A Python library to assist with ETL processing for:
- Amazon Redshift (
COPY
,UNLOAD
) - Snowflake (
COPY INTO <table>
,COPY INTO <location>
)
In addition:
- The library supports Python 3.7 to 3.9
- DB Driver (Adapter) agnostic. Use your favourite driver that complies with DB-API 2.0
- It provides functionality to download and upload data to S3 buckets, and internal stages (Snowflake)
pip install locopy
or install from conda-forge
conda config --add channels conda-forge
conda install locopy
A virtual or conda environment is highly recommended
$ virtualenv locopy
$ source locopy/bin/activate
$ pip install --upgrade setuptools pip
$ pip install locopy
Rather than using a specific Python DB Driver / Adapter for Postgres (which should supports Amazon
Redshift or Snowflake), locopy
prefers to be agnostic. As an end user you can use any Python
Database API Specification 2.0 package.
The following packages have been tested:
psycopg2
pg8000
snowflake-connector-python
You can use which ever one you prefer by importing the package and passing it
into the constructor input dbapi
.
You need to store your connection parameters in a YAML file (or pass them in directly). The YAML would consist of the following items:
# required to connect to redshift
host: my.redshift.cluster.com
port: 5439
database: db
user: userid
password: password
## optional extras for the dbapi connector
sslmode: require
another_option: 123
If you aren't loading data, you don't need to have AWS tokens set up.
The Redshift connection (Redshift
) can be used like this:
import pg8000
import locopy
with locopy.Redshift(dbapi=pg8000, config_yaml="config.yml") as redshift:
redshift.execute("SELECT * FROM schema.table")
df = redshift.to_dataframe()
print(df)
If you want to load data to Redshift via S3, the Redshift
class inherits from S3
:
import pg8000
import locopy
with locopy.Redshift(dbapi=pg8000, config_yaml="config.yml") as redshift:
redshift.execute("SET query_group TO quick")
redshift.execute("CREATE TABLE schema.table (variable VARCHAR(20)) DISTKEY(variable)")
redshift.load_and_copy(
local_file="example/example_data.csv",
s3_bucket="my_s3_bucket",
table_name="schema.table",
delim=",")
redshift.execute("SELECT * FROM schema.table")
res = redshift.cursor.fetchall()
print(res)
If you want to download data from Redshift to a CSV, or read it into Python
my_profile = "some_profile_with_valid_tokens"
with locopy.Redshift(dbapi=pg8000, config_yaml="config.yml", profile=my_profile) as redshift:
##Optionally provide export if you ALSO want the exported data copied to a flat file
redshift.unload_and_copy(
query="SELECT * FROM schema.table",
s3_bucket="my_s3_bucket",
export_path="my_output_destination.csv")
To load data to S3, you will need to be able to generate AWS tokens, or assume the IAM role on a EC2 instance. There are a few options for doing this, depending on where you're running your script and how you want to handle tokens. Once you have your tokens, they need to be accessible to the AWS command line interface. See http://docs.aws.amazon.com/cli/latest/userguide/cli-chap-getting-started.html#config-settings-and-precedence for more information, but you can:
- Populate environment variables
AWS_ACCESS_KEY_ID
,AWS_SECRET_ACCESS_KEY
, etc. - Leverage the AWS credentials file. If you have multiple profiles configured
you can either call
locopy.Redshift(profile="my-profile")
, or set up an environment variableAWS_DEFAULT_PROFILE
. - If you are on a EC2 instance you can assume the credentials associated with the IAM role attached.
See the docs for more detailed usage instructions and examples including Snowflake.
We welcome and appreciate your contributions! Before we can accept any contributions, we ask that you please be sure to sign the Contributor License Agreement (CLA).
This project adheres to the Open Source Code of Conduct. By participating, you are expected to honor this code.
Roadmap details can be found here