snowchange is a simple python based tool to manage all of your Snowflake objects. It follows an Imperative-style approach to Database Change Management (DCM) and was inspired by the Flyway database migration tool. When combined with a version control system and a CI/CD tool, database changes can be approved and deployed through a pipeline using modern software delivery practices. As such snowchange plays a critical role in enabling Database (or Data) DevOps.
DCM tools (also known as Database Migration, Schema Change Management, or Schema Migration tools) follow one of two approaches: Declarative or Imperative. For a background on Database DevOps, including a discussion on the differences between the Declarative and Imperative approaches, please read the Embracing Agile Software Delivery and DevOps with Snowflake blog post.
For the complete list of changes made to snowchange check out the CHANGELOG.
- Overview
- Project Structure
- Change Scripts
- Change History Table
- Running snowchange
- Getting Started with snowchange
- Integrating With DevOps
- Maintainers
- Legal
snowchange expects a directory structure like the following to exist:
(project_root)
|
|-- folder_1
|-- V1.1.1__first_change.sql
|-- V1.1.2__second_change.sql
|-- R__sp_add_sales.sql
|-- R__fn_get_timezone.sql
|-- folder_2
|-- folder_3
|-- V1.1.3__third_change.sql
|-- R__fn_sort_ascii.sql
The snowchange folder structure is very flexible. The project_root folder is specified with the -f or --root-folder argument. Under the project_root folder you are free to arrange the change scripts any way you see fit. You can have as many subfolders (and nested subfolders) as you would like.
Versioned change scripts follow a similar naming convention to that used by Flyway Versioned Migrations. The script name must follow this pattern (image taken from Flyway docs):
With the following rules for each part of the filename:
- Prefix: The letter 'V' for versioned change
- Version: A unique version number with dots or underscores separating as many number parts as you like
- Separator: __ (two underscores)
- Description: An arbitrary description with words separated by underscores or spaces (can not include two underscores)
- Suffix: .sql
For example, a script name that follows this convention is: V1.1.1__first_change.sql. As with Flyway, the unique version string is very flexible. You just need to be consistent and always use the same convention, like 3 sets of numbers separated by periods. Here are a few valid version strings:
- 1
- 5.2
- 5_2
- 1.2.3.4.5.6.7.8.9
- 205_68
- 20200115113556
- 2020.1.15.11.35.56
Every script within a database folder must have a unique version number. snowchange will check for duplicate version numbers and throw an error if it finds any. This helps to ensure that developers who are working in parallel don't accidently (re-)use the same version number.
Repeatable change scripts follow a similar naming convention to that used by Flyway Versioned Migrations. The script name must follow this pattern (image taken from Flyway docs:
e.g:
- R__sp_add_sales.sql
- R__fn_get_timezone.sql
- R__fn_sort_ascii.sql
All repeatable change scripts are applied each time the utility is run, irrespective of the most recent change in the database. Repeatable scripts could be used for maintaining code that always needs to be applied in its entirety. e.g. stores procedures, functions and view definitions etc.
Just like Flyway, within a single migration run, repeatable scripts are always applied last, after all pending versioned scripts have been executed. Repeatable scripts are applied in the order of their description.
snowchange is designed to be very lightweight and not impose to many limitations. Each change script can have any number of SQL statements within it and must supply the necessary context, like database and schema names. The context can be supplied by using an explicit USE <DATABASE> command or by naming all objects with a three-part name (<database name>.<schema name>.<object name>). snowchange will simply run the contents of each script against the target Snowflake account, in the correct order.
snowchange supports a light weight variable replacement strategy. One important use of variables is to support multiple environments (dev, test, prod) in a single Snowflake account by dynamically changing the database name during deployment.
To use a variable in a change script, use this syntax anywhere in the script: {{ variable1 }}. So the pattern is two left curly braces, followed by a space, followed by the variable name, followed by a space, and finally followed by two right curly braces. And the spaces are important. The format for including variables in change scripts mimics Jinja expressions. Please note that at this point snowchange hasn't been integrated with Jinja, but by using the same syntax for variables and expressions a future migration will be seamless.
To pass variables to snowchange, use the --vars command line parameter like this: --vars '{"variable1": "value", "variable2": "value2"}'. This parameter accepts a flat JSON object formatted as a string. Nested objects and arrays don't make sense at this point and aren't supported.
snowchange will replace any variable placeholders before running your change script code and will throw an error if it finds any variable placeholders that haven't been replaced.
snowchange records all applied changes scripts to the change history table. By default snowchange will attempt to log all activities to the METADATA.SNOWCHANGE.CHANGE_HISTORY table. The name and location of the change history table can be overriden by using the -c (or --change-history-table) parameter. The value passed to the parameter can have a one, two, or three part name (e.g. "TABLE_NAME", or "SCHEMA_NAME.TABLE_NAME", or "DATABASE_NAME.SCHEMA_NAME.TABLE_NAME"). This can be used to support multiple environments (dev, test, prod) or multiple subject areas within the same Snowflake account. By default snowchange will not try to create the change history table, and will fail if the table does not exist.
Additionally, if the --create-change-history-table parameter is given, then snowchange will attempt to create the schema and table associated with the change history table. snowchange will not attempt to create the database for the change history table, so that must be created ahead of time, even when using the --create-change-history-table parameter.
The structure of the CHANGE_HISTORY table is as follows:
| Column Name | Type | Example |
|---|---|---|
| VERSION | VARCHAR | 1.1.1 |
| DESCRIPTION | VARCHAR | First change |
| SCRIPT | VARCHAR | V1.1.1__first_change.sql |
| SCRIPT_TYPE | VARCHAR | V |
| CHECKSUM | VARCHAR | 38e5ba03b1a6d2... |
| EXECUTION_TIME | NUMBER | 4 |
| STATUS | VARCHAR | Success |
| INSTALLED_BY | VARCHAR | SNOWFLAKE_USER |
| INSTALLED_ON | TIMESTAMP_LTZ | 2020-03-17 12:54:33.056 -0700 |
A new row will be added to this table every time a change script has been applied to the database. snowchange will use this table to identify which changes have been applied to the database and will not apply the same version more than once.
Here is the current schema DDL for the change history table (found in the snowchange/cli.py script), in case you choose to create it manually and not use the --create-change-history-table parameter:
CREATE TABLE IF NOT EXISTS SNOWCHANGE.CHANGE_HISTORY
(
VERSION VARCHAR
,DESCRIPTION VARCHAR
,SCRIPT VARCHAR
,SCRIPT_TYPE VARCHAR
,CHECKSUM VARCHAR
,EXECUTION_TIME NUMBER
,STATUS VARCHAR
,INSTALLED_BY VARCHAR
,INSTALLED_ON TIMESTAMP_LTZ
)In order to run snowchange you must have the following:
- You will need to have a recent version of python 3 installed (not 3.9.x as of 1/19/21)
- You will need to have the latest Snowflake Python driver installed
- You will need to create the change history table used by snowchange in Snowflake (see Change History Table above for more details)
- First, you will need to create a database to store your change history table (snowchange will not help you with this)
- Second, you will need to create the change history schema and table. You can do this manually (see Change History Table above for the DDL) or have snowchange create them by running it with the
--create-change-history-tableparameter (just make sure the Snowflake user you're running snowchange with has privileges to create a schema and table in that database)
- You will need to create (or choose) a user account that has privileges to apply the changes in your change script
- Don't forget that this user also needs the SELECT and INSERT privileges on the change history table
snowchange is a single python script located at snowchange/cli.py. It can be executed as follows:
python snowchange/cli.py [-h] [-f ROOT_FOLDER] -a SNOWFLAKE_ACCOUNT -u SNOWFLAKE_USER -r SNOWFLAKE_ROLE -w SNOWFLAKE_WAREHOUSE [-d SNOWFLAKE_DATABASE] [-c CHANGE_HISTORY_TABLE] [--vars VARS] [--create-change-history-table] [-ac] [-v]
Or if installed via pip, it can be executed as follows:
snowchange [-h] [-f ROOT_FOLDER] -a SNOWFLAKE_ACCOUNT -u SNOWFLAKE_USER -r SNOWFLAKE_ROLE -w SNOWFLAKE_WAREHOUSE [-d SNOWFLAKE_DATABASE] [-c CHANGE_HISTORY_TABLE] [--vars VARS] [--create-change-history-table] [-ac] [-v]
snowchange supports both password authentication and private key authentication.
In the event both authentication criteria are provided, snowchange will prioritize password authentication.
The Snowflake user password for SNOWFLAKE_USER is required to be set in the environment variable SNOWFLAKE_PASSWORD prior to calling the script. snowchange will fail if the SNOWFLAKE_PASSWORD environment variable is not set.
DEPRECATION NOTICE: The SNOWSQL_PWD environment variable is deprecated but currently still supported. Support for it will be removed in a later version of snowchange. Please use SNOWFLAKE_PASSWORD instead.
The Snowflake user encrypted private key for SNOWFLAKE_USER is required to be in a file with the file path set in the environment variable SNOWFLAKE_PRIVATE_KEY_PATH. Additionally, the password for the encrypted private key file is required to be set in the environment variable SNOWFLAKE_PRIVATE_KEY_PASSPHRASE. These two environment variables must be set prior to calling the script. snowchange will fail if the SNOWFLAKE_PRIVATE_KEY_PATH and SNOWFLAKE_PRIVATE_KEY_PASSPHRASE environment variables are not set.
Here is the list of supported parameters to the script:
| Parameter | Description |
|---|---|
| -h, --help | Show the help message and exit |
| -f ROOT_FOLDER, --root-folder ROOT_FOLDER | (Optional) The root folder for the database change scripts. The default is the current directory. |
| -a SNOWFLAKE_ACCOUNT, --snowflake-account SNOWFLAKE_ACCOUNT | The name of the snowflake account (e.g. xy12345.east-us-2.azure). See Usage Notes for the account Parameter (for the connect Method) for more details on how to structure the account name. |
| -u SNOWFLAKE_USER, --snowflake-user SNOWFLAKE_USER | The name of the snowflake user |
| -r SNOWFLAKE_ROLE, --snowflake-role SNOWFLAKE_ROLE | The name of the role to use |
| -w SNOWFLAKE_WAREHOUSE, --snowflake-warehouse SNOWFLAKE_WAREHOUSE | The name of the default warehouse to use. Can be overridden in the change scripts. |
| -d SNOWFLAKE_DATABASE, --snowflake-database SNOWFLAKE_DATABASE | The name of the default database to use. Can be overridden in the change scripts. |
| -c CHANGE_HISTORY_TABLE, --change-history-table CHANGE_HISTORY_TABLE | (Optional) Used to override the default name of the change history table (which is METADATA.SNOWCHANGE.CHANGE_HISTORY) |
| --vars VARS | (Optional) Define values for the variables to replaced in change scripts, given in JSON format (e.g. '{"variable1": "value1", "variable2": "value2"}') |
| --create-change-history-table | (Optional) Create the change history table if it does not exist. The default is 'False'. |
| -ac, --autocommit | (Optional) Enable autocommit feature for DML commands. The default is 'False'. |
| -v, --verbose | (Optional) Display verbose debugging details during execution. The default is 'False'. |
| --dry-run | (Optional) Run snowchange in dry run mode. the default is 'False'. |
The demo folder in this project repository contains a snowchange demo project for you to try out. This demo is based on the standard Snowflake Citibike demo which can be found in the Snowflake Hands-on Lab. It contains the following database change scripts:
| Change Script | Description |
|---|---|
| v1.1__initial_database_objects.sql | Create the initial Citibike demo objects including file formats, stages, and tables. |
| v1.2__load_tables_from_s3.sql | Load the Citibike and weather data from the Snowlake lab S3 bucket. |
The Citibike data for this demo comes from the NYC Citi Bike bike share program.
To get started with snowchange and these demo Citibike scripts follow these steps:
- Make sure you've completed the Prerequisites steps above
- Get a copy of this snowchange repository (either via a clone or download)
- Open a shell and change directory to your copy of the snowchange repository
- Run snowchange (see Running the Script above) with your Snowflake account details and the
demo/citibikefolder as the root folder (make sure you use the full path)
Here is a sample DevOps development lifecycle with snowchange:
If your build agent has a recent version of python 3 installed, the script can be ran like so:
pip install snowchange --upgrade
snowchange [-h] [-f ROOT_FOLDER] -a SNOWFLAKE_ACCOUNT -u SNOWFLAKE_USER -r SNOWFLAKE_ROLE -w SNOWFLAKE_WAREHOUSE [-d SNOWFLAKE_DATABASE] [-c CHANGE_HISTORY_TABLE] [--vars VARS] [--create-change-history-table] [-ac] [-v]
Or if you prefer docker, set the environment variables and run like so:
docker run -it --rm \
--name snowchange-script \
-v "$PWD":/usr/src/snowchange \
-w /usr/src/snowchange \
-e ROOT_FOLDER \
-e SNOWFLAKE_ACCOUNT \
-e SNOWFLAKE_USER \
-e SNOWFLAKE_ROLE \
-e SNOWFLAKE_WAREHOUSE \
-e SNOWFLAKE_PASSWORD \
python:3 /bin/bash -c "pip install snowchange --upgrade && snowchange -f $ROOT_FOLDER -a $SNOWFLAKE_ACCOUNT -u $SNOWFLAKE_USER -r $SNOWFLAKE_ROLE -w $SNOWFLAKE_WAREHOUSE"
Either way, don't forget to set the SNOWFLAKE_PASSWORD environment variable if using password authentication!
- James Weakley (@jamesweakley)
- Jeremiah Hansen (@jeremiahhansen)
This is a community-developed script, not an official Snowflake offering. It comes with no support or warranty. However, feel free to raise a github issue if you find a bug or would like a new feature.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this tool 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.