recce-nightly 0.1.0.20240115

Environment diff tool for dbt

Recce

Recce is a PR review tool designed for DBT projects. It facilitates the comparison of results between two environments, such as development and production, and helps in identifying any differences.

Features

• Lineage diff
• Schema diff
• Row count diff
• Ad-Hoc Query and Query Diff
• Profile Diff
• Value Diff
• Checklist

Use cases

Recce is primarily designed for PR Review purposes. However, it can be expanded to the following use cases:

1. During development, we can verify new results by contrasting them with those from production prior to pushing the changes.
2. While reviewing PR, you can grasp the extent of the changes and their impact before merging.
3. For troubleshooting, you can execute ad-hoc diff queries to pinpoint the root causes.

Usage

Prerequisites

You have to have at least two environments in your dbt project. For example, one is for development and another is for production. You can prepare two targets with separate schemas in your DBT profile. Here is profiles.yml example

jaffle_shop:
  target: dev
  outputs:
    dev:
      type: duckdb
      path: jaffle_shop.duckdb
      schema: dev
    prod:
      type: duckdb
      path: jaffle_shop.duckdb
      schema: main

Getting Started

5 minutes walkthrough by jaffle shop example

1. Installation

   pip install recce
   

2. Go to your DBT project

   cd your-dbt-project/
   

3. Prepare base artifacts: DBT generates artifacts when every invocation. You can find these files in the target/ folder.

   artifacts	DBT command
   manifest.json	dbt run, dbt build, ..
   catalog.json (optional)	dbt docs generate

   Copy the artifacts for the base environment to target-base/ folder.

4. Run the recce server.

   recce server
   

   Recce would diff environments between target/ and target-base/

Lineage Diff

Ad-Hoc Query and Query Diff

You can use any dbt macros installed in your project.

select * from {{ ref("mymodel") }}

In the query diff, which involves comparing the results from two different environments, the browser is required to pull all result data to the client side. Consequently, minimizing the data volume in the query results is essential for efficiency and performance.

[!TIP] Hotkeys Cmd + Enter: Run query Cmd + Shift + Enter: Run query diff

Schema diff

To use schema diff, make sure that both environments have catalog.json.

Row count diff

Profile diff

Profile diff uses the get_profile from dbt_profiler. Make sure that this package is installed in your project.

packages:
  - package: data-mie/dbt_profiler
    version: <version>

Please reference [dbt-profiler] to understand the definition of each profiling stats.

Value diff

1. Added: New added PKs.
2. Removed: Removed PKs.
3. Matched: For a column, the count of matched value of common PKs.
4. Matched %: For a column, the ratio of matched over common PKs.

PK: Primary key

Value diff uses the compare_column_values from audit-helper. Make sure that this package is installed in your project.

packages:
  - package: dbt-labs/audit_helper
    version: <version>

Value diff requires to select a column as private key. The catalog.json is required to list the available columns.

Checklist

When you feel that your query is worth recording, you can add it to the checklist and name the title and leave a description. This is very helpful for subsequent posting in PR comments.

Q&A

Q: How recce connect to my data warehouse? Does recce support my data warehouse?

recce use the dbt adapter to connect to your warehouse. So it should work for your data warehouse.

Q: What credential does recce connect to the two environments?

Recce uses the same target in the profile to connect your warehouse. If you use the default target dev, it uses the credentials to connect to both environments. So please make sure that the credential can access both environments.