Skip to content

Linter guide

Linter

Linting is a powerful tool for improving code quality and consistency. It enables you to automatically validate model definition, ensuring they adhere to your team's best practices.

When a SQLMesh plan is created, each model's code is checked for compliance with a set of rules you choose.

SQLMesh provides built-in rules, and you can define custom rules. This improves code quality and helps detect issues early in the development cycle when they are simpler to debug.

Rules

Each linting rule is responsible for identifying a pattern in a model's code.

Some rules validate that a pattern is not present, such as not allowing SELECT * in a model's outermost query. Other rules validate that a pattern is present, like ensuring that every model's owner field is specified. We refer to both of these below as "validating a pattern".

Rules are defined in Python. Each rule is an individual Python class that inherits from SQLMesh's Rule base class and defines the logic for validating a pattern.

We display a portion of the Rule base class's code below (full source code). Its methods and properties illustrate the most important components of the subclassed rules you define.

Each rule class you create has four vital components:

  1. Name: the class's name is used as the rule's name.
  2. Description: the class should define a docstring that provides a short explanation of the rule's purpose.
  3. Pattern validation logic: the class should define a check_model() method containing the core logic that validates the rule's pattern. The method can access any Model attribute.
  4. Rule violation logic: if a rule's pattern is not validated, the rule is "violated" and the class should return a RuleViolation object. The RuleViolation object should include the contextual information a user needs to understand and fix the problem.
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
# Class name used as rule's name
class Rule:
    # Docstring provides rule's description
    """The base class for a rule."""

    # Pattern validation logic goes in `check_model()` method
    @abc.abstractmethod
    def check_model(self, model: Model) -> t.Optional[RuleViolation]:
        """The evaluation function that checks for a violation of this rule."""

    # Rule violation object returned by `violation()` method
    def violation(self, violation_msg: t.Optional[str] = None) -> RuleViolation:
        """Return a RuleViolation instance if this rule is violated"""
        return RuleViolation(rule=self, violation_msg=violation_msg or self.summary)

Built-in rules

SQLMesh includes a set of predefined rules that check for potential SQL errors or enforce code style.

An example of the latter is the NoSelectStar rule, which prohibits a model from using SELECT * in its query's outer-most select statement.

Here is code for the built-in NoSelectStar rule class, with the different components annotated:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
# Rule's name is the class name `NoSelectStar`
class NoSelectStar(Rule):
    # Docstring explaining rule
    """Query should not contain SELECT * on its outer most projections, even if it can be expanded."""

    def check_model(self, model: Model) -> t.Optional[RuleViolation]:
        # If this model does not contain a SQL query, there is nothing to validate
        if not isinstance(model, SqlModel):
            return None

        # Use the query's `is_star` property to detect the `SELECT *` pattern.
        # If present, call the `violation()` method to return a `RuleViolation` object.
        return self.violation() if model.query.is_star else None

Here are all of SQLMesh's built-in linting rules:

Name Check type Explanation
ambiguousorinvalidcolumn Correctness SQLMesh found duplicate columns or was unable to determine whether a column is duplicated or not
invalidselectstarexpansion Correctness The query's top-level selection may be SELECT *, but only if SQLMesh can expand the SELECT * into individual columns
noselectstar Stylistic The query's top-level selection may not be SELECT *, even if SQLMesh can expand the SELECT * into individual columns
nomissingaudits Governance SQLMesh did not find any audits in the model's configuration to test data quality.

User-defined rules

You may define custom rules to implement your team's best practices.

For instance, you could ensure all models have an owner by defining the following linting rule:

linter/user.py
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
import typing as t

from sqlmesh.core.linter.rule import Rule, RuleViolation
from sqlmesh.core.model import Model

class NoMissingOwner(Rule):
    """Model owner should always be specified."""

    def check_model(self, model: Model) -> t.Optional[RuleViolation]:
        # Rule violated if the model's owner field (`model.owner`) is not specified
        return self.violation() if not model.owner else None

Place a rule's code in the project's linter/ directory. SQLMesh will load all subclasses of Rule from that directory.

If the rule is specified in the project's configuration file, SQLMesh will run it when: - A plan is created during sqlmesh plan - The command sqlmesh lint is ran

SQLMesh will error if a model violates the rule, informing you which model(s) violated the rule. In this example, full_model.sql violated the NoMissingOwner rule, essentially halting execution:

$ sqlmesh plan

Linter errors for .../models/full_model.sql:
 - nomissingowner: Model owner should always be specified.

Error: Linter detected errors in the code. Please fix them before proceeding.

Or through the standalone command, for faster iterations:

$ sqlmesh lint

Linter errors for .../models/full_model.sql:
 - nomissingowner: Model owner should always be specified.

Error: Linter detected errors in the code. Please fix them before proceeding.

Use sqlmesh lint --help for more information.

Applying linting rules

Specify which linting rules a project should apply in the project's configuration file.

Rules are specified as lists of rule names under the linter key. Globally enable or disable linting with the enabled key, which is false by default.

NOTE: you must set the enabled key to true key to apply the project's linting rules.

Specific linting rules

This example specifies that the "ambiguousorinvalidcolumn" and "invalidselectstarexpansion" linting rules should be enforced:

1
2
3
linter:
  enabled: true
  rules: ["ambiguousorinvalidcolumn", "invalidselectstarexpansion"]

1
2
3
4
5
6
7
8
from sqlmesh.core.config import Config, LinterConfig

config = Config(
    linter=LinterConfig(
        enabled=True,
        rules=["ambiguousorinvalidcolumn", "invalidselectstarexpansion"]
    )
)

All linting rules

Apply every built-in and user-defined rule by specifying "ALL" instead of a list of rules:

1
2
3
linter:
  enabled: True
  rules: "ALL"

1
2
3
4
5
6
7
8
from sqlmesh.core.config import Config, LinterConfig

config = Config(
    linter=LinterConfig(
        enabled=True,
        rules="all",
    )
)

If you want to apply all rules except for a few, you can specify "ALL" and list the rules to ignore in the ignored_rules key:

1
2
3
4
linter:
  enabled: True
  rules: "ALL" # apply all built-in and user-defined rules and error if violated
  ignored_rules: ["noselectstar"] # but don't run the `noselectstar` rule

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
from sqlmesh.core.config import Config, LinterConfig

config = Config(
    linter=LinterConfig(
        enabled=True,
        # apply all built-in and user-defined linting rules and error if violated
        rules="all",
         # but don't run the `noselectstar` rule
        ignored_rules=["noselectstar"]
    )
)

Exclude a model from linting

You can specify that a specific model ignore a linting rule by specifying ignored_rules in its MODEL block.

This example specifies that the model docs_example.full_model should not run the invalidselectstarexpansion rule:

1
2
3
4
MODEL(
  name docs_example.full_model,
  ignored_rules ["invalidselectstarexpansion"] # or "ALL" to turn off linting completely
);

Rule violation behavior

Linting rule violations raise an error by default, preventing the project from running until the violation is addressed.

You may specify that a rule's violation should not error and only log a warning by specifying it in the warn_rules key instead of the rules key.

1
2
3
4
5
6
linter:
  enabled: True
  # error if `ambiguousorinvalidcolumn` rule violated
  rules: ["ambiguousorinvalidcolumn"]
  # but only warn if "invalidselectstarexpansion" is violated
  warn_rules: ["invalidselectstarexpansion"]

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
from sqlmesh.core.config import Config, LinterConfig

config = Config(
    linter=LinterConfig(
        enabled=True,
        # error if `ambiguousorinvalidcolumn` rule violated
        rules=["ambiguousorinvalidcolumn"],
        # but only warn if "invalidselectstarexpansion" is violated
        warn_rules=["invalidselectstarexpansion"],
    )
)

SQLMesh will raise an error if the same rule is included in more than one of the rules, warn_rules, and ignored_rules keys since they should be mutually exclusive.