Slim CI with state comparison in dbt

Introduction

Slim CI helps run only the parts of your data project that changed. It saves time and computing power by comparing the current state with the last run.

When you want faster testing of your data models after small changes.

When your data project is large and full runs take too long.

When you want to avoid running unchanged models to save resources.

When you want to catch errors quickly during development.

When you want to keep your continuous integration (CI) process efficient.

Syntax

dbt

dbt build --state path/to/previous/run --select state:modified

--state points to the folder with the previous run's manifest and run results.

--select state:modified runs only models that changed since the last run.

Examples

Runs only models that changed compared to the previous run stored in ./target/previous_run.

dbt

dbt build --state ./target/previous_run --select state:modified

Runs tests only on models that were modified since the last run.

dbt

dbt test --state ./target/previous_run --select state:modified

Runs modified models and their downstream dependencies.

dbt

dbt run --state ./target/previous_run --select state:modified+

Sample Program

This Python script runs a dbt build command that uses slim CI with state comparison. It runs only the models that changed since the last run saved in ./target/previous_run. The output shows which models were run.

dbt

import os
import subprocess

# Assume previous run artifacts are saved in './target/previous_run'
previous_state_path = './target/previous_run'

# Command to run only modified models compared to previous run
command = [
    'dbt', 'build',
    '--state', previous_state_path,
    '--select', 'state:modified'
]

# Run the command and capture output
result = subprocess.run(command, capture_output=True, text=True)

print(result.stdout)

Important Notes

Make sure to save the previous run's artifacts (manifest.json, run_results.json) in the folder you point to with --state.

Slim CI only works if you have a previous run to compare with.

You can combine state:modified with other selectors like + to include dependencies.

Summary

Slim CI runs only changed models by comparing current and previous states.

Use --state to point to previous run artifacts and --select state:modified to select changed models.

This saves time and resources during development and testing.

Practice

1. What is the main purpose of using --state and --select state:modified in dbt's Slim CI?

easy

A. To delete all previous run artifacts

B. To run all models regardless of changes

C. To create a backup of the current state

D. To run only the models that have changed since the last run

Solution

1. Step 1: Understand the role of --state

   The --state flag points dbt to the previous run's artifacts, allowing comparison between runs.

2. Step 2: Understand --select state:modified

   This selects only the models that have changed compared to the previous state, skipping unchanged models.

3. Final Answer:

   To run only the models that have changed since the last run -> Option D

4. Quick Check:

   Slim CI runs changed models only [OK]

Hint: Slim CI runs changed models only, saving time [OK]

Common Mistakes:

• Thinking it runs all models every time
• Confusing state with backup or deletion
• Assuming it runs unchanged models too

2. Which of the following is the correct syntax to run dbt with Slim CI to select only modified models?

easy

A. dbt run --state path/to/previous --select all

B. dbt run --select state:modified --state path/to/previous

C. dbt run --select modified:state --state=previous_path

D. dbt run --state-modified --select path/to/previous

Solution

1. Step 1: Identify correct flag usage

   The correct flags are --state to specify previous run artifacts and --select state:modified to select changed models.

2. Step 2: Match syntax to flags

   dbt run --select state:modified --state path/to/previous correctly uses --select state:modified and --state path/to/previous in proper order and format.

3. Final Answer:

   dbt run --select state:modified --state path/to/previous -> Option B

4. Quick Check:

   Correct flags and order [OK]

Hint: Use --state then --select state:modified for Slim CI [OK]

Common Mistakes:

• Mixing up flag names or order
• Using incorrect flag syntax
• Selecting all models instead of modified

3. Given the following dbt command run twice with changes in model 'sales':

dbt run --state ./target/run1 --select state:modified
# After modifying 'sales' model
 dbt run --state ./target/run1 --select state:modified

What will the second command do?

medium

A. Run only the 'sales' model because it was modified

B. Run all models again ignoring state

C. Run no models because nothing changed

D. Throw an error due to incorrect state path

Solution

1. Step 1: Understand the first run

   The first command runs all models and saves state in ./target/run1.

2. Step 2: Analyze the second run with modification

   The second command compares current state to ./target/run1 and runs only models that changed, here 'sales'.

3. Final Answer:

   Run only the 'sales' model because it was modified -> Option A

4. Quick Check:

   Modified model runs only [OK]

Hint: Second run runs only changed models using saved state [OK]

Common Mistakes:

• Assuming all models run again
• Thinking no models run if only one changed
• Confusing state path usage

4. You run dbt run --state ./previous_run --select state:modified but get an error saying the state path is invalid. What is the most likely cause?

medium

A. The path './previous_run' does not contain valid dbt run artifacts

B. You forgot to add '--select all' flag

C. The '--state' flag must come after '--select'

D. The command requires '--full-refresh' to work

Solution

1. Step 1: Check the meaning of the error

   An invalid state path error means dbt cannot find or read previous run artifacts at the given path.

2. Step 2: Evaluate other options

   Flags order or missing '--select all' do not cause this error; '--full-refresh' is unrelated.

3. Final Answer:

   The path './previous_run' does not contain valid dbt run artifacts -> Option A

4. Quick Check:

   Invalid state path means missing artifacts [OK]

Hint: Ensure state path has valid previous run artifacts [OK]

Common Mistakes:

• Assuming flag order causes error
• Thinking '--full-refresh' fixes state errors
• Forgetting to generate previous run artifacts

5. You want to optimize your CI pipeline using Slim CI. Your previous run artifacts are stored in ./target/previous. You have models 'A', 'B', and 'C'. Only 'B' was modified. Which command will run only the modified model and also run its dependent models?

hard

A. dbt run --state ./target/previous --select state:modified

B. dbt run --state ./target/previous --select state:modified --full-refresh

C. dbt run --state ./target/previous --select state:modified+

D. dbt run --select state:modified --state ./target/previous --exclude B

Solution

1. Step 1: Understand the '+' operator in selection

   The '+' after state:modified selects modified models plus their dependents.

2. Step 2: Match command to requirement

   dbt run --state ./target/previous --select state:modified+ uses state:modified+ with correct --state path, running 'B' and its dependents.

3. Step 3: Evaluate other options

   dbt run --state ./target/previous --select state:modified runs only modified models without dependents; C adds unnecessary full-refresh; D excludes 'B' incorrectly.

4. Final Answer:

   dbt run --state ./target/previous --select state:modified+ -> Option C

5. Quick Check:

   Modified plus dependents run [OK]

Hint: Add '+' to include dependents of modified models [OK]

Common Mistakes:

• Not including dependents with '+'
• Using --full-refresh unnecessarily
• Excluding modified models by mistake