import io import logging from datetime import tzinfo import pytz from abc import ABC, abstractmethod from typing import Iterable, Optional, Any, Union, Sequence, Dict, Generator, BinaryIO from pytz.exceptions import UnknownTimeZoneError from clickhouse_connect import common from clickhouse_connect.common import version from clickhouse_connect.datatypes.registry import get_from_name from clickhouse_connect.datatypes.base import ClickHouseType from clickhouse_connect.datatypes import dynamic as dynamic_module from clickhouse_connect.driver import tzutil from clickhouse_connect.driver.common import dict_copy, StreamContext, coerce_int, coerce_bool from clickhouse_connect.driver.constants import CH_VERSION_WITH_PROTOCOL, PROTOCOL_VERSION_WITH_LOW_CARD from clickhouse_connect.driver.exceptions import ProgrammingError, OperationalError, DataError from clickhouse_connect.driver.external import ExternalData from clickhouse_connect.driver.insert import InsertContext from clickhouse_connect.driver.options import check_arrow, check_pandas, check_numpy, check_polars, pd, arrow, pl, IS_PANDAS_2 from clickhouse_connect.driver.summary import QuerySummary from clickhouse_connect.driver.models import ColumnDef, SettingDef, SettingStatus from clickhouse_connect.driver.query import QueryResult, to_arrow, to_arrow_batches, QueryContext, arrow_buffer from clickhouse_connect.driver.binding import quote_identifier io.DEFAULT_BUFFER_SIZE = 1024 * 256 logger = logging.getLogger(__name__) arrow_str_setting = 'output_format_arrow_string_as_string' # pylint: disable=too-many-lines # pylint: disable=too-many-public-methods,too-many-arguments,too-many-positional-arguments,too-many-instance-attributes class Client(ABC): """ Base ClickHouse Connect client """ compression: str = None write_compression: str = None protocol_version = 0 valid_transport_settings = set() optional_transport_settings = set() database = None max_error_message = 0 apply_server_timezone = False utc_tz_aware = False show_clickhouse_errors = True def __init__(self, database: str, query_limit: int, uri: str, query_retries: int, server_host_name: Optional[str], apply_server_timezone: Optional[Union[str, bool]], utc_tz_aware: Optional[bool], show_clickhouse_errors: Optional[bool]): """ Shared initialization of ClickHouse Connect client :param database: database name :param query_limit: default LIMIT for queries :param uri: uri for error messages :param utc_tz_aware: Default timezone behavior when the active timezone resolves to UTC. If True, timezone-aware UTC datetimes are returned; otherwise legacy naive datetimes are used. """ self.query_limit = coerce_int(query_limit) self.query_retries = coerce_int(query_retries) if database and not database == '__default__': self.database = database if show_clickhouse_errors is not None: self.show_clickhouse_errors = coerce_bool(show_clickhouse_errors) self.server_host_name = server_host_name self.uri = uri self.utc_tz_aware = bool(utc_tz_aware) self._init_common_settings(apply_server_timezone) def _init_common_settings(self, apply_server_timezone: Optional[Union[str, bool]]): self.server_tz, dst_safe = pytz.UTC, True self.server_version, server_tz = \ tuple(self.command('SELECT version(), timezone()', use_database=False)) try: server_tz = pytz.timezone(server_tz) server_tz, dst_safe = tzutil.normalize_timezone(server_tz) if apply_server_timezone is None: apply_server_timezone = dst_safe self.apply_server_timezone = apply_server_timezone == 'always' or coerce_bool(apply_server_timezone) self.server_tz = server_tz except UnknownTimeZoneError: logger.warning('Warning, server is using an unrecognized timezone %s, will use UTC default', server_tz) if not self.apply_server_timezone and not tzutil.local_tz_dst_safe: logger.warning('local timezone %s may return unexpected times due to Daylight Savings Time/' + 'Summer Time differences', tzutil.local_tz.tzname(None)) readonly = 'readonly' if not self.min_version('19.17'): readonly = common.get_setting('readonly') server_settings = self.query(f'SELECT name, value, {readonly} as readonly FROM system.settings LIMIT 10000') self.server_settings = {row['name']: SettingDef(**row) for row in server_settings.named_results()} if self.min_version(CH_VERSION_WITH_PROTOCOL) and common.get_setting('use_protocol_version'): # Unfortunately we have to validate that the client protocol version is actually used by ClickHouse # since the query parameter could be stripped off (in particular, by CHProxy) test_data = self.raw_query('SELECT 1 AS check', fmt='Native', settings={ 'client_protocol_version': PROTOCOL_VERSION_WITH_LOW_CARD }) if test_data[8:16] == b'\x01\x01\x05check': self.protocol_version = PROTOCOL_VERSION_WITH_LOW_CARD if self._setting_status('date_time_input_format').is_writable: self.set_client_setting('date_time_input_format', 'best_effort') if self._setting_status('allow_experimental_json_type').is_set and \ self._setting_status('cast_string_to_dynamic_use_inference').is_writable: self.set_client_setting('cast_string_to_dynamic_use_inference', '1') if self.min_version('24.8') and not self.min_version('24.10'): dynamic_module.json_serialization_format = 0 def _validate_settings(self, settings: Optional[Dict[str, Any]]) -> Dict[str, str]: """ This strips any ClickHouse settings that are not recognized or are read only. :param settings: Dictionary of setting name and values :return: A filtered dictionary of settings with values rendered as strings """ validated = {} invalid_action = common.get_setting('invalid_setting_action') for key, value in settings.items(): str_value = self._validate_setting(key, value, invalid_action) if str_value is not None: validated[key] = value return validated def _validate_setting(self, key: str, value: Any, invalid_action: str) -> Optional[str]: str_value = str(value) if value is True: str_value = '1' elif value is False: str_value = '0' if key not in self.valid_transport_settings: setting_def = self.server_settings.get(key) current_setting = self.get_client_setting(key) if setting_def and setting_def.value == str_value and (current_setting is None or current_setting == setting_def.value): return None # don't send settings that are already the expected value if setting_def is None or setting_def.readonly: if key in self.optional_transport_settings: return None if invalid_action == 'send': logger.warning('Attempting to send unrecognized or readonly setting %s', key) elif invalid_action == 'drop': logger.warning('Dropping unrecognized or readonly settings %s', key) return None else: raise ProgrammingError(f'Setting {key} is unknown or readonly') from None return str_value def _setting_status(self, key: str) -> SettingStatus: comp_setting = self.server_settings.get(key) if not comp_setting: return SettingStatus(False, False) return SettingStatus(comp_setting.value != '0', comp_setting.readonly != 1) def _prep_query(self, context: QueryContext): if context.is_select and not context.has_limit and self.query_limit: limit = f'\n LIMIT {self.query_limit}' if isinstance(context.query, bytes): return context.final_query + limit.encode() return context.final_query + limit return context.final_query def _check_tz_change(self, new_tz) -> Optional[tzinfo]: if new_tz: try: new_tzinfo = pytz.timezone(new_tz) if new_tzinfo != self.server_tz: return new_tzinfo except UnknownTimeZoneError: logger.warning('Unrecognized timezone %s received from ClickHouse', new_tz) return None @abstractmethod def _query_with_context(self, context: QueryContext): pass @abstractmethod def set_client_setting(self, key, value): """ Set a clickhouse setting for the client after initialization. If a setting is not recognized by ClickHouse, or the setting is identified as "read_only", this call will either throw a Programming exception or attempt to send the setting anyway based on the common setting 'invalid_setting_action' :param key: ClickHouse setting name :param value: ClickHouse setting value """ @abstractmethod def get_client_setting(self, key) -> Optional[str]: """ :param key: The setting key :return: The string value of the setting, if it exists, or None """ @abstractmethod def set_access_token(self, access_token: str): """ Set the ClickHouse access token for the client :param access_token: Access token string """ # pylint: disable=unused-argument,too-many-locals def query(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, Union[str, Dict[str, str]]]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, column_oriented: Optional[bool] = None, use_numpy: Optional[bool] = None, max_str_len: Optional[int] = None, context: QueryContext = None, query_tz: Optional[Union[str, tzinfo]] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> QueryResult: """ Main query method for SELECT, DESCRIBE and other SQL statements that return a result matrix. For parameters, see the create_query_context method :return: QueryResult -- data and metadata from response """ if query and query.lower().strip().startswith('select __connect_version__'): return QueryResult([[f'ClickHouse Connect v.{version()} ⓒ ClickHouse Inc.']], None, ('connect_version',), (get_from_name('String'),)) kwargs = locals().copy() del kwargs['self'] query_context = self.create_query_context(**kwargs) if query_context.is_command: response = self.command(query, parameters=query_context.parameters, settings=query_context.settings, external_data=query_context.external_data, transport_settings=query_context.transport_settings) if isinstance(response, QuerySummary): return response.as_query_result() return QueryResult([response] if isinstance(response, list) else [[response]]) return self._query_with_context(query_context) def query_column_block_stream(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, Union[str, Dict[str, str]]]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, context: QueryContext = None, query_tz: Optional[Union[str, tzinfo]] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> StreamContext: """ Variation of main query method that returns a stream of column oriented blocks. For parameters, see the create_query_context method. :return: StreamContext -- Iterable stream context that returns column oriented blocks """ return self._context_query(locals(), use_numpy=False, streaming=True).column_block_stream def query_row_block_stream(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, Union[str, Dict[str, str]]]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, context: QueryContext = None, query_tz: Optional[Union[str, tzinfo]] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> StreamContext: """ Variation of main query method that returns a stream of row oriented blocks. For parameters, see the create_query_context method. :return: StreamContext -- Iterable stream context that returns blocks of rows """ return self._context_query(locals(), use_numpy=False, streaming=True).row_block_stream def query_rows_stream(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, Union[str, Dict[str, str]]]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, context: QueryContext = None, query_tz: Optional[Union[str, tzinfo]] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> StreamContext: """ Variation of main query method that returns a stream of row oriented blocks. For parameters, see the create_query_context method. :return: StreamContext -- Iterable stream context that returns blocks of rows """ return self._context_query(locals(), use_numpy=False, streaming=True).rows_stream @abstractmethod def raw_query(self, query: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, fmt: str = None, use_database: bool = True, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> bytes: """ Query method that simply returns the raw ClickHouse format bytes :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param fmt: ClickHouse output format :param use_database: Send the database parameter to ClickHouse so the command will be executed in the client database context. :param external_data: External data to send with the query :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: bytes representing raw ClickHouse return value based on format """ @abstractmethod def raw_stream(self, query: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, fmt: str = None, use_database: bool = True, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> io.IOBase: """ Query method that returns the result as an io.IOBase iterator :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param fmt: ClickHouse output format :param use_database Send the database parameter to ClickHouse so the command will be executed in the client database context. :param external_data: External data to send with the query. :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: io.IOBase stream/iterator for the result """ # pylint: disable=duplicate-code,unused-argument def query_np(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, str]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, max_str_len: Optional[int] = None, context: QueryContext = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None): """ Query method that returns the results as a numpy array. For parameter values, see the create_query_context method :return: Numpy array representing the result set """ check_numpy() self._add_integration_tag("numpy") return self._context_query(locals(), use_numpy=True).np_result # pylint: disable=duplicate-code,too-many-arguments,unused-argument def query_np_stream(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, str]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, max_str_len: Optional[int] = None, context: QueryContext = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> StreamContext: """ Query method that returns the results as a stream of numpy arrays. For parameter values, see the create_query_context method :return: Generator that yield a numpy array per block representing the result set """ check_numpy() self._add_integration_tag("numpy") return self._context_query(locals(), use_numpy=True, streaming=True).np_stream # pylint: disable=duplicate-code,unused-argument def query_df(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, str]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, max_str_len: Optional[int] = None, use_na_values: Optional[bool] = None, query_tz: Optional[str] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, context: QueryContext = None, external_data: Optional[ExternalData] = None, use_extended_dtypes: Optional[bool] = None, transport_settings: Optional[Dict[str, str]] = None): """ Query method that results the results as a pandas dataframe. For parameter values, see the create_query_context method :return: Pandas dataframe representing the result set """ check_pandas() self._add_integration_tag("pandas") return self._context_query(locals(), use_numpy=True, as_pandas=True).df_result # pylint: disable=duplicate-code,unused-argument def query_df_stream(self, query: Optional[str] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, str]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, max_str_len: Optional[int] = None, use_na_values: Optional[bool] = None, query_tz: Optional[str] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, context: QueryContext = None, external_data: Optional[ExternalData] = None, use_extended_dtypes: Optional[bool] = None, transport_settings: Optional[Dict[str, str]] = None) -> StreamContext: """ Query method that returns the results as a StreamContext. For parameter values, see the create_query_context method :return: Generator that yields a Pandas dataframe per block representing the result set """ check_pandas() self._add_integration_tag("pandas") return self._context_query(locals(), use_numpy=True, as_pandas=True, streaming=True).df_stream def create_query_context(self, query: Optional[Union[str, bytes]] = None, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, query_formats: Optional[Dict[str, str]] = None, column_formats: Optional[Dict[str, Union[str, Dict[str, str]]]] = None, encoding: Optional[str] = None, use_none: Optional[bool] = None, column_oriented: Optional[bool] = None, use_numpy: Optional[bool] = False, max_str_len: Optional[int] = 0, context: Optional[QueryContext] = None, query_tz: Optional[Union[str, tzinfo]] = None, column_tzs: Optional[Dict[str, Union[str, tzinfo]]] = None, utc_tz_aware: Optional[bool] = None, use_na_values: Optional[bool] = None, streaming: bool = False, as_pandas: bool = False, external_data: Optional[ExternalData] = None, use_extended_dtypes: Optional[bool] = None, transport_settings: Optional[Dict[str, str]] = None) -> QueryContext: """ Creates or updates a reusable QueryContext object :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param query_formats: See QueryContext __init__ docstring :param column_formats: See QueryContext __init__ docstring :param encoding: See QueryContext __init__ docstring :param use_none: Use None for ClickHouse NULL instead of default values. Note that using None in Numpy arrays will force the numpy array dtype to 'object', which is often inefficient. This effect also will impact the performance of Pandas dataframes. :param column_oriented: Deprecated. Controls orientation of the QueryResult result_set property :param use_numpy: Return QueryResult columns as one-dimensional numpy arrays :param max_str_len: Limit returned ClickHouse String values to this length, which allows a Numpy structured array even with ClickHouse variable length String columns. If 0, Numpy arrays for String columns will always be object arrays :param context: An existing QueryContext to be updated with any provided parameter values :param query_tz: Either a string or a pytz tzinfo object. (Strings will be converted to tzinfo objects). Values for any DateTime or DateTime64 column in the query will be converted to Python datetime.datetime objects with the selected timezone. :param column_tzs: A dictionary of column names to tzinfo objects (or strings that will be converted to tzinfo objects). The timezone will be applied to datetime objects returned in the query :param utc_tz_aware: Override the client default for handling UTC results. True forces timezone-aware UTC datetimes while False returns naive UTC datetimes. :param use_na_values: Deprecated alias for use_advanced_dtypes :param as_pandas Return the result columns as pandas.Series objects :param streaming Marker used to correctly configure streaming queries :param external_data ClickHouse "external data" to send with query :param use_extended_dtypes: Only relevant to Pandas Dataframe queries. Use Pandas "missing types", such as pandas.NA and pandas.NaT for ClickHouse NULL values, as well as extended Pandas dtypes such as IntegerArray and StringArray. Defaulted to True for query_df methods :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: Reusable QueryContext """ resolved_utc_tz_aware = self.utc_tz_aware if utc_tz_aware is None else utc_tz_aware if context: return context.updated_copy(query=query, parameters=parameters, settings=settings, query_formats=query_formats, column_formats=column_formats, encoding=encoding, server_tz=self.server_tz, use_none=use_none, column_oriented=column_oriented, use_numpy=use_numpy, max_str_len=max_str_len, query_tz=query_tz, column_tzs=column_tzs, utc_tz_aware=resolved_utc_tz_aware, as_pandas=as_pandas, use_extended_dtypes=use_extended_dtypes, streaming=streaming, external_data=external_data, transport_settings=transport_settings) if use_numpy and max_str_len is None: max_str_len = 0 if use_extended_dtypes is None: use_extended_dtypes = use_na_values if as_pandas and use_extended_dtypes is None: use_extended_dtypes = True return QueryContext(query=query, parameters=parameters, settings=settings, query_formats=query_formats, column_formats=column_formats, encoding=encoding, server_tz=self.server_tz, use_none=use_none, column_oriented=column_oriented, use_numpy=use_numpy, max_str_len=max_str_len, query_tz=query_tz, column_tzs=column_tzs, utc_tz_aware=resolved_utc_tz_aware, use_extended_dtypes=use_extended_dtypes, as_pandas=as_pandas, streaming=streaming, apply_server_tz=self.apply_server_timezone, external_data=external_data, transport_settings=transport_settings) def query_arrow(self, query: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, use_strings: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None): """ Query method using the ClickHouse Arrow format to return a PyArrow table :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param use_strings: Convert ClickHouse String type to Arrow string type (instead of binary) :param external_data: ClickHouse "external data" to send with query :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: PyArrow.Table """ check_arrow() self._add_integration_tag("arrow") settings = self._update_arrow_settings(settings, use_strings) return to_arrow(self.raw_query(query, parameters, settings, fmt='Arrow', external_data=external_data, transport_settings=transport_settings)) def query_arrow_stream(self, query: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, use_strings: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> StreamContext: """ Query method that returns the results as a stream of Arrow tables :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param use_strings: Convert ClickHouse String type to Arrow string type (instead of binary) :param external_data: ClickHouse "external data" to send with query :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: Generator that yields a PyArrow.Table for per block representing the result set """ check_arrow() self._add_integration_tag("arrow") settings = self._update_arrow_settings(settings, use_strings) return to_arrow_batches(self.raw_stream(query, parameters, settings, fmt='ArrowStream', external_data=external_data, transport_settings=transport_settings)) def query_df_arrow(self, query: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, use_strings: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None, dataframe_library: str = "pandas" ) -> Union["pd.DataFrame", "pl.DataFrame"]: """ Query method using the ClickHouse Arrow format to return a DataFrame with PyArrow dtype backend. This provides better performance and memory efficiency compared to the standard query_df method, though fewer output formatting options. :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param use_strings: Convert ClickHouse String type to Arrow string type (instead of binary) :param external_data: ClickHouse "external data" to send with query :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :param dataframe_library: Library to use for DataFrame creation ("pandas" or "polars") :return: DataFrame (pandas or polars based on dataframe_library parameter) """ check_arrow() if dataframe_library == "pandas": check_pandas() self._add_integration_tag("pandas") if not IS_PANDAS_2: raise ProgrammingError("PyArrow-backed dtypes are only supported when using pandas 2.x.") def converter(table: arrow.Table) -> pd.DataFrame: return table.to_pandas(types_mapper=pd.ArrowDtype, safe=False) elif dataframe_library == "polars": check_polars() self._add_integration_tag("polars") def converter(table: arrow.Table) -> pl.DataFrame: return pl.from_arrow(table) else: raise ValueError(f"dataframe_library must be 'pandas' or 'polars', got '{dataframe_library}'") arrow_table = self.query_arrow( query=query, parameters=parameters, settings=settings, use_strings=use_strings, external_data=external_data, transport_settings=transport_settings, ) return converter(arrow_table) def query_df_arrow_stream(self, query: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, settings: Optional[Dict[str, Any]] = None, use_strings: Optional[bool] = None, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None, dataframe_library: str = "pandas") -> StreamContext: """ Query method that returns the results as a stream of DataFrames with PyArrow dtype backend. Each DataFrame represents a block from the ClickHouse response. :param query: Query statement/format string :param parameters: Optional dictionary used to format the query :param settings: Optional dictionary of ClickHouse settings (key/string values) :param use_strings: Convert ClickHouse String type to Arrow string type (instead of binary) :param external_data: ClickHouse "external data" to send with query :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :param dataframe_library: Library to use for DataFrame creation ("pandas" or "polars") :return: StreamContext that yields DataFrames (pandas or polars based on dataframe_library parameter) """ check_arrow() if dataframe_library == "pandas": check_pandas() self._add_integration_tag("pandas") if not IS_PANDAS_2: raise ProgrammingError("PyArrow-backed dtypes are only supported when using pandas 2.x.") def converter(table: "arrow.Table") -> "pd.DataFrame": return table.to_pandas(types_mapper=pd.ArrowDtype, safe=False) elif dataframe_library == "polars": check_polars() self._add_integration_tag("polars") def converter(table: arrow.Table) -> pl.DataFrame: return pl.from_arrow(table) else: raise ValueError(f"dataframe_library must be 'pandas' or 'polars', got '{dataframe_library}'") settings = self._update_arrow_settings(settings, use_strings) raw_stream = self.raw_stream( query, parameters, settings, fmt="ArrowStream", external_data=external_data, transport_settings=transport_settings ) reader = arrow.ipc.open_stream(raw_stream) def df_generator(): for batch in reader: yield converter(batch) return StreamContext(raw_stream, df_generator()) def _update_arrow_settings(self, settings: Optional[Dict[str, Any]], use_strings: Optional[bool]) -> Dict[str, Any]: settings = dict_copy(settings) if self.database: settings['database'] = self.database str_status = self._setting_status(arrow_str_setting) if use_strings is None: if str_status.is_writable and not str_status.is_set: settings[arrow_str_setting] = '1' # Default to returning strings if possible elif use_strings != str_status.is_set: if not str_status.is_writable: raise OperationalError(f'Cannot change readonly {arrow_str_setting} to {use_strings}') settings[arrow_str_setting] = '1' if use_strings else '0' return settings @abstractmethod def command(self, cmd: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None, data: Union[str, bytes] = None, settings: Dict[str, Any] = None, use_database: bool = True, external_data: Optional[ExternalData] = None, transport_settings: Optional[Dict[str, str]] = None) -> Union[str, int, Sequence[str], QuerySummary]: """ Client method that returns a single value instead of a result set :param cmd: ClickHouse query/command as a python format string :param parameters: Optional dictionary of key/values pairs to be formatted :param data: Optional 'data' for the command (for INSERT INTO in particular) :param settings: Optional dictionary of ClickHouse settings (key/string values) :param use_database: Send the database parameter to ClickHouse so the command will be executed in the client database context. Otherwise, no database will be specified with the command. This is useful for determining the default user database :param external_data: ClickHouse "external data" to send with command/query :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: Decoded response from ClickHouse as either a string, int, or sequence of strings, or QuerySummary if no data returned """ @abstractmethod def ping(self) -> bool: """ Validate the connection, does not throw an Exception (see debug logs) :return: ClickHouse server is up and reachable """ def insert(self, table: Optional[str] = None, data: Sequence[Sequence[Any]] = None, column_names: Union[str, Iterable[str]] = '*', database: Optional[str] = None, column_types: Sequence[ClickHouseType] = None, column_type_names: Sequence[str] = None, column_oriented: bool = False, settings: Optional[Dict[str, Any]] = None, context: InsertContext = None, transport_settings: Optional[Dict[str, str]] = None) -> QuerySummary: """ Method to insert multiple rows/data matrix of native Python objects. If context is specified arguments other than data are ignored :param table: Target table :param data: Sequence of sequences of Python data :param column_names: Ordered list of column names or '*' if column types should be retrieved from the ClickHouse table definition :param database: Target database -- will use client default database if not specified. :param column_types: ClickHouse column types. If set then column data does not need to be retrieved from the server :param column_type_names: ClickHouse column type names. If set then column data does not need to be retrieved from the server :param column_oriented: If true the data is already "pivoted" in column form :param settings: Optional dictionary of ClickHouse settings (key/string values) :param context: Optional reusable insert context to allow repeated inserts into the same table with different data batches :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: QuerySummary with summary information, throws exception if insert fails """ if (context is None or context.empty) and data is None: raise ProgrammingError('No data specified for insert') from None if context is None: context = self.create_insert_context(table, column_names, database, column_types, column_type_names, column_oriented, settings, transport_settings=transport_settings) if data is not None: if not context.empty: raise ProgrammingError('Attempting to insert new data with non-empty insert context') from None context.data = data return self.data_insert(context) def insert_df(self, table: str = None, df=None, database: Optional[str] = None, settings: Optional[Dict] = None, column_names: Optional[Sequence[str]] = None, column_types: Sequence[ClickHouseType] = None, column_type_names: Sequence[str] = None, context: InsertContext = None, transport_settings: Optional[Dict[str, str]] = None) -> QuerySummary: """ Insert a pandas DataFrame into ClickHouse. If context is specified arguments other than df are ignored :param table: ClickHouse table :param df: two-dimensional pandas dataframe :param database: Optional ClickHouse database :param settings: Optional dictionary of ClickHouse settings (key/string values) :param column_names: An optional list of ClickHouse column names. If not set, the DataFrame column names will be used :param column_types: ClickHouse column types. If set then column data does not need to be retrieved from the server :param column_type_names: ClickHouse column type names. If set then column data does not need to be retrieved from the server :param context: Optional reusable insert context to allow repeated inserts into the same table with different data batches :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: QuerySummary with summary information, throws exception if insert fails """ check_pandas() self._add_integration_tag("pandas") if context is None: if column_names is None: column_names = df.columns elif len(column_names) != len(df.columns): raise ProgrammingError('DataFrame column count does not match insert_columns') from None return self.insert(table, df, column_names, database, column_types=column_types, column_type_names=column_type_names, settings=settings, transport_settings=transport_settings, context=context) def insert_arrow(self, table: str, arrow_table, database: str = None, settings: Optional[Dict] = None, transport_settings: Optional[Dict[str, str]] = None) -> QuerySummary: """ Insert a PyArrow table DataFrame into ClickHouse using raw Arrow format :param table: ClickHouse table :param arrow_table: PyArrow Table object :param database: Optional ClickHouse database :param settings: Optional dictionary of ClickHouse settings (key/string values) :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) """ check_arrow() self._add_integration_tag("arrow") full_table = table if '.' in table or not database else f'{database}.{table}' compression = self.write_compression if self.write_compression in ('zstd', 'lz4') else None column_names, insert_block = arrow_buffer(arrow_table, compression) return self.raw_insert(full_table, column_names, insert_block, settings, 'Arrow', transport_settings) def insert_df_arrow(self, table: str, df: Union["pd.DataFrame", "pl.DataFrame"], database: Optional[str] = None, settings: Optional[Dict] = None, transport_settings: Optional[Dict[str, str]] = None) -> QuerySummary: """ Insert a pandas DataFrame with PyArrow backend or a polars DataFrame into ClickHouse using Arrow format. This method is optimized for DataFrames that already use Arrow format, providing better performance than the standard insert_df method. Validation is performed and an exception will be raised if this requirement is not met. Polars DataFrames are natively Arrow-based and don't require additional validation. :param table: ClickHouse table name :param df: Pandas DataFrame with PyArrow dtype backend or Polars DataFrame :param database: Optional ClickHouse database name :param settings: Optional dictionary of ClickHouse settings (key/string values) :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: QuerySummary with summary information, throws exception if insert fails """ check_arrow() if pd is not None and isinstance(df, pd.DataFrame): df_lib = "pandas" elif pl is not None and isinstance(df, pl.DataFrame): df_lib = "polars" else: if pd is None and pl is None: raise ImportError("A DataFrame library (pandas or polars) must be installed to use insert_df_arrow.") raise TypeError(f"df must be either a pandas DataFrame or polars DataFrame, got {type(df).__name__}") if df_lib == "pandas": if not IS_PANDAS_2: raise ProgrammingError("PyArrow-backed dtypes are only supported when using pandas 2.x.") non_arrow_cols = [col for col, dtype in df.dtypes.items() if not isinstance(dtype, pd.ArrowDtype)] if non_arrow_cols: raise ProgrammingError( f"insert_df_arrow requires all columns to use PyArrow dtypes. Non-Arrow columns found: [{', '.join(non_arrow_cols)}]. " ) try: arrow_table = arrow.Table.from_pandas(df, preserve_index=False) except Exception as e: raise DataError(f"Failed to convert pandas DataFrame to Arrow table: {e}") from e else: try: arrow_table = df.to_arrow() except Exception as e: raise DataError(f"Failed to convert polars DataFrame to Arrow table: {e}") from e self._add_integration_tag(df_lib) return self.insert_arrow( table=table, arrow_table=arrow_table, database=database, settings=settings, transport_settings=transport_settings, ) def create_insert_context(self, table: str, column_names: Optional[Union[str, Sequence[str]]] = None, database: Optional[str] = None, column_types: Sequence[ClickHouseType] = None, column_type_names: Sequence[str] = None, column_oriented: bool = False, settings: Optional[Dict[str, Any]] = None, data: Optional[Sequence[Sequence[Any]]] = None, transport_settings: Optional[Dict[str, str]] = None) -> InsertContext: """ Builds a reusable insert context to hold state for a duration of an insert :param table: Target table :param database: Target database. If not set, uses the client default database :param column_names: Optional ordered list of column names. If not set, all columns ('*') will be assumed in the order specified by the table definition :param database: Target database -- will use client default database if not specified :param column_types: ClickHouse column types. Optional Sequence of ClickHouseType objects. If neither column types nor column type names are set, actual column types will be retrieved from the server. :param column_type_names: ClickHouse column type names. Specified column types by name string :param column_oriented: If true the data is already "pivoted" in column form :param settings: Optional dictionary of ClickHouse settings (key/string values) :param data: Initial dataset for insert :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) :return: Reusable insert context """ full_table = table if '.' not in table: if database: full_table = f'{quote_identifier(database)}.{quote_identifier(table)}' else: full_table = quote_identifier(table) column_defs = [] if column_types is None and column_type_names is None: describe_result = self.query(f'DESCRIBE TABLE {full_table}', settings=settings) column_defs = [ColumnDef(**row) for row in describe_result.named_results() if row['default_type'] not in ('ALIAS', 'MATERIALIZED')] if column_names is None or isinstance(column_names, str) and column_names == '*': column_names = [cd.name for cd in column_defs] column_types = [cd.ch_type for cd in column_defs] elif isinstance(column_names, str): column_names = [column_names] if len(column_names) == 0: raise ValueError('Column names must be specified for insert') if not column_types: if column_type_names: column_types = [get_from_name(name) for name in column_type_names] else: column_map = {d.name: d for d in column_defs} try: column_types = [column_map[name].ch_type for name in column_names] except KeyError as ex: raise ProgrammingError(f'Unrecognized column {ex} in table {table}') from None if len(column_names) != len(column_types): raise ProgrammingError('Column names do not match column types') from None return InsertContext(full_table, column_names, column_types, column_oriented=column_oriented, settings=settings, transport_settings=transport_settings, data=data) def min_version(self, version_str: str) -> bool: """ Determine whether the connected server is at least the submitted version For Altinity Stable versions like 22.8.15.25.altinitystable the last condition in the first list comprehension expression is added :param version_str: A version string consisting of up to 4 integers delimited by dots :return: True if version_str is greater than the server_version, False if less than """ try: server_parts = [int(x) for x in self.server_version.split('.') if x.isnumeric()] server_parts.extend([0] * (4 - len(server_parts))) version_parts = [int(x) for x in version_str.split('.')] version_parts.extend([0] * (4 - len(version_parts))) except ValueError: logger.warning('Server %s or requested version %s does not match format of numbers separated by dots', self.server_version, version_str) return False for x, y in zip(server_parts, version_parts): if x > y: return True if x < y: return False return True # pylint: disable=no-self-use def _add_integration_tag(self, name: str) -> None: """Transport hook to surface 3rd party lib integration info (default: no-op).""" return @abstractmethod def data_insert(self, context: InsertContext) -> QuerySummary: """ Subclass implementation of the data insert :context: InsertContext parameter object :return: No return, throws an exception if the insert fails """ @abstractmethod def raw_insert(self, table: str, column_names: Optional[Sequence[str]] = None, insert_block: Union[str, bytes, Generator[bytes, None, None], BinaryIO] = None, settings: Optional[Dict] = None, fmt: Optional[str] = None, compression: Optional[str] = None, transport_settings: Optional[Dict[str, str]] = None) -> QuerySummary: """ Insert data already formatted in a bytes object :param table: Table name (whether qualified with the database name or not) :param column_names: Sequence of column names :param insert_block: Binary or string data already in a recognized ClickHouse format :param settings: Optional dictionary of ClickHouse settings (key/string values) :param fmt: Valid clickhouse format :param compression: Recognized ClickHouse `Accept-Encoding` header compression value :param transport_settings: Optional dictionary of transport level settings (HTTP headers, etc.) """ @abstractmethod def close(self): """ Subclass implementation to close the connection to the server/deallocate the client """ @abstractmethod def close_connections(self): """ Subclass implementation to disconnect all "re-used" client connections """ def _context_query(self, lcls: dict, **overrides): kwargs = lcls.copy() kwargs.pop('self') kwargs.update(overrides) return self._query_with_context((self.create_query_context(**kwargs))) def __enter__(self): return self def __exit__(self, exc_type, exc_value, exc_traceback): self.close()