feat: preserve major version zero on breaking changes (when using conventional commits) #2486

bajtos · 2020-03-09T12:31:50Z

Description

According to semver, major version zero (0.y.z) is for initial development. Anything MAY change at any time. The public API SHOULD NOT be considered stable. The version 1.0.0 defines the (initial stable) public API.

In my opinion formed by my experience building https://github.com/strongloop/loopback-next, in order to allow monorepos to use major version zero meaningfully, the transition from 0.x to 1.x must be explicitly requested by the user. Breaking changes MUST NOT automatically bump the major version from 0.x to 1.x.

The usual convention is to use semver-patch bumps for bugfix releases and semver-minor for everything else, including breaking changes. This matches the behavior of ^ operator as implemented by npm (see docs of the semver module).

This commit implements the convention described above for packages in major version zero:

a patch-level change bumps semver-patch version (NO CHANGE)
a minor-level change bumps semver-minor version (NO CHANGE)
a major-level (breaking) change bumps semver-minor version (NEW)

Motivation and Context

In LoopBack, we have been several times bitten by lerna version bumping up our packages from 0.x to 1.0 (see loopbackio/loopback-next#1068). Most recently, we dropped support for Node.js 8.x because it reached end of life. We described this change as breaking, and so in the next release, lerna wanted to graduate our experimental packages with unstable API to stable 1.0. That was not desired at all!

Related discussions and alternative solutions:

[conventional-commits] Add a flag to disable conventional-recommended-bump #1195 [conventional-commits] Add a flag to disable conventional-recommended-bump
Do not bump to 1.0.0 when issuing a breaking change < 1.0.0 conventional-changelog/conventional-changelog#239 Do not bump to 1.0.0 when issuing a breaking change < 1.0.0
Recommend different bumping strategy for 0.x versions conventional-changelog/conventional-changelog#294 Recommend different bumping strategy for 0.x versions
Incorrect Increment for versions < 1.0 npm/node-semver#177 Incorrect Increment for versions < 1.0 (rejected)
https://www.npmjs.com/package/shifted-semver-increment (no longer maintained)

How Has This Been Tested?

I added a new unit test, verified that it fails against the current implementation and passes with my patch applied.

I am not sure if a single test is enough for a change like this. Please suggest what other scenarios to cover by tests to give us enough confidence, I am happy to add more tests as needed.

Types of changes

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

My code follows the code style of this project.
My change requires a change to the documentation. -- I am not sure about this?
I have updated the documentation accordingly. -- I think no doc changes are needed.
I have read the CONTRIBUTING document.
I have added tests to cover my changes.
All new and existing tests passed. (lerna bootstrap --npm-client yarn is failing for me, the error seems unrelated to my changes.)

According to semver, major version zero (0.y.z) is for initial development. Anything MAY change at any time. The public API SHOULD NOT be considered stable. The version 1.0.0 defines the (initial stable) public API. To allow monorepos to use major version zero meaningfully, the transition from 0.x to 1.x must be explicitly requested by the user. Breaking changes MUST NOT automatically bump the major version from 0.x to 1.x. The usual convention is to use semver-patch bumps for bugfix releases and semver-minor for everything else, including breaking changes. This matches the behavior of `^` operator as implemented by `npm`. This commit implements the convention described above: - a patch-level change bumps semver-patch version (NO CHANGE) - a minor-level change bumps semver-minor version (NO CHANGE) - a major-level (breaking) change bumps semver-minor version (NEW) Signed-off-by: Miroslav Bajtoš <mbajtoss@gmail.com>

bajtos · 2020-03-09T12:36:31Z

Discussion points:

At high-level, is it acceptable to have different bumping behavior for 0.x versions implemented inside lerna?
Should we introduce a new CLI option allowing users to choose between the old or the new behavior? We already have similar options --conventional-prerelease and --conventional-graduate, see https://github.com/lerna/lerna/tree/master/commands/version#prerelease
How and where to document the new behavior (and the new option if it's introduced)?

raymondfeng · 2020-03-09T15:48:32Z

core/conventional-commits/lib/recommend-version.js

+            // as implemented by `npm`.
+            //
+            if (releaseType === "major") {
+              releaseType = "minor";


I guess if --conventional-graduate is true, we will bump the release to the new major, such as 0.1.0 -> 1.0.0?

Personally, I would find such rule as possibly confusing. Also, how would such rule work in large monorepos having packages both in pre-release versions (2.0.0-beta.1) and major zero versions (0.2.5)?

I am currently envisioning the following process for graduating from 0.x to 1.0.0:

Add a commit changing the version in package.json to 1.0.0-1 (or a similar pre-release). This will be just a commit, no release/tag is associated with it. The commit can have a nice descriptive message for the changelog, e.g. feat: graduate to 1.0.

Run lerna version with --conventional-graduate to bump the temporary version 1.0.0-1 to final 1.0.0.

Please note --conventional-graduate allows a list of packages. Whether it's passed in as a CLI option or configured in lerna.json, we probably should honor it. See https://github.com/lerna/lerna/blob/master/commands/version/README.md#--conventional-graduate.

I don't see --conventional-graduate affecting 0.x ranges, as they are not "prereleases" in the sense that option is intended to modify. 0.x ranges, as I understand them, mean "API unstable", not "this is a prerelease of a specified quality".

tpluscode · 2020-04-18T15:03:58Z

@bajtos please refer to #2549 I've just created.

NPM treats minor change to 0.x versions as breaking, ie. will not update 0.4 to 0.5. This can be tried on https://semver.npmjs.com

Thus, for 0.x packages only breaking change should bump minor and any other change should bump patch

bajtos · 2020-04-24T15:24:48Z

Thank you @tpluscode for the link to the issue you have created. I am confused about why are you posting this here? My pull request have been waiting for review by project maintainers for more than a month by now. I don't see how filling a new issue is going to help progress forward?

Are you perhaps trying to say that my pull request is implementing incorrect logic for determining which part of the 0.x version to bump up? If yes, then could you please be more specific about what should I change?

tpluscode · 2020-04-24T18:39:59Z

If yes, then could you please be more specific about what should I change?

I'm sorry, I thought you would tell me. I was only basing my comment on the description and not code.

Here is where our understanding differs on incrementing 0.x packages. You say

semver-patch bumps for bugfix releases
semver-minor for everything else, including breaking changes

Where I think it should be

semver-minor for breaking changes
semver-patch for anything else

I say that because npm treats minor bump on 0.x as breaking change and will not update automatically. I will however update for example 1.1 -> 1.2.

bajtos · 2020-04-30T07:04:19Z

Thank you @tpluscode for clarification!

Here is where our understanding differs on incrementing 0.x packages. You say

semver-patch bumps for bugfix releases
semver-minor for everything else, including breaking changes

Where I think it should be

semver-minor for breaking changes
semver-patch for anything else

I say that because npm treats minor bump on 0.x as breaking change and will not update automatically. I will however update for example 1.1 -> 1.2.

You have raised a valid point 👍

Personally, I don't care that much whether new features bump semver-minor or semver-patch version. I am fine to implement whichever rule that gets widest consensus.

Anything would be better than the current behavior, where breaking changes are bumping 0.x to 1.0 :(

tpluscode · 2020-04-30T07:08:19Z

Anything would be better than the current behavior, where breaking changes are bumping 0.x to 1.0 :(

Agreed. This is why I create a separate issue. In the event that your PR gets merged at its current state and it'd still need tweaking to align 100% with npm's treatment of 0.x version.

Unless you'd agree to adjust right here and we can kill two birds with one stone

evocateur

I'm not a huge fan of the 0.x range for many reasons, but I agree we should adhere closer to how npm deals with those ranges for a more congruent developer experience.

core/conventional-commits/__tests__/conventional-commits.test.js

evocateur · 2020-05-24T21:00:12Z

core/conventional-commits/lib/recommend-version.js

+            // the (initial stable) public API.
+            //
+            // To allow monorepos to use major version zero meaningfully,
+            // the transition from 0.x to 1.x must be explicitly requested


I think some info-level logging indicating what is happening would be helpful. I appreciate the detailed comment.

evocateur · 2020-05-24T21:02:29Z

core/conventional-commits/lib/recommend-version.js

+            // as implemented by `npm`.
+            //
+            if (releaseType === "major") {
+              releaseType = "minor";


I don't see --conventional-graduate affecting 0.x ranges, as they are not "prereleases" in the sense that option is intended to modify. 0.x ranges, as I understand them, mean "API unstable", not "this is a prerelease of a specified quality".

evocateur · 2020-05-24T21:03:51Z

Thanks for your patience!

bajtos · 2020-05-28T10:39:34Z

@evocateur thank you for landing my pull request! ❤️

it includes what we were patching: lerna/lerna#2486

bajtos mentioned this pull request Mar 9, 2020

Build: lerna publish upgrades from 0.x.y to 1.0.0 for breaking changes loopbackio/loopback-next#1068

Closed

raymondfeng reviewed Mar 9, 2020

View reviewed changes

evocateur approved these changes May 24, 2020

View reviewed changes

evocateur merged commit 6126e6c into lerna:master May 24, 2020

evocateur mentioned this pull request May 24, 2020

fix: do not require conventional commits for releases #2528

Closed

9 tasks

bajtos deleted the feat/preserve-zero-major branch May 28, 2020 10:39

brianjquinn mentioned this pull request Aug 27, 2020

How to tag and publish an initial v1.0.0 with conventional commits? #2325

Closed

bajtos mentioned this pull request Nov 19, 2020

Version is not bumping 0.x packages correctly #2549

Closed

achingbrain mentioned this pull request Jul 8, 2021

Publish an initial v1.0.0 with conventional commits #2739

Open

turadg added a commit to Agoric/agoric-sdk that referenced this pull request Feb 23, 2022

chore(root): upgrade Lerna to latest

69569b4

it includes what we were patching: lerna/lerna#2486

turadg added a commit to Agoric/agoric-sdk that referenced this pull request Feb 24, 2022

chore(root): upgrade Lerna to latest

8a08fb4

it includes what we were patching: lerna/lerna#2486

turadg added a commit to Agoric/agoric-sdk that referenced this pull request Feb 24, 2022

chore(root): upgrade Lerna to latest

e78169b

it includes what we were patching: lerna/lerna#2486

turadg mentioned this pull request Feb 24, 2022

build: no-bail when testing/linting whole repo Agoric/agoric-sdk#4646

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: preserve major version zero on breaking changes (when using conventional commits) #2486

feat: preserve major version zero on breaking changes (when using conventional commits) #2486

bajtos commented Mar 9, 2020 •

edited

Loading

bajtos commented Mar 9, 2020

raymondfeng Mar 9, 2020

bajtos Mar 9, 2020

raymondfeng Mar 9, 2020

evocateur May 24, 2020

tpluscode commented Apr 18, 2020

bajtos commented Apr 24, 2020

tpluscode commented Apr 24, 2020 •

edited

Loading

bajtos commented Apr 30, 2020

tpluscode commented Apr 30, 2020

evocateur left a comment

evocateur May 24, 2020

evocateur May 24, 2020

evocateur commented May 24, 2020

bajtos commented May 28, 2020

feat: preserve major version zero on breaking changes (when using conventional commits) #2486

feat: preserve major version zero on breaking changes (when using conventional commits) #2486

Conversation

bajtos commented Mar 9, 2020 • edited Loading

Description

Motivation and Context

How Has This Been Tested?

Types of changes

Checklist:

bajtos commented Mar 9, 2020

raymondfeng Mar 9, 2020

Choose a reason for hiding this comment

bajtos Mar 9, 2020

Choose a reason for hiding this comment

raymondfeng Mar 9, 2020

Choose a reason for hiding this comment

evocateur May 24, 2020

Choose a reason for hiding this comment

tpluscode commented Apr 18, 2020

bajtos commented Apr 24, 2020

tpluscode commented Apr 24, 2020 • edited Loading

bajtos commented Apr 30, 2020

tpluscode commented Apr 30, 2020

evocateur left a comment

Choose a reason for hiding this comment

evocateur May 24, 2020

Choose a reason for hiding this comment

evocateur May 24, 2020

Choose a reason for hiding this comment

evocateur commented May 24, 2020

bajtos commented May 28, 2020

bajtos commented Mar 9, 2020 •

edited

Loading

tpluscode commented Apr 24, 2020 •

edited

Loading