ID Stitching dbt Package

This dbt package stitches together identifiers in an ID graph table.

Documentation: rudderlabs.github.io/id-stitching-docs

Overview

The primary ouput of this package is id_graph. There are a few intermediate models used to create this model.

Model	Description
queries	Generates select statements which pull IDs from your tables.
edges	Combines the results of those select statement to create a table containing edges (IDs) the first time it is run, and matches edges on subsequent runs.
check_edges	Determines if there are still edges to match.
id_graph	Creates an ID graph table.

Installation

Check dbt Hub for the latest installation instructions, or read the docs for more information on installing packages.

Configuration

Set ID columns and IDs to exclude in dbt_project.yml:

vars:
  id-columns: ('anonymous_id', 'user_id', 'email')
  ids-to-exclude: ('sources','user@company.com')

This package searches your data warehouse for tables that include multiple columns defined in id-columns. Any IDs defined in ids-to-exclude are disregarded.

Usage

The edges model must be run enough times to match all edges (IDs). Five or six passes is usually sufficient. The check_edges model will show 0 when all edges have been matched. Edit your job commands for dbt Cloud or run.sh script for dbt CLI to run the edges model however many times is necessary.

dbt Cloud

Create a job with the following commands:

dbt run --full-refresh --select queries edges
dbt run --select edges
dbt run --select edges
dbt run --select edges
dbt run --select edges
dbt run --select edges check_edges id_graph

dbt CLI

Run the included run.sh shell script:

./run.sh

Additional intstrumentation can be created to evaluate the check_edges model to determine programatically whether to run the edges model subsequent times.

License

MIT