Skip to content

Latest commit

 

History

History
109 lines (81 loc) · 3.84 KB

0000-npm-diff.md

File metadata and controls

109 lines (81 loc) · 3.84 KB

npm diff

Summary

Add a new command that enables diff workflows (similar to git diff) for packages published in the registry.

Motivation

  • Package Consumers: Complements npm audit and npm outdated workflows by providing insight on what changed across different versions of a package.
  • Package Authors: Enables diff packlist-tracked-only file changes prior to publishing a new version of a package, or while debugging past changes.

Detailed Explanation

Introduce a new npm diff command that accepts one or more package specs and uses these specs to compare files and print diff patches. Contents are fetched using pacote, accepting any type of specs that the npm installer understands.

npm diff --diff=<spec>

Using a single <spec> allows users to retrieve diff output between an existing valid version of a package with the same name found in the actual dependency tree and the exact <spec> match from the registry.

npm diff --diff=<spec-a> --diff=<spec-b>

Fetches/reads contents of two versions of a package, represented by <spec-a> and <spec-b> and prints diff output.

npm diff (no arguments)

Meant as a helper tool to package authors, prints diff output between the current file system dirs/files (tracked by packlist) and the last published version of that package.

Rationale

  • A very common workflow is using npm outdated to figure out what dependencies need update and then manually reviewing what changed in between the current version you have in your project and whatever you can update to. That workflow involves multiple steps, some of them being extra mental hurdle (such as keeping track of the semver versions change and their meaning) some other very manual such as jumping around to repos, scanning through changelogs, veryfing semver contract is respected, etc.
  • It feels very natural to have a npm diff from the point of view of a user of git diffs - It does provide much more quick insight and transparency on code that makes its way into our projects in a (medium) that most developers are already familiar with.

Alternatives

  • Not have a native npm diff command and deffer to userland tools.

Implementation

Parses user args and reads package contents for two different specs and run a diff of each file, providing an usable patch output for users to work with.

Parsing diff arguments logic

What should happen if using npm diff (no args) ? What should be the expected behavior when using npm diff --diff=<pkg> ? What if I use two --diff=<pkg>@<version> arguments?

Here's some of the sketch ideas that were initially put together to try and answer these questions:

  • IF NO arg:
    • READ package.json name
      • IF FOUND:
        • RETURN package-json-name@tag - file:.
      • ELSE:
        • THROW usage
  • IF arg is package-name only:
    • READ arborist load actual tree AND PARSE package-name
      • IF NOT SPEC TYPE REGISTRY:
        • THROW unsupported spec type
      • IF FOUND:
        • RETURN package-name@FOUND_VERSION
      • ELSE:
        • RETURN package-name@latest
  • IF arg is validl semver range|version:
    • READ package-name from either arg1 or arg2
      • IF FOUND:
        • RETURN other-arg-parsed-name@arg
      • ELSE:
        • READ package.json name
          • IF FOUND:
            • RETURN package-json-name@arg
          • IF NO package.json:
            • THROW error need to run from package dir to user versions

This argument parsing part of the implementation lives in the npm cli:

Generating patch diff contents

Prior Art