Support diffing against a non-existing composer.lock file

[wikando-ck]

Summary

When comparing against a composer.lock that doesn't exist yet (e.g. a path on a branch where no lockfile has been committed), the diff now treats it as an empty lockfile ({}) instead of throwing an error. All packages in the target are reported as new installs.

This covers two scenarios:

• Local file path — the path looks like a composer.lock file but doesn't exist on disk (e.g. composer diff /new/project/composer.lock composer.lock)
• Git ref path — the file doesn't exist in the referenced commit (e.g. composer diff HEAD:new-dir/composer.lock)

Motivation

We use composer-bin to manage tooling in separate directories away from the main composer.json/composer.lock. When a new tool directory is introduced on a branch, there is no corresponding composer.lock on the base branch yet. Running composer diff against that non-existing base lockfile currently throws an error, making it impossible to review the newly added dependencies.

Changes

• Handle non-existing local composer.lock paths by returning {} instead of falling through to git
• Detect missing-file errors from git show and return {} instead of throwing
• Add tests for both local and git-ref missing lockfile scenarios
• Document the usage in README

[IonBazan]

Thanks for the bug report and your contribution. Could you help to check why the Windows build is failing? This must have something to do with using : as both git separator and disk identifier.
Please ignore Composer v1 failures for now.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (cb9542e) to head (c8e7b3e).

Additional details and impacted files

@@             Coverage Diff             @@
##                main       #46   +/-   ##
===========================================
  Coverage     100.00%   100.00%           
- Complexity       242       247    +5     
===========================================
  Files             22        22           
  Lines            611       632   +21     
===========================================
+ Hits             611       632   +21     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[wikando-ck]

Fixed the windows issue and covered that by tests.
Squashed a fix for the duplicate empty line to fix the styleci into the first commit.

[IonBazan]

Thanks for following it up and sorry for late review. This looks almost good but needs some clarification (see below).

[IonBazan]

This will cause decoding the json twice. Also, if the file exists but is not a well-formed JSON, I wouldn't consider it a skippable IO error and throw exception here. Let me know what you think.

[IonBazan]

Somehow there is an escaped mutant here now 🤔

[IonBazan]

Previously there was a GIT_SEPARATOR strpos() check. Is it replaced by trial and error instead to better handle Windows absolute paths?