snowflake-ddl-extractor

A Python CLI tool that connects to a Snowflake database, extracts DDL scripts for all objects using GET_DDL(), and saves them into a structured local directory tree mirroring the database hierarchy.

What It Does

• Connects to any Snowflake database with four authentication methods
• Extracts DDL for tables, views, procedures, functions, tasks, streams, stages, file formats, sequences, and pipes
• Writes each DDL to a .sql file organized by schema and object type
• Schema-qualifies all object names in the DDL (e.g., CREATE TABLE SCHEMA.MY_TABLE)
• Strips the database name from all references inside scripts so DDLs are portable
• Handles overloaded procedures/functions (same name, different signatures) without data loss
• Runs idempotent refreshes: overwrites existing files, removes stale ones, prunes empty directories
• Extracts concurrently using a thread pool for speed

Installation

pip install snowflake-ddl-extractor

Or install from source:

git clone https://github.com/imsaurav/snowflake-ddl-extractor.git
cd snowflake-ddl-extractor
pip install -e .

Quick Start

Password Authentication

export SNOWFLAKE_PASSWORD='my-secret'
snow-extract --account myaccount --user myuser --db MY_DATABASE --out-dir ./ddl_output

SSO (Browser) Authentication

snow-extract --account myaccount --user myuser --db MY_DATABASE --auth sso --out-dir ./ddl_output

Key-Pair Authentication

export SNOWFLAKE_PRIVATE_KEY_PATH=~/.ssh/snowflake_rsa_key.p8
snow-extract --account myaccount --user myuser --db MY_DATABASE --auth keypair --out-dir ./ddl_output

OAuth Authentication

export SNOWFLAKE_OAUTH_TOKEN='eyJhbGci...'
snow-extract --account myaccount --user myuser --db MY_DATABASE --auth oauth --out-dir ./ddl_output

You can also run it as a Python module:

python -m snow_ddl_extractor --account myaccount --user myuser --db MY_DATABASE --auth sso

CLI Reference

snow-extract [OPTIONS]

Options:
  -a, --account TEXT              Snowflake account identifier  [required]
  -u, --user TEXT                 Snowflake login user  [required]
      --database, --db TEXT       Target database  [required]
  -o, --out-dir DIRECTORY         Root output directory  [default: .]
      --auth [password|keypair|sso|oauth]
                                  Authentication method  [default: password]
      --password TEXT             Snowflake password
      --private-key-path TEXT     Path to PEM private key file
      --private-key-passphrase TEXT
                                  Passphrase for the private key
      --oauth-token TEXT          OAuth access token
      --role TEXT                 Snowflake role
      --warehouse TEXT            Snowflake warehouse
      --include-schemas TEXT      Comma-separated schemas to include
      --exclude-schemas TEXT      Comma-separated schemas to exclude
      --include-types TEXT        Comma-separated object types to include
      --exclude-types TEXT        Comma-separated object types to exclude
  -w, --workers INTEGER           Concurrent threads  [default: 10]
      --dry-run                   Preview what would be extracted, no files written
  -v, --verbose                   Enable debug logging
      --version                   Show version and exit
  -h, --help                      Show this message and exit

Output Structure

{out_dir}/{DATABASE}/
├── {SCHEMA_A}/
│   ├── TABLES/
│   │   ├── CUSTOMERS.sql
│   │   └── ORDERS.sql
│   ├── VIEWS/
│   │   └── V_DAILY_SALES.sql
│   ├── PROCEDURES/
│   │   ├── LOAD_DATA.sql
│   │   └── LOAD_DATA(VARCHAR,NUMBER).sql    <- overloaded variant
│   ├── FUNCTIONS/
│   ├── TASKS/
│   ├── STREAMS/
│   ├── STAGES/
│   ├── FILE_FORMATS/
│   ├── SEQUENCES/
│   └── PIPES/
└── {SCHEMA_B}/
    └── ...

All folder names are uppercase. Filenames match the Snowflake object name. Overloaded procedures/functions include the argument signature in the filename to avoid collisions.

DDL Processing

Each extracted DDL goes through two transformations before being written:

1. Database prefix removal -- All occurrences of the database name are stripped from the script. For example, DEV_DB.SCHEMA.MY_TABLE becomes SCHEMA.MY_TABLE. This makes the DDLs portable across environments (dev/uat/prod) without find-and-replace.

2. Schema qualification -- The object name in the CREATE statement is prefixed with its schema name. For example, CREATE TABLE MY_TABLE becomes CREATE TABLE SCHEMA_A.MY_TABLE. This ensures every script is self-contained and can be executed without setting a schema context first.

Filtering

By Schema

Extract only specific schemas:

snow-extract --account myacc --user myuser --db MY_DB --auth sso \
  --include-schemas ANALYTICS,STAGING

Exclude schemas:

snow-extract --account myacc --user myuser --db MY_DB --auth sso \
  --exclude-schemas PUBLIC

INFORMATION_SCHEMA is always excluded automatically.

By Object Type

Extract only tables and views:

snow-extract --account myacc --user myuser --db MY_DB --auth sso \
  --include-types TABLES,VIEWS

Extract everything except stages and pipes:

snow-extract --account myacc --user myuser --db MY_DB --auth sso \
  --exclude-types STAGES,PIPES

Valid type names: TABLES, VIEWS, PROCEDURES, FUNCTIONS, TASKS, STREAMS, STAGES, FILE_FORMATS, SEQUENCES, PIPES (case-insensitive).

Dry Run

Preview what would be extracted without writing any files:

snow-extract --account myacc --user myuser --db MY_DB --auth sso --dry-run

This discovers all objects, prints a summary table, and exits.

Progress Bar and Summary Table

During extraction, a progress bar shows real-time status:

Extracting DDLs  [################----]  80%

After extraction, a summary table breaks down counts by schema and object type:

  Schema              TABLES  VIEWS  PROCEDURES  total
  ----------------------------------------------------
  ANALYTICS               10      5           3     18
  STAGING                  8      2           1     11
  ----------------------------------------------------
  TOTAL                   18      7           4     29

Idempotent Refresh

On subsequent runs the tool:

1. Overwrites existing .sql files with the latest DDL
2. Removes .sql files for objects that no longer exist in Snowflake
3. Prunes empty directories automatically

This keeps the local directory tree an exact mirror of the database.

Environment Variables

All connection options can be set via environment variables to avoid passing secrets on the command line:

Supported Object Types

Dependencies

• snowflake-connector-python >= 3.6.0
• click >= 8.0
• cryptography >= 41.0

License

MIT