# -*- coding: utf-8 -*-
#
# Copyright (c) 2015 Gouthaman Balaraman
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may not
# use this file except in compliance with the License. You may obtain a copy of
# the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations under
# the License.
#
"""
Support for SQLAlchemy. Provides SQLAlchemyTarget for storing in databases
supported by SQLAlchemy. The user would be responsible for installing the
required database driver to connect using SQLAlchemy.
Minimal example of a job to copy data to database using SQLAlchemy is as shown
below:
.. code-block:: python
from sqlalchemy import String
import luigi
from luigi.contrib import sqla
class SQLATask(sqla.CopyToTable):
# columns defines the table schema, with each element corresponding
# to a column in the format (args, kwargs) which will be sent to
# the sqlalchemy.Column(*args, **kwargs)
columns = [
(["item", String(64)], {"primary_key": True}),
(["property", String(64)], {})
]
connection_string = "sqlite://" # in memory SQLite database
table = "item_property" # name of the table to store data
def rows(self):
for row in [("item1", "property1"), ("item2", "property2")]:
yield row
if __name__ == '__main__':
task = SQLATask()
luigi.build([task], local_scheduler=True)
If the target table where the data needs to be copied already exists, then
the column schema definition can be skipped and instead the reflect flag
can be set as True. Here is a modified version of the above example:
.. code-block:: python
from sqlalchemy import String
import luigi
from luigi.contrib import sqla
class SQLATask(sqla.CopyToTable):
# If database table is already created, then the schema can be loaded
# by setting the reflect flag to True
reflect = True
connection_string = "sqlite://" # in memory SQLite database
table = "item_property" # name of the table to store data
def rows(self):
for row in [("item1", "property1"), ("item2", "property2")]:
yield row
if __name__ == '__main__':
task = SQLATask()
luigi.build([task], local_scheduler=True)
In the above examples, the data that needs to be copied was directly provided by
overriding the rows method. Alternately, if the data comes from another task, the
modified example would look as shown below:
.. code-block:: python
from sqlalchemy import String
import luigi
from luigi.contrib import sqla
from luigi.mock import MockTarget
class BaseTask(luigi.Task):
def output(self):
return MockTarget("BaseTask")
def run(self):
out = self.output().open("w")
TASK_LIST = ["item%d\\tproperty%d\\n" % (i, i) for i in range(10)]
for task in TASK_LIST:
out.write(task)
out.close()
class SQLATask(sqla.CopyToTable):
# columns defines the table schema, with each element corresponding
# to a column in the format (args, kwargs) which will be sent to
# the sqlalchemy.Column(*args, **kwargs)
columns = [
(["item", String(64)], {"primary_key": True}),
(["property", String(64)], {})
]
connection_string = "sqlite://" # in memory SQLite database
table = "item_property" # name of the table to store data
def requires(self):
return BaseTask()
if __name__ == '__main__':
task1, task2 = SQLATask(), BaseTask()
luigi.build([task1, task2], local_scheduler=True)
In the above example, the output from `BaseTask` is copied into the
database. Here we did not have to implement the `rows` method because
by default `rows` implementation assumes every line is a row with
column values separated by a tab. One can define `column_separator`
option for the task if the values are say comma separated instead of
tab separated.
You can pass in database specific connection arguments by setting the connect_args
dictionary. The options will be passed directly to the DBAPI's connect method as
keyword arguments.
The other option to `sqla.CopyToTable` that can be of help with performance aspect is the
`chunk_size`. The default is 5000. This is the number of rows that will be inserted in
a transaction at a time. Depending on the size of the inserts, this value can be tuned
for performance.
See here for a `tutorial on building task pipelines using luigi
<http://gouthamanbalaraman.com/blog/building-luigi-task-pipeline.html>`_ and
using `SQLAlchemy in workflow pipelines <http://gouthamanbalaraman.com/blog/sqlalchemy-luigi-workflow-pipeline.html>`_.
Author: Gouthaman Balaraman
Date: 01/02/2015
"""
import abc
import collections
import datetime
import itertools
import logging
import luigi
import os
import sqlalchemy
[docs]
class SQLAlchemyTarget(luigi.Target):
"""
Database target using SQLAlchemy.
This will rarely have to be directly instantiated by the user.
Typical usage would be to override `luigi.contrib.sqla.CopyToTable` class
to create a task to write to the database.
"""
marker_table = None
_engine_dict = {} # dict of sqlalchemy engine instances
Connection = collections.namedtuple("Connection", "engine pid")
def __init__(self, connection_string, target_table, update_id, echo=False, connect_args=None):
"""
Constructor for the SQLAlchemyTarget.
:param connection_string: SQLAlchemy connection string
:type connection_string: str
:param target_table: The table name for the data
:type target_table: str
:param update_id: An identifier for this data set
:type update_id: str
:param echo: Flag to setup SQLAlchemy logging
:type echo: bool
:param connect_args: A dictionary of connection arguments
:type connect_args: dict
:return:
"""
if connect_args is None:
connect_args = {}
self.target_table = target_table
self.update_id = update_id
self.connection_string = connection_string
self.echo = echo
self.connect_args = connect_args
self.marker_table_bound = None
def __str__(self):
return self.target_table
@property
def engine(self):
"""
Return an engine instance, creating it if it doesn't exist.
Recreate the engine connection if it wasn't originally created
by the current process.
"""
pid = os.getpid()
conn = SQLAlchemyTarget._engine_dict.get(self.connection_string)
if not conn or conn.pid != pid:
# create and reset connection
engine = sqlalchemy.create_engine(
self.connection_string,
connect_args=self.connect_args,
echo=self.echo
)
SQLAlchemyTarget._engine_dict[self.connection_string] = self.Connection(engine, pid)
return SQLAlchemyTarget._engine_dict[self.connection_string].engine
[docs]
def touch(self):
"""
Mark this update as complete.
"""
if self.marker_table_bound is None:
self.create_marker_table()
table = self.marker_table_bound
id_exists = self.exists()
with self.engine.begin() as conn:
if not id_exists:
ins = table.insert().values(update_id=self.update_id, target_table=self.target_table,
inserted=datetime.datetime.now())
else:
ins = table.update().where(sqlalchemy.and_(table.c.update_id == self.update_id,
table.c.target_table == self.target_table)).\
values(update_id=self.update_id, target_table=self.target_table,
inserted=datetime.datetime.now())
conn.execute(ins)
assert self.exists()
[docs]
def exists(self):
row = None
if self.marker_table_bound is None:
self.create_marker_table()
with self.engine.begin() as conn:
table = self.marker_table_bound
s = sqlalchemy.select([table]).where(sqlalchemy.and_(table.c.update_id == self.update_id,
table.c.target_table == self.target_table)).limit(1)
row = conn.execute(s).fetchone()
return row is not None
[docs]
def create_marker_table(self):
"""
Create marker table if it doesn't exist.
Using a separate connection since the transaction might have to be reset.
"""
if self.marker_table is None:
self.marker_table = luigi.configuration.get_config().get('sqlalchemy', 'marker-table', 'table_updates')
engine = self.engine
with engine.begin() as con:
metadata = sqlalchemy.MetaData()
if not con.dialect.has_table(con, self.marker_table):
self.marker_table_bound = sqlalchemy.Table(
self.marker_table, metadata,
sqlalchemy.Column("update_id", sqlalchemy.String(128), primary_key=True),
sqlalchemy.Column("target_table", sqlalchemy.String(128)),
sqlalchemy.Column("inserted", sqlalchemy.DateTime, default=datetime.datetime.now()))
metadata.create_all(engine)
else:
metadata.reflect(only=[self.marker_table], bind=engine)
self.marker_table_bound = metadata.tables[self.marker_table]
[docs]
def open(self, mode):
raise NotImplementedError("Cannot open() SQLAlchemyTarget")
[docs]
class CopyToTable(luigi.Task):
"""
An abstract task for inserting a data set into SQLAlchemy RDBMS
Usage:
* subclass and override the required `connection_string`, `table` and `columns` attributes.
* optionally override the `schema` attribute to use a different schema for
the target table.
"""
_logger = logging.getLogger('luigi-interface')
echo = False
connect_args = {}
@property
@abc.abstractmethod
def connection_string(self):
return None
@property
@abc.abstractmethod
def table(self):
return None
# specify the columns that define the schema. The format for the columns is a list
# of tuples. For example :
# columns = [
# (["id", sqlalchemy.Integer], dict(primary_key=True)),
# (["name", sqlalchemy.String(64)], {}),
# (["value", sqlalchemy.String(64)], {})
# ]
# The tuple (args_list, kwargs_dict) here is the args and kwargs
# that need to be passed to sqlalchemy.Column(*args, **kwargs).
# If the tables have already been setup by another process, then you can
# completely ignore the columns. Instead set the reflect value to True below
columns = []
# Specify the database schema of the target table, if supported by the
# RDBMS. Note that this doesn't change the schema of the marker table.
# The schema MUST already exist in the database, or this will task fail.
schema = ''
# options
column_separator = "\t" # how columns are separated in the file copied into postgres
chunk_size = 5000 # default chunk size for insert
reflect = False # Set this to true only if the table has already been created by alternate means
[docs]
def create_table(self, engine):
"""
Override to provide code for creating the target table.
By default it will be created using types specified in columns.
If the table exists, then it binds to the existing table.
If overridden, use the provided connection object for setting up the table in order to
create the table and insert data using the same transaction.
:param engine: The sqlalchemy engine instance
:type engine: object
"""
def construct_sqla_columns(columns):
retval = [sqlalchemy.Column(*c[0], **c[1]) for c in columns]
return retval
needs_setup = (len(self.columns) == 0) or (False in [len(c) == 2 for c in self.columns]) if not self.reflect else False
if needs_setup:
# only names of columns specified, no types
raise NotImplementedError("create_table() not implemented for %r and columns types not specified" % self.table)
else:
# if columns is specified as (name, type) tuples
with engine.begin() as con:
if self.schema:
metadata = sqlalchemy.MetaData(schema=self.schema)
else:
metadata = sqlalchemy.MetaData()
try:
if not con.dialect.has_table(con, self.table, self.schema or None):
sqla_columns = construct_sqla_columns(self.columns)
self.table_bound = sqlalchemy.Table(self.table, metadata, *sqla_columns)
metadata.create_all(engine)
else:
full_table = '.'.join([self.schema, self.table]) if self.schema else self.table
metadata.reflect(only=[self.table], bind=engine)
self.table_bound = metadata.tables[full_table]
except Exception as e:
self._logger.exception(self.table + str(e))
[docs]
def update_id(self):
"""
This update id will be a unique identifier for this insert on this table.
"""
return self.task_id
[docs]
def output(self):
return SQLAlchemyTarget(
connection_string=self.connection_string,
target_table=self.table,
update_id=self.update_id(),
connect_args=self.connect_args,
echo=self.echo)
[docs]
def rows(self):
"""
Return/yield tuples or lists corresponding to each row to be inserted.
This method can be overridden for custom file types or formats.
"""
with self.input().open('r') as fobj:
for line in fobj:
yield line.strip("\n").split(self.column_separator)
[docs]
def run(self):
self._logger.info("Running task copy to table for update id %s for table %s" % (self.update_id(), self.table))
output = self.output()
engine = output.engine
self.create_table(engine)
with engine.begin() as conn:
rows = iter(self.rows())
ins_rows = [dict(zip(("_" + c.key for c in self.table_bound.c), row))
for row in itertools.islice(rows, self.chunk_size)]
while ins_rows:
self.copy(conn, ins_rows, self.table_bound)
ins_rows = [dict(zip(("_" + c.key for c in self.table_bound.c), row))
for row in itertools.islice(rows, self.chunk_size)]
self._logger.info("Finished inserting %d rows into SQLAlchemy target" % len(ins_rows))
output.touch()
self._logger.info("Finished inserting rows into SQLAlchemy target")
[docs]
def copy(self, conn, ins_rows, table_bound):
"""
This method does the actual insertion of the rows of data given by ins_rows into the
database. A task that needs row updates instead of insertions should overload this method.
:param conn: The sqlalchemy connection object
:param ins_rows: The dictionary of rows with the keys in the format _<column_name>. For example
if you have a table with a column name "property", then the key in the dictionary
would be "_property". This format is consistent with the bindparam usage in sqlalchemy.
:param table_bound: The object referring to the table
:return:
"""
bound_cols = dict((c, sqlalchemy.bindparam("_" + c.key)) for c in table_bound.columns)
ins = table_bound.insert().values(bound_cols)
conn.execute(ins, ins_rows)
