🧪 Code PushUp plugin for tracking code coverage. ☂️
This plugin allows you to measure and track code coverage on your project. It accepts the LCOV coverage format and merges coverage results from any test suites provided.
Measured coverage types are mapped to Code PushUp audits in the following way
- The value is in range 0-100 and represents the code coverage for all passed results (covered / total)
- the score is value converted to 0-1 range
- missing coverage is mapped to issues in the audit details (uncalled functions, uncovered branches or lines)
Important
In order to successfully run your coverage tool and gather coverage results directly within the plugin, all your tests need to pass!
-
If you haven't already, install @code-pushup/cli and create a configuration file.
-
Prepare either existing code coverage result files or a command for a coverage tool of your choice that will generate the results. Set lcov as the reporter to the configuration (example for Jest here).
-
Add this plugin to the
plugins
array in your Code PushUp CLI config file (e.g.code-pushup.config.js
).Pass paths to the code coverage results in LCOV format and optionally define your code coverage tool to be run first. All coverage types are measured by default. If you wish to focus on a subset of offered types of coverage, define them in
coverageTypes
.📌 Please note that when you define the tool command, you still need to define the paths to all relevant coverage results.
The configuration will look similarly to the following:
import coveragePlugin from '@code-pushup/coverage-plugin'; export default { // ... plugins: [ // ... await coveragePlugin({ reports: ['coverage/lcov.info'], coverageToolCommand: { command: 'npx', args: ['jest', '--coverage', '--coverageReporters=lcov'], }, }), ], };
-
(Optional) Reference individual audits or the provided plugin group which you wish to include in custom categories (use
npx code-pushup print-config
to list audits and groups).💡 Assign weights based on what influence each coverage type should have on the overall category score (assign weight 0 to only include as extra info, without influencing category score).
export default { // ... categories: [ { slug: 'code-coverage', title: 'Code coverage', refs: [ { type: 'group', plugin: 'coverage', slug: 'coverage', weight: 1, }, // ... ], }, // ... ], };
-
Run the CLI with
npx code-pushup collect
and view or upload report (refer to CLI docs).
Code coverage is a metric that indicates what percentage of source code is executed by unit tests. It can give insights into test effectiveness and uncover parts of source code that would otherwise go untested.
- Statement coverage: Measures how many statements are executed in at least one test.
- Line coverage: Measures how many lines are executed in at least one test. Unlike statement coverage, any partially executed line counts towards line coverage.
- Condition coverage: Measures all condition values (
true
/false
) evaluated for a conditional statement in at least one test. - Branch coverage: Measures how many branches are executed as a result of conditional statements (
if
/else
and other) in at least one test. In case of short-circuit logic, only executed paths are counted in. Unlike condition coverage, it does not ensure all combinations of condition values are tested. - Function coverage: Measures how many functions are called in at least one test. Argument values, usage of optional arguments or default values is irrelevant for this metric.
Important
Please note that code coverage is not the same as test coverage. Test coverage measures the amount of acceptance criteria covered by tests and is hard to formally verify. This means that code coverage cannot guarantee that the designed software caters to the business requirements.
If you want to know more code coverage and how each type of coverage is measured, go to Software Testing Help.
The LCOV format was originally used by GCOV tool for coverage results in C/C++ projects. It recognises the following entities:
- TN [test name]
- SF [source file]
- FN [line number] [function name]
- FNF [number of functions found]
- FNH [number of functions hit]
- FNDA [number of hits] [function name]
- BRDA [line number] [block number] [branch name] [number of hits]
- BRF [number of branches found]
- BRH [number of branches taken]
- DA [line number] [number of hits]
- LF [lines found]
- LH [lines hit]
Here is the source of the information above.
Note
Branch name is usually a number indexed from 0, indicating either truthy/falsy condition or loop conditions.
The plugin accepts the following parameters:
coverageTypes
: An array of types of coverage that you wish to track. Supported values:function
,branch
,line
. Defaults to all available types.reports
: Array of information about files with code coverage results. LCOV format is supported for now.- For a single project, providing paths to results as strings is enough.
- If you have a monorepo, both path to results (
resultsPath
) and path from the root to project the results belong to (pathToProject
) need to be provided for the LCOV format. For Nx monorepos, you can use our helper functiongetNxCoveragePaths
to get the path information automatically.
- (optional)
coverageToolCommand
: If you wish to run your coverage tool to generate the results first, you may define it here. - (optional)
perfectScoreThreshold
: If your coverage goal is not 100%, you may define it here in range 0-1. Any score above the defined threshold will be given the perfect score. The value will stay unaffected.
This plugin provides a group for convenient declaration in your config. When defined this way, all measured coverage type audits have the same weight.
// ...
categories: [
{
slug: 'code-coverage',
title: 'Code coverage',
refs: [
{
type: 'group',
plugin: 'coverage',
slug: 'coverage',
weight: 1,
},
// ...
],
},
// ...
],
Each coverage type still has its own audit. So when you want to include a subset of coverage types or assign different weights to them, you can do so in the following way:
// ...
categories: [
{
slug: 'code-coverage',
title: 'Code coverage',
refs: [
{
type: 'audit',
plugin: 'coverage',
slug: 'function-coverage',
weight: 2,
},
{
type: 'audit',
plugin: 'coverage',
slug: 'branch-coverage',
weight: 1,
},
// ...
],
},
// ...
],
An audit is an aggregation of all results for one coverage type passed to the plugin.
For functions and branches, an issue points to a single instance of a branch or function not covered in any test and counts as an error. In line coverage, one issue groups any amount of consecutive lines together to reduce the total amount of issues and counts as a warning.
For instance, the following can be an audit output for line coverage.
{
"slug": "line-coverage",
"displayValue": "95 %",
"score": 0.95,
"value": 95,
"details": {
"issues": [
{
"message": "Lines 7-9 are not covered in any test case.",
"severity": "warning",
"source": {
"file": "packages/cli/src/lib/utils.ts",
"position": {
"startLine": 7,
"endLine": 9
}
}
}
]
}
}
At the moment, the LCOV results include (empty-report)
functions with missing coverage. These point to various imports or exports, not actual functions. For that reason, they are omitted from the results.
As a part of the plugin, there is a getNxCoveragePaths
helper for setting up paths to coverage results if you are using Nx. The helper accepts all relevant targets (e.g. test
or unit-test
) and searches for a coverage path option.
Jest and Vitest configuration options are currently supported:
- For
@nx/jest
executor it looks for thecoverageDirectory
option. - For
@nx/vite
executor it looks for thereportsDirectory
option.
Important
Please note that you need to set up the coverage directory option in your project.json
target options. Test configuration files are not searched.