-
Notifications
You must be signed in to change notification settings - Fork 901
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Draft PR for issue 1946 - snowflake integration #2029
Changes from 7 commits
0a4c79e
a657d3a
95e272b
8643609
d4976f7
7ccef62
6d22856
da3d463
56d37af
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,8 @@ | ||
"""Provides I/O modules for Snowflake.""" | ||
|
||
__all__ = ["SnowflakeTableDataSet"] | ||
|
||
from contextlib import suppress | ||
|
||
with suppress(ImportError): | ||
from .snowflake_dataset import SnowflakeTableDataSet |
Original file line number | Diff line number | Diff line change | ||||
---|---|---|---|---|---|---|
@@ -0,0 +1,252 @@ | ||||||
"""``SnowflakeTableDataSet`` to load and save data to a snowflake backend.""" | ||||||
|
||||||
import copy | ||||||
import re | ||||||
from typing import Any, Dict, Optional | ||||||
|
||||||
import pandas as pd | ||||||
from snowflake.snowpark import Session | ||||||
|
||||||
from kedro.io.core import AbstractDataSet, DataSetError | ||||||
|
||||||
KNOWN_PIP_INSTALL = { | ||||||
"snowflake.snowpark": "snowflake.snowpark", | ||||||
} | ||||||
|
||||||
DRIVER_ERROR_MESSAGE = """ | ||||||
A module/driver is missing when connecting to Snowflake | ||||||
\n\n | ||||||
""" | ||||||
|
||||||
|
||||||
def _find_known_drivers(module_import_error: ImportError) -> Optional[str]: | ||||||
"""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}") | ||||||
|
||||||
|
||||||
class SnowflakeTableDataSet(AbstractDataSet[pd.DataFrame, pd.DataFrame]): | ||||||
"""`SnowflakeTableDataSet` loads data from a snowflake table and saves a pandas | ||||||
dataframe to a table through snowpark. 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/en/13/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 adding a catalog entry with | ||||||
`YAML API <https://kedro.readthedocs.io/en/stable/data/\ | ||||||
data_catalog.html#use-the-data-catalog-with-the-yaml-api>`_: | ||||||
|
||||||
.. code-block:: yaml | ||||||
|
||||||
>>> shuttles_table_dataset: | ||||||
>>> type: snowflake.SnowflakeTableDataSet | ||||||
>>> 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_creds: | ||||||
|
||||||
Example using Python API: | ||||||
:: | ||||||
|
||||||
>>> from kedro.extras.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 = { | ||||||
>>> "account": "", | ||||||
>>> "user": "", | ||||||
>>> "password": "", | ||||||
>>> "role": "", | ||||||
>>> "warehouse": "", | ||||||
>>> "database": "", | ||||||
>>> "schema": "" | ||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Why are |
||||||
>>> } | ||||||
>>> data_set = SnowflakeTableDataSet(table_name=table_name, | ||||||
>>> credentials=credentials) | ||||||
>>> | ||||||
>>> data_set.save(data) | ||||||
>>> reloaded = data_set.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 | ||||||
sessions: Dict[str, Any] = {} | ||||||
|
||||||
def __init__( | ||||||
self, | ||||||
table_name: str, | ||||||
credentials: Dict[str, Any], | ||||||
load_args: Dict[str, Any] = None, | ||||||
save_args: 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. 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/en/13/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/en/13/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/en/13/core/engines.html#database-urls | ||||||
It has ``index=False`` in the default parameters. | ||||||
|
||||||
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: | ||||||
raise DataSetError("Please configure expected credentials") | ||||||
|
||||||
# print(self._load_args) | ||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more.
Suggested change
|
||||||
|
||||||
# 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._credentials = credentials["credentials"] | ||||||
|
||||||
# self._connection_str = credentials["con"] | ||||||
self._session = self._get_session(self._credentials) | ||||||
|
||||||
@classmethod | ||||||
def _get_session(cls, credentials: dict) -> None: | ||||||
"""Given a connection string, create singleton connection | ||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. We did something similar in pandas.SQL*DataSet - is this the same pattern? There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Yes, that's the same pattern as the SQLDataSet, this would change when implementing a session hook, similar to the SparkDataSet |
||||||
to be used across all instances of `SQLQueryDataSet` that | ||||||
need to connect to the same source. | ||||||
connection_params = { | ||||||
"account": "", | ||||||
"user": "", | ||||||
"password": "", | ||||||
"role": "", | ||||||
"warehouse": "", | ||||||
"database": "", | ||||||
"schema": "" | ||||||
} | ||||||
""" | ||||||
try: | ||||||
session = Session.builder.configs(credentials).create() | ||||||
except ImportError as import_error: | ||||||
raise _get_missing_module_error(import_error) from import_error | ||||||
except Exception as exception: | ||||||
raise exception | ||||||
return session | ||||||
|
||||||
def _describe(self) -> Dict[str, Any]: | ||||||
load_args = copy.deepcopy(self._load_args) | ||||||
save_args = copy.deepcopy(self._save_args) | ||||||
return dict( | ||||||
table_name=self._load_args["table_name"], | ||||||
load_args=load_args, | ||||||
save_args=save_args, | ||||||
) | ||||||
|
||||||
def _load(self) -> pd.DataFrame: | ||||||
sp_df = self._session.table(self._load_args["table_name"]) | ||||||
return sp_df.to_pandas() | ||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I don't think we want to return as a Pandas, I would return a SnowPark DataFrame and let the user do the pandas casting themselves. There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. ➕ to that, Snowpark DataFrames are lazy, so the user node could potentially leverage that. |
||||||
|
||||||
def _save(self, data: pd.DataFrame) -> None: | ||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more.
Suggested change
I'm not actually sure on the type signature, but the push here is to accept either option and handle gracefully. There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more.
|
||||||
# pd df to snowpark df | ||||||
sp_df = self._session.create_dataframe(data) | ||||||
table_name = [ | ||||||
self._credentials.get("database"), | ||||||
self._credentials.get("schema"), | ||||||
self._save_args["name"], | ||||||
] | ||||||
sp_df.write.mode(self._save_args["mode"]).save_as_table( | ||||||
table_name, | ||||||
column_order=self._save_args["column_order"], | ||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Couldn't this be inferred from the passed dataframe and only fallback to |
||||||
table_type=self._save_args["table_type"], | ||||||
) | ||||||
|
||||||
def _exists(self) -> bool: | ||||||
session = self.sessions[self._credentials["account"]] # type: ignore | ||||||
schema = self._load_args.get("schema", None) | ||||||
exists = self._load_args["table_name"] in session.table_names(schema) | ||||||
return exists |
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -37,7 +37,7 @@ Pillow~=9.0 | |
plotly>=4.8.0, <6.0 | ||
pre-commit>=2.9.2, <3.0 # The hook `mypy` requires pre-commit version 2.9.2. | ||
psutil==5.8.0 | ||
pyarrow>=1.0, <7.0 | ||
pyarrow>=1.0, <9.0 | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Hopefully this doesn't break any other bits of Kedro! There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Hopefully not! But found this dependency update required for the snowpark df.to_pandas() method to work. |
||
pylint>=2.5.2, <3.0 | ||
pyproj~=3.0 | ||
pyspark>=2.2, <4.0 | ||
|
@@ -50,6 +50,7 @@ requests-mock~=1.6 | |
requests~=2.20 | ||
s3fs>=0.3.0, <0.5 # Needs to be at least 0.3.0 to make use of `cachable` attribute on S3FileSystem. | ||
SQLAlchemy~=1.2 | ||
snowflake-snowpark-python~=0.12.0 | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. 1.0.0 came out 1st November |
||
tables~=3.6.0; platform_system == "Windows" and python_version<'3.9' | ||
tables~=3.6; platform_system != "Windows" | ||
tensorflow~=2.0 | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we want to raise an error here?