Skip to content

Latest commit

 

History

History
124 lines (102 loc) · 4.86 KB

Debug-custom-ESLint-check-tests.md

File metadata and controls

124 lines (102 loc) · 4.86 KB

Table of contents

Introduction

When you write tests for a custom ESLint rule, you may sometimes have trouble understanding why the tests fail. This page will give you some tips to help you debug your failing tests.

Print statements

Printing out the values of variables in your lint rule is one of the most useful debugging strategies because it lets you check whether the variables in your rule have the values you expect. For ESLint rules, you can add console.log(), console.error(), or console.info() statements to your custom ESLint rule code. The output will appear along with the test status output.

For example, suppose we have a lint rule that raises an error every time .click() or .setValue() is called on an object instead of action.click() or action.setValue(), which we would like people to use instead. If we add console.log() statement to print DEBUGGING found lint error whenever our lint rule raises an error, we see this output when running the tests:

  e2e-action
    valid
      ✓ it('should test a feature', function() {
        action.click("Element", elem);
      });
      ✓ it('should test a feature', function() {
        action.setValue("Element", elem, "keys");
      });
      ✓ it('should test a feature', function() {
        console.log(elem.click);
      });
      ✓ it('should test a feature', function() {
        console.log(elem.setValue);
      });
    invalid
DEBUGGING found lint error
      ✓ it('should test a feature', function() {
        elem.click();
      });
DEBUGGING found lint error
      ✓ it('should test a feature', function() {
        elem.setValue("keys");
      });
DEBUGGING found lint error
DEBUGGING found lint error
      ✓ it('should test a feature', function() {
        $('.e2e-test')).click();
      });
      ✓ it('should test a feature', function() {
        $('.e2e-test')).setValue("keys");
      });

However, there are a few issues that you may run into with this approach:

  • All the tests are named should test a feature. This is because we generate tests dynamically based on the test cases you provide in your *.spec.js file. If you are used to using fit to run a single frontend test in isolation, this dynamic test generation means that fit won't work.
  • Our debugging statements don't appear in a consistent place with respect to the test status report. For example, the debugging output for the last test appears grouped with the output from the second-to-last test. You shouldn't assume that output from your debugging statements will be synchronized with the test status messages.

These points make debugging a little more difficult, but there's an easy fix: Comment-out all the tests in your *.spec.js file except the one that you want to debug. Then you'll know that all the debugging statements came from that one test.

Inspect the AST

If you are confused about why your linter is raising or not raising an error on a particular test case, we recommend copying your test case into AST Explorer and examining the AST (abstract syntax tree). Then you can use print statements in your linter code to examine the AST nodes being visited.

For example, suppose we have this rule to forbid console.log() statements:

export default function(context) {
  return {
    CallExpression(node) {
      console.log(
        `Considering node starting at line ${node.loc.start.line}`);
      if (
        node.callee.object.name == 'console' &&
        node.callee.property.name == 'log'
      ) {
        context.report({
          node,
          message: 'Do not use console.log()',
        });
      }
    }
  };
};

Now, suppose we run our lint rule on the following test case:

console.log("Hello");
foo(1)

The test fails with:

Considering node starting at line 1
Considering node starting at line 2
TypeError: node.callee.object is undefined

This tells us that the error occurs on the second line of the test code. If we copy this test case into AST Explorer, we can see that foo(1) is a CallExpression but its callee attribute has no object attribute:

Screenshot of AST Explorer with test case and AST.

This is why we're getting that error! We could fix this problem by checking that callee has the object and property attributes at the start of our rule:

export default function(context) {
  return {
    CallExpression(node) {
      if (
        "callee" in node &&
        "object" in node.callee &&
        "property" in node.callee &&
        node.callee.object.name == 'console' &&
        node.callee.property.name == 'log'
      ) {
        context.report({
          node,
          message: 'Do not use console.log()',
        });
      }
    }
  };
};