builtins.getFlake: Support path values

[edolstra]

Motivation

This allows doing

builtins.getFlake ./..

instead of the hacky

builtins.getFlake (builtins.flakeRefToString { type = "path"; path = self.sourceInfo.outPath; narHash = self.narHash; });

Depends on #337.

Issue #302.

Context

Summary by CodeRabbit

• New Features

  • New ways to lock a flake using a directory or explicit flake input, giving more flexible lock workflows.

• Refactor

  • Optimized flake resolution to handle path-based and string-based inputs with shared preparatory logic, preserving behavior.

• Tests

  • Added functional coverage for the new path-based flake resolution alongside the legacy method.

[github-actions]

🚀 Deployed on https://6986249bd4ffe2380fbdd38e--determinate-nix.netlify.app

[coderabbitai]

📝 Walkthrough

Walkthrough

Refactors builtins.getFlake to branch on argument type (path vs non-path) with shared LockFlags preparation, and adds two new lockFlake overloads (FlakeRef and SourcePath) that delegate to a primary overload accepting an explicit Flake.

Changes

Cohort / File(s)	Summary
Flake resolution (primops) src/libflake/flake-primops.cc	getFlake now prepares LockFlags once, branches on argument type: path branch realises the SourcePath and calls flake path-based logic; non-path branch forces the value and performs string-based FlakeRef parsing and validation using the shared LockFlags.
Locking API / core flake logic src/libflake/flake.cc, src/libflake/include/nix/flake/flake.hh	Introduces primary lockFlake overload that accepts an explicit Flake parameter and two delegating overloads: one taking FlakeRef and one taking SourcePath (flakeDir). Delegating overloads compute or read the Flake and call the primary overload.
Tests tests/functional/flakes/get-flake.sh	Adds a path-based getFlake test (builtins.getFlake ./..) alongside existing string-based test and exposes both outputs to verify both access patterns evaluate the same value.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• Make path inputs lazy #312 — touches lockFlake/getFlake interactions and lock-file cache semantics overlapping this PR's API and flow changes.
• builtins.getFlake fixes #337 — modifies builtins.getFlake and related tests; closely related to the path vs string branching work.

Suggested labels

flake-regression-test

Suggested reviewers

• cole-h

Poem

🐰 Hopping through paths and strings so neat,
LockFlags pooled once — what a rhythmic beat.
Overloads align and small changes wink,
A flake finds its home with a happy blink.

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The pull request title 'builtins.getFlake: Support path values' clearly and concisely summarizes the main feature addition: enabling path values to be passed to the builtins.getFlake function.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch getflake-path

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@src/libflake/flake.cc`:
- Around line 959-962: The overload of lockFlake that constructs a fakeRef via
parseFlakeRef("flake:dummy") causes topRef.input.getSourcePath() to be null and
prevents writing lockfiles for local flakes; change the overload (the function
that currently creates fakeRef and calls readFlake) to detect when flakeDir is a
local/rootFS path and construct a path-based FlakeRef whose input.sourcePath is
flakeDir (instead of parseFlakeRef("flake:dummy")), or alternatively ensure
callers pass writeLockFile=false or an explicit outputLockFilePath; update the
call site that uses fakeRef, readFlake, and lockFlake to use the new path-based
FlakeRef so writeLockFile logic can find a valid topRef.input.getSourcePath().

🧹 Nitpick comments (1)

src/libexpr/primops.cc (1)

2371-2374: Avoid double symlink resolution in __narHash.

realisePath already defaults to full symlink resolution, so resolving again adds extra I/O. Consider passing path directly to fetchToStore2.

♻️ Proposed tweak

-            auto hash =
-                fetchToStore2(state.fetchSettings, *state.store, path.resolveSymlinks(), FetchMode::DryRun).second;
+            auto hash =
+                fetchToStore2(state.fetchSettings, *state.store, path, FetchMode::DryRun).second;

[cole-h]

Otherwise LGTM.