Flags allow you to isolate coverage reports for different tests. This is particularly helpful if you have multiple types of tests (e.g., unit, integration, frontend, backend, etc), or you're employing a monorepo setup where you'd like to encapsulate each project's test coverage independently.
Flags feature on the Codecov Compare page
As of June 2019, Codecov has suspended flags for Codecov.io users on the Codecov Compare page.
Flags were causing the compare page to consistently timeout for a minority of users. The Codecov team is investigating refactoring flags to make the compare page more usable, even with many flags turned on.
The compare page will still work for comparing changed code
The flags feature will still be viewable on each individual commit view
Flags will still be functional on pull and merge request status checks
We will continue to iterate on this feature and push out updates here.
As a minimal example, consider a project called Monorepo that has the following folder structure:
Monorepo /moduleA /moduleB
To generate a flag and individual coverage gates per project in the mono repo the codecov.yml can be structured as follows:
coverage: status: project: default: target: auto moduleA: target: auto flags: - moduleA moduleB: target: auto flags: - moduleB comment: layout: "reach, diff, flags, files" behavior: default require_changes: false require_base: yes require_head: yes branches: null flags: moduleA: paths: - src/moduleA/code.py carryforward: false moduleB: paths: - src/moduleB/code2.py carryforward: true
This yaml configuration will fail a Pull Request if...
• ...the entire Monorepo project is below 90% covered (this is the
default specification of the yaml above)
• ...the moduleA coverage is less than 90%
• ...the moduleB coverage is less than 50%
It is important to note that for flagging to work, separate coverage reports must be uploaded and flagged per project in the mono repo. For example, in a CI run for the MonoRepo project example, assume that coverage for moduleA is written to
Then that report would be uploaded as follows:
bash <(curl https://codecov.io/bash) -t <token> -f moduleA/output/coverage.xml -F moduleA
You Must Upload using -F
To properly leverage flags and you must ensure you are uploading coverage reports with the appropriate flag name.
The moduleA flag can be used for any number of reports (unit tests, integration tests, etc) by flagging each report with
moduleA. The reports will be merged and contribute the project's total coverage and the total coverage of the files covered under the
However, specifying multiple flags for a single report upload can result in erroneous coverage unless the contents of that report fully encompass each flag. For example, one report that contains coverage information for moduleA and moduleB, can be uploaded in a manner that is valid but technically incorrect:
bash <(curl https://codecov.io/bash) -t <token> -f coverage.xml -F moduleA -F moduleB
This will apply the entire coverage of the report to both flags, resulting in incorrect coverage. Specifically, the
moduleB flags will display the coverage of the entire uploaded report, not just the subset of the report that happens to cover the files under their purview.
For coverage information to be properly tracked per flag, the flags must be applied at upload time.
To apply a flag simply submit coverage reports with the
-F flagname included in the upload command; Codecov will do the rest.
The following example illustrates how you may run specific test groups and upload coverage, flagging each coverage group. This strategy for uploading is preferred as it cleanly associates a single flag with a single report.
# example running unittests only py.test --cov=./ -k tests/unittests/ bash <(curl -s https://codecov.io/bash) -c -F unittests # example running integration tests only py.test --cov=./ -k tests/integration/ bash <(curl -s https://codecov.io/bash) -c -F integration # example running ui tests only npm test bash <(curl -s https://codecov.io/bash) -c -F ui
Remove old reports after uploading (optional)
-cargument to clear the workspace of all coverage reports before running you next set of tests.
Flags must be lowercase, alphanumeric, periods, or hyphens and not exceed 45 characters
Only lowercase, alphanumeric values are accepted.
If you do not test all of your repo code on each commit, Codecov uses a feature called Carryforward Flags to help only update coverage on tests that were run. Carryforward Flags is built on top of basic Flags.
Carryforward Flag Usage
It is highly recommended to read the full documentation on carryforward flags before using them in your project.
Carryforward Flags are used by appending your YAML with
flags: ui: paths: - ui_1.py - ui_2.py carryforward: true unit: paths: - unit_1.py - unit_2.py carryforward: true enterprise: paths: - ent_1.py - ent_2.py carryforward: false # If no Carryfoward flag specified in YAML, the # default configuration is false.
Flags can be used to create custom notifications to your repository provider. This makes it easy to see per-flag coverage information alongside pull requests.
You can specify Flags in your Codecov Yaml for statuses and all notifications.
coverage: status: project: default: off frontend: flags: - frontend backend: target: 50% flags: - backend api: target: 89% flags: - api flags: # filter the folder(s) you wish to measure by that flag backend: # only include files in the backend folder paths: - app/backend/ frontend: paths: - app/frontend/ api: paths: - app/api/ tests: paths: - tests/
Codecov provides a strategy to isolate specific builds from the master report while maintaining the report's integrity. When reports are not joined into the master report, they will be ignored for comparison, although they will remain accessible for source overlay, api, badges, and graphing.
A nightly build is an example of this feature. What follows is a Yaml configuration for a nightly build.
flags: nightly: joined: false
Now that we have configured the flag
nightly to not join into the master report, let's upload a report flagged as
bash <(curl -s https://codecov.io/bash) -F nightly
-F nightly we mark all the coverage report data for this build as
nightly coverage data.
Updated about a month ago