#265 — render: macro lens section MVP (Slice B)

.github/workflows/ci.yml

@@ -341,6 +341,7 @@ jobs:
     #   - 26 after review_scope_variants (cute-dbt#302)
     #   - 27 after review_pr_anchor (cute-dbt#303)
     #   - 28 after agent_skill (cute-dbt#304)
+    #   - 29 after macro_perspective (cute-dbt#265)
     # The README is not a feature. The job name deliberately carries no
     # number (cute-dbt#68 — the old `==6` display label went stale across
     # several bumps): the enforced count lives ONLY in the `expected=`
@@ -358,7 +359,7 @@ jobs:
       - name: features/*.feature count matches expected
         run: |
           set -euo pipefail
-          expected=28
+          expected=29
           count=$(find features -maxdepth 1 -name '*.feature' | wc -l | tr -d ' ')
           echo "feature-count: expected ${expected}, found ${count}"
           if [ "$count" != "$expected" ]; then

features/macro_perspective.feature

@@ -0,0 +1,62 @@
+# Maps: cute-dbt#265 (epic #265 — the macro perspective lens). A macro
+# edit is invisible to cute-dbt's model-and-unit-test scope selection: a
+# macros/*.sql file matches no model `original_file_path` and macros are
+# not unit-test targets, so a macro change slips through entirely. Slice B
+# gives the report a "macro changed" section — the changed macro's body
+# diff, the count of root-project models it reaches (the reverse
+# `macro_blast_radius`), and those models as a collapsible directory tree.
+#
+# cute-dbt#265 Slice B (epic #288 default-OFF posture): the whole macro
+# lens is gated behind the `macro-lens` experiment. Every scenario
+# asserting the section opts in via the experimental-switch Given
+# (CUTE_DBT_EXPERIMENTAL on the subprocess); the switch-off scenario at
+# the bottom pins the default posture (no section even with a macro edit,
+# byte-stable non-macro goldens). The section copy says "macro changed",
+# NEVER a `state:modified.macros` selector name (critique S2): the
+# macro-body-changed scope is a synthesized class, not a dbt state
+# selector.
+Feature: A changed macro is called out, never silently invisible
+  As a PR reviewer
+  I want a macros/*.sql edit surfaced with the models it reaches
+  So that a macro change is part of the review, not an invisible side door
+
+  Scenario: A macro edit in pr-diff mode renders the section with a body diff and the impacted-model tree
+    Given a current manifest with a root-project macro called by two models
+    And the working tree carries that macro's source file
+    And the PR diff edits the macro's body
+    And the experimental switch enables macro-lens
+    When I run cute-dbt report in pr-diff mode against the macro patch
+    Then the exit code is 0
+    And the report carries the macro-lens section
+    And the macro-lens section names the changed macro
+    And the macro-lens section carries the macro body diff
+    And the macro-lens section reports the impacted-model count as 2
+    And the macro-lens section lists both impacted models in the directory tree
+    And the macro-lens fidelity chip reads "heuristic"
+    And the macro-lens section never names a "state:modified.macros" selector
+
+  Scenario: No macro edit leaves the section absent and the report byte-stable
+    Given a current manifest with a root-project macro called by two models
+    And the working tree carries that macro's source file
+    And the PR diff edits only a model's SQL, not the macro
+    And the experimental switch enables macro-lens
+    When I run cute-dbt report in pr-diff mode against the macro patch
+    Then the exit code is 0
+    And the report carries no macro-lens section
+
+  Scenario: With the macro-lens switch off a macro edit renders no section
+    Given a current manifest with a root-project macro called by two models
+    And the working tree carries that macro's source file
+    And the PR diff edits the macro's body
+    When I run cute-dbt report in pr-diff mode against the macro patch
+    Then the exit code is 0
+    And the report carries no macro-lens section
+
+  Scenario: A vendor-package macro edit is filtered out of the lens
+    Given a current manifest with a vendor-package macro called by a root-project model
+    And the working tree carries that macro's source file
+    And the PR diff edits the macro's body
+    And the experimental switch enables macro-lens
+    When I run cute-dbt report in pr-diff mode against the macro patch
+    Then the exit code is 0
+    And the report carries no macro-lens section

lefthook.yml

@@ -89,7 +89,9 @@ pre-push:
     feature-count:
       # Asserts the .feature-count invariant; the enforced count lives ONLY
       # in the `expected=` line below (cute-dbt#68 — numbers in labels and
-      # comments go stale). History: bumped to 28 in cute-dbt#304 when
+      # comments go stale). History: bumped to 29 in cute-dbt#265 when
+      # macro_perspective.feature joined the contract; previously
+      # bumped to 28 in cute-dbt#304 when
       # agent_skill.feature joined the contract; previously
       # bumped in cute-dbt#303 when review_pr_anchor.feature joined; previously
       # bumped in cute-dbt#302 when review_scope_variants.feature joined; previously
@@ -111,7 +113,7 @@ pre-push:
       # Mirrors the `feature-count` CI job; the README is not a feature.
       run: |
         set -euo pipefail
-        expected=28
+        expected=29
         count=$(find features -maxdepth 1 -name '*.feature' | wc -l | tr -d ' ')
         if [ "$count" != "$expected" ]; then
           echo "feature-count: expected ${expected} .feature files, found ${count}"