) from e

    @property
    def searchescape(self) -> str:
        """
        The ODBC search pattern escape character, as returned by
        SQLGetInfo(SQL_SEARCH_PATTERN_ESCAPE), used to escape special characters
        such as '%' and '_' in LIKE clauses. These are driver specific.

        Returns:
            str: The search pattern escape character (usually '\' or another character)
        """
        if not hasattr(self, "_searchescape") or self._searchescape is None:
            try:
                escape_char = self.getinfo(GetInfoConstants.SQL_SEARCH_PATTERN_ESCAPE.value)
                # Some drivers might return this as an integer memory address
                # or other non-string format, so ensure we have a string
                if not isinstance(escape_char, str):
                    # Default to backslash if not a string
                    escape_char = "\\"
                self._searchescape = escape_char
            except Exception as e:
                # Log the exception for debugging, but do not expose sensitive info
                logger.debug(
                    "warning",
                    "Failed to retrieve search escape character, using default '\\'. "
                    "Exception: %s",
                    type(e).__name__,
                )
                self._searchescape = "\\"
        return self._searchescape

    def cursor(self) -> Cursor:
        """
        Return a new Cursor object using the connection.

        This method creates and returns a new cursor object that can be used to
        execute SQL queries and fetch results. The cursor is associated with the
        current connection and allows interaction with the database.

        Returns:
            Cursor: A new cursor object for executing SQL queries.

        Raises:
            DatabaseError: If there is an error while creating the cursor.
            InterfaceError: If there is an error related to the database interface.
        """
        logger.debug(
            "cursor: Creating new cursor - timeout=%d, total_cursors=%d",
            self._timeout,
            len(self._cursors),
        )
        if self._closed:
            logger.error("cursor: Cannot create cursor on closed connection")
            # raise InterfaceError
            raise InterfaceError(
                driver_error="Cannot create cursor on closed connection",
                ddbc_error="Cannot create cursor on closed connection",
            )

        cursor = Cursor(self, timeout=self._timeout)
        self._cursors.add(cursor)  # Track the cursor
        logger.debug("cursor: Cursor created successfully - total_cursors=%d", len(self._cursors))
        return cursor

    def add_output_converter(self, sqltype: Union[int, type], func: Callable[[Any], Any]) -> None:
        """
        Register an output converter function that will be called whenever a value
        with the given SQL type is read from the database.

        Thread-safe implementation that protects the converters dictionary with a lock.

        Changes apply on the next fetch, including for an already executed result set.

        ⚠️ WARNING: Registering an output converter will cause the supplied Python function
        to be executed on every matching database value. Do not register converters from
        untrusted sources, as this can result in arbitrary code execution and security
        vulnerabilities. This API should never be exposed to untrusted or external input.

        Args:
            sqltype (int or type): The type to convert. Either:
                - an integer ODBC SQL type code (pyodbc-compatible), which can be one of
                  the standard constants (e.g. SQL_VARCHAR, SQL_DECIMAL) or a
                  database-specific value (e.g. -151 for the SQL Server geometry type). The
                  converter fires for columns whose ODBC SQL type matches exactly, so
                  distinct types such as DECIMAL and NUMERIC can have separate converters; or
                - a Python type (e.g. ``decimal.Decimal``, ``str``, ``bytes``), which fires
                  for every column whose value materializes to that Python type (so, for
                  example, a single ``decimal.Decimal`` converter matches DECIMAL, NUMERIC,
                  MONEY and SMALLMONEY columns alike). The supported Python types are
                  ``str``, ``bytes``, ``bool``, ``int``, ``float``, ``decimal.Decimal``,
                  ``datetime.date``, ``datetime.time``, ``datetime.datetime`` and
                  ``uuid.UUID``.
                When both an integer-keyed and a Python-type-keyed converter could apply to
                the same column, the integer SQL-type converter takes precedence.

                For pyodbc compatibility the ``sqltype`` argument is not validated: any key
                is accepted and stored. A key that matches neither an exact integer ODBC SQL
                type code nor an exact materialized Python type (for example a ``set``, a
                ``dict``, ``object``, or a ``decimal.Decimal`` subclass) is stored but never
                dispatched — registering it is a harmless no-op rather than an error.
            func (callable): The converter function, called with a single parameter (the
                            value) that returns the converted value. The converter is not
                            invoked for SQL NULL values (they are returned as ``None``
                            unchanged). For non-NULL values the parameter is the value
                            already materialized as its Python type (e.g. a
                            ``decimal.Decimal`` or ``datetime.datetime``); string values are
                            passed as their UTF-16LE-encoded ``bytes``.

        Returns:
            None
        """
        with self._converters_lock:
            self._output_converters[sqltype] = func
            self._converters_generation += 1
            # Pass to the underlying connection if native implementation supports it
            if hasattr(self._conn, "add_output_converter"):
                self._conn.add_output_converter(sqltype, func)
        logger.info(f"Added output converter for SQL type {sqltype}")

    def get_output_converter(self, sqltype: Union[int, type]) -> Optional[Callable[[Any], Any]]:
        """
        Get the output converter function for the specified SQL type.

        Thread-safe implementation that protects the converters dictionary with a lock.

        Args:
            sqltype (int or type): The SQL type value or Python type to get the converter for

        Returns:
            callable or None: The converter function or None if no converter is registered

        Note:
            ⚠️ The returned converter function will be executed on database values. Only use
            converters from trusted sources.
        """
        with self._converters_lock:
            return self._output_converters.get(sqltype)

    def remove_output_converter(self, sqltype: Union[int, type]) -> None:
        """
        Remove the output converter function for the specified SQL type.

        Thread-safe implementation that protects the converters dictionary with a lock.

        Existing cursors use the updated converters on their next fetch.

        Args:
            sqltype (int or type): The SQL type value to remove the converter for

        Returns:
            None
        """
        with self._converters_lock:
            if sqltype in self._output_converters:
                del self._output_converters[sqltype]
                self._converters_generation += 1
                # Pass to the underlying connection if native implementation supports it
                if hasattr(self._conn, "remove_output_converter"):
                    self._conn.remove_output_converter(sqltype)
        logger.info(f"Removed output converter for SQL type {sqltype}")

    def clear_output_converters(self) -> None:
        """
        Remove all output converter functions.

        Thread-safe implementation that protects the converters dictionary with a lock.

        Existing cursors stop applying converters on their next fetch.

        Returns:
            None
        """
        with self._converters_lock:
            self._output_converters.clear()
            self._converters_generation += 1
            # Pass to the underlying connection if native implementation supports it
            if hasattr(self._conn, "clear_output_converters"):
                self._conn.clear_output_converters()
        logger.info("Cleared all output converters")

    def execute(self, sql: str, *args: Any) -> Cursor:
        """
        Creates a new Cursor object, calls its execute method, and returns the new cursor.

        This is a convenience method that is not part of the DB API. Since a new Cursor
        is allocated by each call, this should not be used if more than one SQL statement
        needs to be executed on the connection.

        Note on cursor lifecycle management:
        - Each call creates a new cursor that is tracked by the connection's internal WeakSet
        - Cursors are automatically dereferenced/closed when they go out of scope
        - For long-running applications or loops, explicitly call cursor.close() when done
          to release resources immediately rather than waiting for garbage collection

        Args:
            sql (str): The SQL query to execute.
            *args: Parameters to be passed to the query.

        Returns:
            Cursor: A new cursor with the executed query.

        Raises:
            DatabaseError: If there is an error executing the query.
            InterfaceError: If the connection is closed.

        Example:
            # Automatic cleanup (cursor goes out of scope after the operation)
            row = connection.execute("SELECT name FROM users WHERE id = ?", 123).fetchone()

            # Manual cleanup for more explicit resource management
            cursor = connection.execute("SELECT * FROM large_table")
            try:
                # Use cursor...
                rows = cursor.fetchall()
            finally:
                cursor.close()  # Explicitly release resources
        """
        cursor = self.cursor()
        try:
            # Add the cursor to our tracking set BEFORE execution
            # This ensures it's tracked even if execution fails
            self._cursors.add(cursor)

            # Now execute the query
            cursor.execute(sql, *args)
            return cursor
        except Exception:
            # If execution fails, close the cursor to avoid leaking resources
            cursor.close()
            raise

    def batch_execute(
        self,
        statements: List[str],
        params: Optional[List[Union[None, Any, Tuple[Any, ...], List[Any]]]] = None,
        reuse_cursor: Optional[Cursor] = None,
        auto_close: bool = False,
    ) -> Tuple[List[Union[List["Row"], int]], Cursor]:
        """
        Execute multiple SQL statements efficiently using a single cursor.

        This method allows executing multiple SQL statements in sequence using a single
        cursor, which is more efficient than creating a new cursor for each statement.

        Args:
            statements (list): List of SQL statements to execute
            params (list, optional): List of parameter sets corresponding to statements.
                Each item can be None, a single parameter, or a sequence of parameters.
                If None, no parameters will be used for any statement.
            reuse_cursor (Cursor, optional): Existing cursor to reuse instead of creating a new one.
                If None, a new cursor will be created.
            auto_close (bool): Whether to close the cursor after execution if a new one was created.
                Defaults to False. Has no effect if reuse_cursor is provided.

        Returns:
            tuple: (results, cursor) where:
                - results is a list of execution results, one for each statement
                - cursor is the cursor used for execution (useful if you want to keep using it)

        Raises:
            TypeError: If statements is not a list or if params is provided but not a list
            ValueError: If params is provided but has different length than statements
            DatabaseError: If there is an error executing any of the statements
            InterfaceError: If the connection is closed

        Example:
            # Execute multiple statements with a single cursor
            results, _ = conn.batch_execute([
                "INSERT INTO users VALUES (?, ?)",
                "UPDATE stats SET count = count + 1",
                "SELECT * FROM users"
            ], [
                (1, "user1"),
                None,
                None
            ])

            # Last result contains the SELECT results
            for row in results[-1]:
                print(row)

            # Reuse an existing cursor
            my_cursor = conn.cursor()
            results, _ = conn.batch_execute([
                "SELECT * FROM table1",
                "SELECT * FROM table2"
            ], reuse_cursor=my_cursor)

            # Cursor remains open for further use
            my_cursor.execute("SELECT * FROM table3")
        """
        # Validate inputs
        if not isinstance(statements, list):
            raise TypeError("statements must be a list of SQL statements")

        if params is not None:
            if not isinstance(params, list):
                raise TypeError("params must be a list of parameter sets")
            if len(params) != len(statements):
                raise ValueError("params list must have the same length as statements list")
        else:
            # Create a list of None values with the same length as statements
            params = [None] * len(statements)

        # Determine which cursor to use
        is_new_cursor = reuse_cursor is None
        cursor = self.cursor() if is_new_cursor else reuse_cursor

        # Execute statements and collect results
        results = []
        try:
            for i, (stmt, param) in enumerate(zip(statements, params)):
                try:
                    # Execute the statement with parameters if provided
                    if param is not None:
                        cursor.execute(stmt, param)
                    else:
                        cursor.execute(stmt)

                    # For SELECT statements, fetch all rows
                    # For other statements, get the row count
                    if cursor.description is not None:
                        # This is a SELECT statement or similar that returns rows
                        results.append(cursor.fetchall())
                    else:
                        # This is an INSERT, UPDATE, DELETE or similar that doesn't return rows
                        results.append(cursor.rowcount)

                    logger.debug(f"Executed batch statement {i+1}/{len(statements)}")

                except Exception as e:
                    # If a statement fails, include statement context in the error
                    logger.debug(
                        "error",
                        f"Error executing statement {i+1}/{len(statements)}: {e}",
                    )
                    raise

        except Exception:
            # If an error occurs and auto_close is True, close the cursor
            if auto_close:
                try:
                    # Close the cursor regardless of whether it's reused or new
                    cursor.close()
                    logger.debug(
                        "debug",
                        "Automatically closed cursor after batch execution error",
                    )
                except Exception as close_err:
                    logger.debug(
                        "warning",
                        f"Error closing cursor after execution failure: {close_err}",
                    )
            # Re-raise the original exception
            raise

        # Close the cursor if requested and we created a new one
        if is_new_cursor and auto_close:
            cursor.close()
            logger.debug("Automatically closed cursor after batch execution")

        return results, cursor

    def getinfo(self, info_type: int) -> Union[str, int, bool, None]:
        """
        Return general information about the driver and data source.

        Args:
            info_type (int): The type of information to return. See the ODBC
                             SQLGetInfo documentation for the supported values.

        Returns:
            The requested information. The type of the returned value depends
            on the information requested. For registered ODBC types, character values (including
            "Y"/"N") return strings; numeric values and bitmasks return unsigned
            integers. Native retrieval failures, including unsupported types,
            timeouts, and connection loss, are logged and return None.

        Note:
            SQL_DRIVER_HDBC, SQL_DRIVER_HENV and SQL_DRIVER_HLIB are implemented
            by the ODBC Driver Manager, which this driver bypasses. Correct IDs
            do not imply that the selected native provider supports these queries.
            A native provider may return cached metadata even after connection loss;
            getinfo() is not a connection-health check.
            Deprecated GetInfoConstants names are retained for API compatibility,
            not as valid requests. Their integer values may request unrelated information.

        Raises:
            DatabaseError: If a numeric byte result does not match its ODBC type's width.
            InterfaceError: If the connection is closed.
        """
        if self._closed:
            raise InterfaceError(
                driver_error="Cannot get info on closed connection",
                ddbc_error="Cannot get info on closed connection",
            )

        # Check that info_type is an integer
        if not isinstance(info_type, int):
            raise ValueError(f"info_type must be an integer, got {type(info_type).__name__}")

        # Check for invalid info_type values
        if info_type < 0:
            logger.debug("Invalid info_type: %d. Must be non-negative.", info_type)
            return None

        # Get the raw result from the C++ layer
        try:
            raw_result = self._conn.get_info(info_type)
        except Exception as e:  # pylint: disable=broad-exception-caught
            # Preserve the legacy logged-None contract for native retrieval failures.
            logger.warning(f"getinfo({info_type}) failed: {e}")
            return None

        if raw_result is None:
            return None

        # Check if the result is already a simple type
        if isinstance(raw_result, (str, int, bool)):
            return raw_result

        # If it's a dictionary with data and metadata
        if isinstance(raw_result, dict) and "data" in raw_result:
            # Extract data and metadata from the raw result
            data = raw_result["data"]
            length = raw_result["length"]

            logger.debug(
                "getinfo: info_type=%d, length=%r, data_type=%s",
                info_type,
                length,
                type(data),
            )

            # Explicit numeric types take precedence over the legacy high-ID
            # string fallback (e.g. SQL_MAX_IDENTIFIER_LEN is numeric at 10005).
            return_type = _GETINFO_RETURN_TYPES.get(info_type)
            is_string_type = return_type is str or (
                return_type is None and info_type > INFO_TYPE_STRING_THRESHOLD
            )
            if is_string_type:
                # For string data, ensure we properly handle the byte array
                if isinstance(data, bytes):
                    # Make sure we use the correct amount of data based on length
                    actual_data = data[:length]

                    # SQLGetInfoW returns UTF-16LE encoded strings (wide-character ODBC API)
                    # Try UTF-16LE first (expected), then UTF-8 as fallback
                    for encoding in ("utf-16-le", "utf-8"):
                        try:
                            return actual_data.decode(encoding).rstrip("\0")
                        except UnicodeDecodeError:
                            continue

                    # All decodings failed
                    logger.debug(
                        "Failed to decode string in getinfo (info_type=%d) with supported encodings. "
                        "Returning None to avoid silent corruption.",
                        info_type,
                    )
                    return None
                else:
                    # If it's not bytes, return as is
                    return data
            elif isinstance(return_type, struct.Struct):
                if isinstance(data, bytes):
                    if (
                        type(length) is not int
                        or length != return_type.size
                        or len(data) < return_type.size
                    ):
                        raise DatabaseError(
                            driver_error=f"Invalid numeric result length for getinfo({info_type})",
                            ddbc_error=(
                                f"Expected {return_type.size} bytes; "
                                f"got length={length} with {len(data)} bytes of data"
                            ),
                        )
                    return return_type.unpack_from(data)[0]
                # Legacy non-byte payloads must not lose precision or change bool to int.
                if isinstance(data, str) and data.isdecimal():
                    try:
                        return int(data)
                    except ValueError:
                        logger.debug(
                            "Numeric getinfo compatibility value exceeds the integer conversion "
                            "limit; returning it unchanged"
                        )
                return data

            # Preserve legacy handling for unregistered, driver-specific info types.
            if isinstance(data, bytes):
                # Try to convert to string first
                try:
                    return data[:length].decode("utf-8").rstrip("\0")
                except UnicodeDecodeError:
                    pass

                # Try to convert to int for short binary data
                try:
                    if length <= 8:
                        return int.from_bytes(data[:length], "little", signed=True)
                except Exception:  # pylint: disable=broad-exception-caught
                    pass

                # Return as is if we can't determine
                return data

            return data

        return raw_result  # Return as-is

    def commit(self) -> None:
        """
        Commit the current transaction.

        This method commits the current transaction to the database, making all
        changes made during the transaction permanent. It should be called after
        executing a series of SQL statements that modify the database to ensure
        that the changes are saved.

        Raises:
            InterfaceError: If the connection is closed.
            DatabaseError: If there is an error while committing the transaction.
        """
        # Check if connection is closed
        if self._closed or self._conn is None:
            raise InterfaceError(
                driver_error="Cannot commit on a closed connection",
                ddbc_error="Cannot commit on a closed connection",
            )

        # Commit the current transaction
        try:
            self._conn.commit()
        except RuntimeError as e:
            _raise_connection_error(e)
        logger.info("Transaction committed successfully.")

    def rollback(self) -> None:
        """
        Roll back the current transaction.

        This method rolls back the current transaction, undoing all changes made
        during the transaction. It should be called if an error occurs during the
        transaction or if the changes should not be saved.

        Raises:
            InterfaceError: If the connection is closed.
            DatabaseError: If there is an error while rolling back the transaction.
        """
        # Check if connection is closed
        if self._closed or self._conn is None:
            raise InterfaceError(
                driver_error="Cannot rollback on a closed connection",
                ddbc_error="Cannot rollback on a closed connection",
            )

        # Roll back the current transaction
        try:
            self._conn.rollback()
        except RuntimeError as e:
            _raise_connection_error(e)
        logger.info("Transaction rolled back successfully.")

    def close(self) -> None:
        """
        Close the connection now (rather than whenever .__del__() is called).

        This method closes the connection to the database, releasing any resources
        associated with it. After calling this method, the connection object should
        not be used for any further operations. The same applies to all cursor objects
        trying to use the connection. Note that closing a connection without committing
        the changes first will cause an implicit rollback to be performed.

        Raises:
            DatabaseError: If there is an error while closing the connection.
        """
        # Close the connection
        if self._closed:
            return

        # Close all cursors first, but don't let one failure stop the others
        if hasattr(self, "_cursors"):
            # Convert to list to avoid modification during iteration
            cursors_to_close = list(self._cursors)
            close_errors = []

            for cursor in cursors_to_close:
                try:
                    if not cursor.closed:
                        cursor.close()
                except Exception as e:  # pylint: disable=broad-exception-caught
                    # Collect errors but continue closing other cursors
                    close_errors.append(f"Error closing cursor: {e}")
                    logger.warning(f"Error closing cursor: {e}")

            # If there were errors closing cursors, log them but continue
            if close_errors:
                logger.debug(
                    "warning",
                    "Encountered %d errors while closing cursors",
                    len(close_errors),
                )

            # Clear the cursor set explicitly to release any internal
            # references
            self._cursors.clear()

        # Close the connection even if cursor cleanup had issues
        try:
            if self._conn:
                # Native close owns transaction cleanup and fail-closed discard.
                # Avoid duplicate attribute probes and untrusted rollback hints.
                try:
                    self._conn.close(rollback_before_disconnect=True)
                except RuntimeError as e:
                    _raise_connection_error(e)
                finally:
                    self._conn = None
        except Exception as e:
            logger.error(f"Error closing database connection: {e}")
            # Re-raise the connection close error as it's more critical
            raise
        finally:
            # Always mark as closed, even if there were errors
            self._closed = True

        logger.info("Connection closed successfully.")

    def _remove_cursor(self, cursor: Cursor) -> None:
        """
        Remove a cursor from the connection's tracking.

        This method is called when a cursor is closed to ensure proper cleanup.

        Args:
            cursor: The cursor to remove from tracking.
        """
        if hasattr(self, "_cursors"):
            try:
                self._cursors.discard(cursor)
            except Exception:
                pass  # Ignore errors during cleanup

    def __enter__(self) -> "Connection":
        """
        Enter the context manager.

        This method enables the Connection to be used with the 'with' statement.
        When entering the context, it simply returns the connection object itself.

        Returns:
            Connection: The connection object itself.

        Example:
            with connect(connection_string) as conn:
                cursor = conn.cursor()
                cursor.execute("INSERT INTO table VALUES (?)", [value])
                # Transaction will be committed automatically when exiting
        """
        logger.info("Entering connection context manager.")
        return self

    def __exit__(self, exc_type: Any, exc_val: Any, exc_tb: Any) -> None:
        """
        Exit the context manager.

        Implements commit-on-success / rollback-on-exception semantics:
        - If the block exits cleanly and autocommit is off, the transaction
          is committed.
        - If an exception is raised and autocommit is off, the transaction
          is rolled back.
        - The connection is always closed when leaving the block.

        If commit() fails on clean exit, the connection is closed and the
        commit exception is raised. On exception exit, cleanup failures
        (rollback or close) are suppressed so the original user exception
        propagates unchanged.
        """
        if self._closed:
            return
        try:
            if not self.autocommit:
                if exc_type is None:
                    self.commit()
                else:
                    self.rollback()
        except Exception:
            try:
                self.close()
            except Exception:
                logger.warning(
                    "Failed to close connection after failed "
                    "commit/rollback in context manager.",
                    exc_info=True,
                )
            if exc_type is None:
                raise
            return
        try:
            self.close()
        except Exception:
            if exc_type is None:
                raise
            logger.warning(
                "Failed to close connection in context manager.",
                exc_info=True,
            )

    def __del__(self) -> None:
        """
        Destructor to ensure the connection is closed when the connection object
        is no longer needed.
        This is a safety net to ensure resources are cleaned up
        even if close() was not called explicitly.
        """
        if "_closed" not in self.__dict__ or not self._closed:
            try:
                self.close()
            except Exception as e:
                # Dont raise exceptions from __del__ to avoid issues during garbage collection
                logger.warning(f"Error during connection cleanup: {e}")