covector

covector

Transparent and flexible change management for publishing packages and assets. Publish and deploy from a single asset repository, monorepos, and even multi-language repositories.

Usage

This library is primarily designed as a CLI, but we do also have a GitHub Action that can be used. The CLI commands can be used within any CI/CD environment. To use this in a production setting, we expect a .changes folder where one would put the config.json. (Eventually an init command can bootstrap this for you. Would love to see a PR adding this.) We typically will recommend adding covector as a dev dependency at the root. Then one can run npm run covector to access the commands. We include a dry run mode that you can test the version and publish commands with.

Primary Commands

Command	Description
init	Initializes the configuration required for covector.
status	Outputs the status letting you know if there are changes and what they are.
config	Pretty prints the config.
version	Will see which packages have changes, run the version bumps on any package changed, and update the changelog.
publish	Will run the getPublishedVersion specified in the config, and any package whose version does not match that return will attempt the publish commands.

GitHub Action

The GitHub Action adds a little extra sugar on top. One of the additional inputs is a version-or-publish command, and it will dynamically choose to run either command depending on the context.

version command

The context is the existence of changes in the .changes folder. if there are changes, we expect that the next command to be run is the version command. The version command will delete those change files, apply the applicable version bumps to your packages, and create changelogs.

publish command

If we find that there are no changes, then we will run the publish command. The publish command is written with the expectation of failure. The design is to fail forward and be able to recover from it cleanly. Publishing on CI has many failure points and it will happen at some point. The publish sequence can be primarily grouped into two main parts. The first part of the command will run the getPublishedVersions from your config which will determine if the current package version is published. Theoretically we could skip this command and it would always publish, but some package managers will fail and exit the command if it's already published. That failure would prevent other packages from being published.

After the published command runs through and checks which packages need to be published, the next sequence is running all of the prepublish, publish, and postpublish commands. The commands are run in groups where the prepublish commands will be run for every package. With no failures, the publish command will be run for every package and with no failures then the postpublish command will be run for every package.

GitHub Release

From this point, the GitHub Action will also create a release and upload any assets if you have opted into this. You can opt into it by passing createRelease: true as an input. To upload assets to the release, add an array of assets in the config. Each item should be an object with a path and name property.

outputs from action

The action will also output some of the information from the commands. This is helpful for chaining further commands together depending on what covector has done. Note that we can only output strings so remember to keep this in mind when accessing or using them in conditionals. One of the outputs is the changes that were applied as a stringified JSON object. This can either be consumed and processed in a later step, or can potentially be directly accessed using something like jq (a Linux command that can pull items from json piped to it).

It also outputs if a successfulPublish which is a boolean (as a string) if it was successful or not, which can be used to determine if follow on commands and jobs and steps need to run. One example might be to push a documentation update after a publish has been successful. The last output is the command that was run. Well not terribly applicable if you were specifying a command directly, it is useful when using the version-or-publish input. When a version command is run, there is no expectation from covector on how to commit or follow up that version command. We recommend using a action that will create a PR for you. This also allows you to do follow-up work that can be included in the commit if you so choose. This output can tell you if the version command was run in which case you would expect to have to create a PR. The PR step can then use a conditional statement to run based on that commandRan.

Applying Changes

As you create PRs and make changes that require a version bump, please add a new markdown file in this folder. You do not note the version number, but rather the type of bump that you expect: major, minor, or patch. The filename is not important, as long as it is a .md, but we recommend it represents the overall change for our sanity.

When you select the version bump required, you do not need to consider depedencies. Only note the package with the actual change, and any packages that depend on that package will be bumped automatically in the process.

Use the following format:

---
"covector": minor
"@covector/apply": patch
---

Change summary goes here

Summaries do not have a specific character limit, but are text only. These summaries are used within the (future implementation of) changelogs. They will give context to the change and also point back to the original PR if more details and context are needed.

Changes will be designated as a major, minor or patch as further described in semver.

Given a version number MAJOR.MINOR.PATCH, increment the:

• MAJOR version when you make incompatible API changes,
• MINOR version when you add functionality in a backwards compatible manner, and
• PATCH version when you make backwards compatible bug fixes.

Additional labels for pre-release and build metadata are available as extensions to the MAJOR.MINOR.PATCH format, but will be discussed prior to usage (as extra steps will be necessary in consideration of merging and publishing).

Additionally, a change may contain an optional tag (section) major:breaking, minor:bug or patch:pref but will be discussed later on.

Power of Configuration

Covector is driven by your configuration, and creates rather open ended use cases. The two to level properties are packages and pkgManagers. The packages is an object of your packages with the key being name of your package (or even a nickname!), a path to it's folder, the package manager, and an array of dependencies if applicable (which operates across languages too!).

Each package is driven by commands. When using a command such as covector publish, we look up the relevant command for each package and run it, publish in this instance. Managing your changes are never that easy though. We support an array of commands letting you chain multiple commands together for each package. Even more, we support both pre and post versions of your commands both which accept arrays of commands. Running a Typescript build, tests, an audit, publishing the packages and applying git tags is no longer too complex of a workflow.

To those of you using a monorepo with multiple packages, the pkgManagers will be quite useful. Each package can opt into a package manager that matches a key in pkgManagers. Specifying a command in one of the pkgManagers will apply it to all packages with that manager key if the package hasn't specified the command.

Your configuration may look something like this.

{
  "gitSiteUrl": "https://www.github.com/jbolda/covector/",
  "additionalBumpTypes": ["housekeeping"],
  "pkgManagers": {
    "javascript": {
      "version": true,
      "getPublishedVersion": "npm view ${ pkg.pkg } version",
      "publish": "npm publish --access public",
      "postpublish": [
        "git tag ${ pkg.pkg }-v${ pkgFile.version }",
        "git push --tags"
      ]
    }
  },
  "packages": {
    "covector": {
      "path": "./packages/covector",
      "manager": "javascript",
      "dependencies": [
        "@covector/apply",
        "@covector/assemble",
        "@covector/files",
        "@covector/changelog"
      ]
    },
    "@covector/apply": {
      "path": "./packages/apply",
      "manager": "javascript"
    },
    "@covector/assemble": {
      "path": "./packages/assemble",
      "manager": "javascript"
    },
    "@covector/files": {
      "path": "./packages/files",
      "manager": "javascript"
    },
    "@covector/changelog": {
      "path": "./packages/changelog",
      "manager": "javascript",
      "dependencies": ["@covector/files"]
    }
  }
}

Prereleases

We have initial support for prereleases. Functionally, you put your repository into "prerelease mode". Every package within the repo will begin apply bumps as prerelease numbers related to the type of bump being applied. Enabling this is currently a manual process with support via CLI commands planned in a future release.

To enable, add a pre.json file with your .changes folder.

{
  "tag": "beta",
  "changes": []
}

During the versioning step, the change files will be recorded in this array and kept within the .changes folder instead of being removed. This provides a record on the previously applied bumps and inform the next bump. It will resolve to premajor, preminor, or prepatch if a bump of that impact has not yet been applied. If a new change added with a bump at or below that "impact", than it resolves to a prerelease bump. As you configure prerelease mode, we recommend making use of covector version --dry-run to help provide feedback.

When you are ready to release the stable version, remove the pre.json and all of your change files. Add a new change file announcing the stable version, and apply the bumps to the packages coming out of prerelease.

Prereleases in general complicate considerations, and introduce many foot-guns. We expect to expand functionality here, but look to do it with guardrails and assistance to hopefully prevent issues.

Command Options

Each command may be specified as a string or an object. The following are equivalent.

A command specified as a string.

{
  "packages": {
    "covector": {
      "path": "./packages/covector",
      "publish": "echo publish"
    }
  }
}

A command specified as an object.

{
  "packages": {
    "covector": {
      "path": "./packages/covector",
      "publish": { "command": "echo publish" }
    }
  }
}

The command is required. The follow are additional options you may specify.

• retries: number[] - If the command fails, opt-in to retrying based on this timeout and frequency. Using [2000, 2000] would try two additional times with a 2 second delay between attempts. It would throw on the last attempt if they all fail.

Built-In Commands

Besides specifying a command with additional options, one may use a built-in command. The following are the currently available commands.

fetch:check

This requires an options.url for use. It will fetch the specified endpoint, and throw if the response returns a code >= 400 or if the JSON response includes an errors. This is useful to check if a version was published. The following would check the registry if this version was published to determine if it needs to run a covector publish. After publishing, it checks with retries on a five second timeout to confirm the publish appears in the registry.

{
  "packages": {
    "covector": {
      "path": "./packages/covector",
      "getPublishedVersion": {
        "use": "fetch:check",
        "options": {
          "url": "https://registry.npmjs.com/${ pkg.pkg }/${ pkg.pkgFile.version }"
        }
      },
      "publish": "npm publish --access public",
      "postpublish": {
        "use": "fetch:check",
        "options": {
          "url": "https://registry.npmjs.com/${ pkg.pkg }/${ pkg.pkgFile.version }"
        },
        "retries": [5000, 5000, 5000]
      }
    }
  }
}

Change Tags

Each bump in a change file could optionally contain a tag (section) in the format of <bump>:<tag>, and will be used to group related changes under one tag in the final changelog:

---
"covector": minor:feat
"@covector/apply": patch:fix
"@covector/assemble": patch
---

Change summary goes here

By default the tags are rednered as is but a longer description could be provided in the config:

{
  "changeTags": {
    "feat": "New Features",
    "bug": "Bug Fixes",
    "pref": "Prefromance Improvements",
    "misc": "Miscellaneous"
  }
}

Note: the order is important, tags will be rendered in the order they are defined in.

By default, untagged changes will be rendered at the top before the tagged changes, but you can specify a default tag if needed.

{
  "changeTags": {
    "feat": "New Features",
    "bug": "Bug Fixes",
    "pref": "Prefromance Improvements",
    "misc": "Miscellaneous"
  },
  "defaultChangeTag": "misc"
}