"""``SQLDataset`` to load and save data to a SQL backend."""
from __future__ import annotations
import copy
import datetime as dt
import re
from pathlib import PurePosixPath
from typing import Any, NoReturn
import fsspec
import pandas as pd
from kedro.io.core import (
AbstractDataset,
DatasetError,
get_filepath_str,
get_protocol_and_path,
)
from sqlalchemy import create_engine, inspect
from sqlalchemy.exc import NoSuchModuleError
__all__ = ["SQLTableDataset", "SQLQueryDataset"]
KNOWN_PIP_INSTALL = {
"psycopg2": "psycopg2",
"mysqldb": "mysqlclient",
"cx_Oracle": "cx_Oracle",
"mssql": "pyodbc",
}
DRIVER_ERROR_MESSAGE = """
A module/driver is missing when connecting to your SQL server. SQLDataset
supports SQLAlchemy drivers. Please refer to
https://docs.sqlalchemy.org/core/engines.html#supported-databases
for more information.
\n\n
"""
def _find_known_drivers(module_import_error: ImportError) -> str | None:
"""Looks up known keywords in a ``ModuleNotFoundError`` so that it can
provide better guideline for the user.
Args:
module_import_error: Error raised while connecting to a SQL server.
Returns:
Instructions for installing missing driver. An empty string is
returned in case error is related to an unknown driver.
"""
# module errors contain string "No module name 'module_name'"
# we are trying to extract module_name surrounded by quotes here
res = re.findall(r"'(.*?)'", str(module_import_error.args[0]).lower())
# in case module import error does not match our expected pattern
# we have no recommendation
if not res:
return None
missing_module = res[0]
if KNOWN_PIP_INSTALL.get(missing_module):
return (
f"You can also try installing missing driver with\n"
f"\npip install {KNOWN_PIP_INSTALL.get(missing_module)}"
)
return None
def _get_missing_module_error(import_error: ImportError) -> DatasetError:
missing_module_instruction = _find_known_drivers(import_error)
if missing_module_instruction is None:
return DatasetError(
f"{DRIVER_ERROR_MESSAGE}Loading failed with error:\n\n{str(import_error)}"
)
return DatasetError(f"{DRIVER_ERROR_MESSAGE}{missing_module_instruction}")
def _get_sql_alchemy_missing_error() -> DatasetError:
return DatasetError(
"The SQL dialect in your connection is not supported by "
"SQLAlchemy. Please refer to "
"https://docs.sqlalchemy.org/core/engines.html#supported-databases "
"for more information."
)
[docs]
class SQLTableDataset(AbstractDataset[pd.DataFrame, pd.DataFrame]):
"""``SQLTableDataset`` loads data from a SQL table and saves a pandas
dataframe to a table. It uses ``pandas.DataFrame`` internally,
so it supports all allowed pandas options on ``read_sql_table`` and
``to_sql`` methods. Since Pandas uses SQLAlchemy behind the scenes, when
instantiating ``SQLTableDataset`` one needs to pass a compatible connection
string either in ``credentials`` (see the example code snippet below) or in
``load_args`` and ``save_args``. Connection string formats supported by
SQLAlchemy can be found here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
``SQLTableDataset`` modifies the save parameters and stores
the data with no index. This is designed to make load and save methods
symmetric.
Example usage for the
`YAML API <https://kedro.readthedocs.io/en/stable/data/\
data_catalog_yaml_examples.html>`_:
.. code-block:: yaml
shuttles_table_dataset:
type: pandas.SQLTableDataset
credentials: db_credentials
table_name: shuttles
load_args:
schema: dwschema
save_args:
schema: dwschema
if_exists: replace
Sample database credentials entry in ``credentials.yml``:
.. code-block:: yaml
db_credentials:
con: postgresql://scott:tiger@localhost/test
Example usage for the
`Python API <https://kedro.readthedocs.io/en/stable/data/\
advanced_data_catalog_usage.html>`_:
.. code-block:: pycon
>>> from kedro_datasets.pandas import SQLTableDataset
>>> import pandas as pd
>>>
>>> data = pd.DataFrame({"col1": [1, 2], "col2": [4, 5], "col3": [5, 6]})
>>> table_name = "table_a"
>>> credentials = {"con": f"sqlite:///{tmp_path / 'test.db'}"}
>>> dataset = SQLTableDataset(table_name=table_name, credentials=credentials)
>>>
>>> dataset.save(data)
>>> reloaded = dataset.load()
>>>
>>> assert data.equals(reloaded)
"""
DEFAULT_LOAD_ARGS: dict[str, Any] = {}
DEFAULT_SAVE_ARGS: dict[str, Any] = {"index": False}
# using Any because of Sphinx but it should be
# sqlalchemy.engine.Engine or sqlalchemy.engine.base.Engine
engines: dict[str, Any] = {}
[docs]
def __init__( # noqa: PLR0913
self,
*,
table_name: str,
credentials: dict[str, Any],
load_args: dict[str, Any] = None,
save_args: dict[str, Any] = None,
metadata: dict[str, Any] = None,
) -> None:
"""Creates a new ``SQLTableDataset``.
Args:
table_name: The table name to load or save data to. It
overwrites name in ``save_args`` and ``table_name``
parameters in ``load_args``.
credentials: A dictionary with a ``SQLAlchemy`` connection string.
Users are supposed to provide the connection string 'con'
through credentials.
To find all supported connection string formats, see here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
load_args: Provided to underlying pandas ``read_sql_table``
function along with the connection string.
To find all supported arguments, see here:
https://pandas.pydata.org/pandas-docs/stable/generated/pandas.read_sql_table.html
To find all supported connection string formats, see here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
save_args: Provided to underlying pandas ``to_sql`` function along
with the connection string.
To find all supported arguments, see here:
https://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.to_sql.html
To find all supported connection string formats, see here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
It has ``index=False`` in the default parameters.
metadata: Any arbitrary metadata.
This is ignored by Kedro, but may be consumed by users or external plugins.
Raises:
DatasetError: When either ``table_name`` or ``con`` is empty.
"""
if not table_name:
raise DatasetError("'table_name' argument cannot be empty.")
if not (credentials and "con" in credentials and credentials["con"]):
raise DatasetError(
"'con' argument cannot be empty. Please "
"provide a SQLAlchemy connection string."
)
# Handle default load and save arguments
self._load_args = copy.deepcopy(self.DEFAULT_LOAD_ARGS)
if load_args is not None:
self._load_args.update(load_args)
self._save_args = copy.deepcopy(self.DEFAULT_SAVE_ARGS)
if save_args is not None:
self._save_args.update(save_args)
self._load_args["table_name"] = table_name
self._save_args["name"] = table_name
self._connection_str = credentials["con"]
self.metadata = metadata
[docs]
@classmethod
def create_connection(cls, connection_str: str) -> None:
"""Given a connection string, create singleton connection
to be used across all instances of ``SQLTableDataset`` that
need to connect to the same source.
"""
try:
engine = create_engine(connection_str)
except ImportError as import_error:
raise _get_missing_module_error(import_error) from import_error
except NoSuchModuleError as exc:
raise _get_sql_alchemy_missing_error() from exc
cls.engines[connection_str] = engine
@property
def engine(self):
"""The ``Engine`` object for the dataset's connection string."""
cls = type(self)
if self._connection_str not in cls.engines:
self.create_connection(self._connection_str)
return cls.engines[self._connection_str]
def _describe(self) -> dict[str, Any]:
load_args = copy.deepcopy(self._load_args)
save_args = copy.deepcopy(self._save_args)
del load_args["table_name"]
del save_args["name"]
return {
"table_name": self._load_args["table_name"],
"load_args": load_args,
"save_args": save_args,
}
def _load(self) -> pd.DataFrame:
return pd.read_sql_table(con=self.engine, **self._load_args)
def _save(self, data: pd.DataFrame) -> None:
data.to_sql(con=self.engine, **self._save_args)
def _exists(self) -> bool:
insp = inspect(self.engine)
schema = self._load_args.get("schema", None)
return insp.has_table(self._load_args["table_name"], schema)
[docs]
class SQLQueryDataset(AbstractDataset[None, pd.DataFrame]):
"""``SQLQueryDataset`` loads data from a provided SQL query. It
uses ``pandas.DataFrame`` internally, so it supports all allowed
pandas options on ``read_sql_query``. Since Pandas uses SQLAlchemy behind
the scenes, when instantiating ``SQLQueryDataset`` one needs to pass
a compatible connection string either in ``credentials`` (see the example
code snippet below) or in ``load_args``. Connection string formats supported
by SQLAlchemy can be found here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
It does not support save method so it is a read only data set.
To save data to a SQL server use ``SQLTableDataset``.
Example usage for the
`YAML API <https://kedro.readthedocs.io/en/stable/data/\
data_catalog_yaml_examples.html>`_:
.. code-block:: yaml
shuttle_id_dataset:
type: pandas.SQLQueryDataset
sql: "select shuttle, shuttle_id from spaceflights.shuttles;"
credentials: db_credentials
Advanced example using the ``stream_results`` and ``chunksize`` options to reduce memory usage:
.. code-block:: yaml
shuttle_id_dataset:
type: pandas.SQLQueryDataset
sql: "select shuttle, shuttle_id from spaceflights.shuttles;"
credentials: db_credentials
execution_options:
stream_results: true
load_args:
chunksize: 1000
Sample database credentials entry in ``credentials.yml``:
.. code-block:: yaml
db_credentials:
con: postgresql://scott:tiger@localhost/test
Example usage for the
`Python API <https://kedro.readthedocs.io/en/stable/data/\
advanced_data_catalog_usage.html>`_:
.. code-block:: pycon
>>> import sqlite3
>>>
>>> from kedro_datasets.pandas import SQLQueryDataset
>>> import pandas as pd
>>>
>>> data = pd.DataFrame({"col1": [1, 2], "col2": [4, 5], "col3": [5, 6]})
>>> sql = "SELECT * FROM table_a"
>>> credentials = {"con": f"sqlite:///{tmp_path / 'test.db'}"}
>>> dataset = SQLQueryDataset(sql=sql, credentials=credentials)
>>>
>>> con = sqlite3.connect(tmp_path / "test.db")
>>> cur = con.cursor()
>>> cur.execute("CREATE TABLE table_a(col1, col2, col3)")
<sqlite3.Cursor object at 0x...>
>>> cur.execute("INSERT INTO table_a VALUES (1, 4, 5), (2, 5, 6)")
<sqlite3.Cursor object at 0x...>
>>> con.commit()
>>> reloaded = dataset.load()
>>>
>>> assert data.equals(reloaded)
Example of usage for MSSQL:
.. code-block:: pycon
>>> credentials = {
... "server": "localhost",
... "port": "1433",
... "database": "TestDB",
... "user": "SA",
... "password": "StrongPassword",
... }
>>> def _make_mssql_connection_str(
... server: str, port: str, database: str, user: str, password: str
... ) -> str:
... import pyodbc
... from sqlalchemy.engine import URL
... driver = pyodbc.drivers()[-1]
... connection_str = (
... f"DRIVER={driver};SERVER={server},{port};DATABASE={database};"
... f"ENCRYPT=yes;UID={user};PWD={password};"
... f"TrustServerCertificate=yes;"
... )
... return URL.create("mssql+pyodbc", query={"odbc_connect": connection_str})
...
>>> connection_str = _make_mssql_connection_str(**credentials) # doctest: +SKIP
>>> dataset = SQLQueryDataset( # doctest: +SKIP
... credentials={"con": connection_str}, sql="SELECT TOP 5 * FROM TestTable;"
... )
>>> df = dataset.load()
In addition, here is an example of a catalog with dates parsing:
.. code-block:: yaml
mssql_dataset:
type: kedro_datasets.pandas.SQLQueryDataset
credentials: mssql_credentials
sql: >
SELECT *
FROM DateTable
WHERE date >= ? AND date <= ?
ORDER BY date
load_args:
params:
- ${begin}
- ${end}
index_col: date
parse_dates:
date: "%Y-%m-%d %H:%M:%S.%f0 %z"
"""
# using Any because of Sphinx but it should be
# sqlalchemy.engine.Engine or sqlalchemy.engine.base.Engine
engines: dict[str, Any] = {}
[docs]
def __init__( # noqa: PLR0913
self,
sql: str = None,
credentials: dict[str, Any] = None,
load_args: dict[str, Any] = None,
fs_args: dict[str, Any] = None,
filepath: str = None,
execution_options: dict[str, Any] | None = None,
metadata: dict[str, Any] = None,
) -> None:
"""Creates a new ``SQLQueryDataset``.
Args:
sql: The sql query statement.
credentials: A dictionary with a ``SQLAlchemy`` connection string.
Users are supposed to provide the connection string 'con'
through credentials. It overwrites `con` parameter in
``load_args`` and ``save_args`` in case it is provided. To find
all supported connection string formats, see here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
load_args: Provided to underlying pandas ``read_sql_query``
function along with the connection string.
To find all supported arguments, see here:
https://pandas.pydata.org/pandas-docs/stable/generated/pandas.read_sql_query.html
To find all supported connection string formats, see here:
https://docs.sqlalchemy.org/core/engines.html#database-urls
fs_args: Extra arguments to pass into underlying filesystem class constructor
(e.g. `{"project": "my-project"}` for ``GCSFileSystem``), as well as
to pass to the filesystem's `open` method through nested keys
`open_args_load` and `open_args_save`.
Here you can find all available arguments for `open`:
https://filesystem-spec.readthedocs.io/en/latest/api.html#fsspec.spec.AbstractFileSystem.open
All defaults are preserved, except `mode`, which is set to `r` when loading.
filepath: A path to a file with a sql query statement.
execution_options: A dictionary with non-SQL advanced options for the connection to
be applied to the underlying engine. To find all supported execution
options, see here:
https://docs.sqlalchemy.org/core/connections.html#sqlalchemy.engine.Connection.execution_options
Note that this is not a standard argument supported by pandas API, but could be
useful for handling large datasets.
metadata: Any arbitrary metadata.
This is ignored by Kedro, but may be consumed by users or external plugins.
Raises:
DatasetError: When either ``sql`` or ``con`` parameters is empty.
"""
if sql and filepath:
raise DatasetError(
"'sql' and 'filepath' arguments cannot both be provided."
"Please only provide one."
)
if not (sql or filepath):
raise DatasetError(
"'sql' and 'filepath' arguments cannot both be empty."
"Please provide a sql query or path to a sql query file."
)
if not (credentials and "con" in credentials and credentials["con"]):
raise DatasetError(
"'con' argument cannot be empty. Please "
"provide a SQLAlchemy connection string."
)
default_load_args: dict[str, Any] = {}
self._load_args = (
{**default_load_args, **load_args}
if load_args is not None
else default_load_args
)
self.metadata = metadata
# load sql query from file
if sql:
self._load_args["sql"] = sql
self._filepath = None
else:
# filesystem for loading sql file
_fs_args = copy.deepcopy(fs_args) or {}
_fs_credentials = _fs_args.pop("credentials", {})
protocol, path = get_protocol_and_path(str(filepath))
self._protocol = protocol
self._fs = fsspec.filesystem(self._protocol, **_fs_credentials, **_fs_args)
self._filepath = path
self._connection_str = credentials["con"]
self._execution_options = execution_options or {}
if "mssql" in self._connection_str:
self.adapt_mssql_date_params()
[docs]
@classmethod
def create_connection(cls, connection_str: str) -> None:
"""Given a connection string, create singleton connection
to be used across all instances of `SQLQueryDataset` that
need to connect to the same source.
"""
try:
engine = create_engine(connection_str)
except ImportError as import_error:
raise _get_missing_module_error(import_error) from import_error
except NoSuchModuleError as exc:
raise _get_sql_alchemy_missing_error() from exc
cls.engines[connection_str] = engine
@property
def engine(self):
"""The ``Engine`` object for the dataset's connection string."""
cls = type(self)
if self._connection_str not in cls.engines:
self.create_connection(self._connection_str)
return cls.engines[self._connection_str]
def _describe(self) -> dict[str, Any]:
load_args = copy.deepcopy(self._load_args)
return {
"sql": str(load_args.pop("sql", None)),
"filepath": str(self._filepath),
"load_args": str(load_args),
"execution_options": str(self._execution_options),
}
def _load(self) -> pd.DataFrame:
load_args = copy.deepcopy(self._load_args)
if self._filepath:
load_path = get_filepath_str(PurePosixPath(self._filepath), self._protocol)
with self._fs.open(load_path, mode="r") as fs_file:
load_args["sql"] = fs_file.read()
return pd.read_sql_query(
con=self.engine.execution_options(**self._execution_options), **load_args
)
def _save(self, data: None) -> NoReturn:
raise DatasetError("'save' is not supported on SQLQueryDataset")
# For mssql only
[docs]
def adapt_mssql_date_params(self) -> None:
"""We need to change the format of datetime parameters.
MSSQL expects datetime in the exact format %y-%m-%dT%H:%M:%S.
Here, we also accept plain dates.
`pyodbc` does not accept named parameters, they must be provided as a list."""
params = self._load_args.get("params", [])
if not isinstance(params, list):
raise DatasetError(
"Unrecognized `params` format. It can be only a `list`, "
f"got {type(params)!r}"
)
new_load_args = []
for value in params:
try:
as_date = dt.date.fromisoformat(value)
new_val = dt.datetime.combine(as_date, dt.time.min)
new_load_args.append(new_val.strftime("%Y-%m-%dT%H:%M:%S"))
except (TypeError, ValueError):
new_load_args.append(value)
if new_load_args:
self._load_args["params"] = new_load_args
