fix: match bare wildcard FQN selectors against nested leaf names

[itsnamangoyal]

Problem

CORE-937: a bare wildcard FQN selector (e.g. -s "f_*", no dot) fails to match models nested in subfolders — it only matches root-level nodes.

Repro:

dbt ls -s "f_*"   # 74 real f_* models under marts/... -> "No nodes selected!"
dbt ls -s "d_*"   # single root-level d_ model -> matches fine

Root cause

is_selected_node() in core/dbt/graph/selector_methods.py flattens the node's FQN and fnmatchs the entire remaining dotted string against the selector once it hits a wildcard token. For a bare pattern like f_* (no dot), this requires the whole flattened FQN tail (e.g. marts.subdir.f_model) to start with f_, which only holds for root-level nodes whose FQN has no folder segments before the leaf name.

Confirmed this behavior is identical (byte-for-byte function) across 1.10.latest, 1.11.latest, and 1.12.latest — every current v1 branch is affected. dbt Fusion (v2) has carried an explicit "bare wildcard-leaf shortcut" in node_selector.rs since May 2025 (an intentional Fusion-only addition, not something ever ported back to Python) — that's the actual origin of the parity gap, not the PR originally cited in the customer-facing Slack thread (which turned out to be an unrelated path:-selector directory-matching fix).

Fix

In the wildcard-matching branch of is_selected_node, if the existing flatten-and-fnmatch check doesn't match, and the selector is a bare wildcard (no dot), fall back to matching against just the FQN's leaf segment. This is additive — the original whole-FQN match (which also supports matching folder names, e.g. *unions* matching a directory literally named unions) still runs first and is unchanged.

Testing

• Added test_qualified_name_bare_wildcard_matches_root_leaf (already passed pre-fix) and test_qualified_name_bare_wildcard_matches_nested_leaf (red before, green after) to tests/unit/graph/test_selector_methods.py, exercised through QualifiedNameSelectorMethod.node_is_match — the real production entry point (which also retries against the package-unscoped FQN).
• Full tests/unit/graph/ + tests/unit/test_graph_selection.py suites (266 tests) pass with no regressions.
• Live-verified against jaffle-shop with the actual dbt v1 CLI:
  • Before fix: dbt ls -s "stg_*" against the nested models/staging/ folder → No nodes selected!
  • After fix: same command → all 6 stg_* models + their generic tests correctly selected
  • Confirmed no change to path:, dotted-FQN prefix, and exact-match selector forms.

Test plan

• Unit tests pass (tests/unit/graph/test_selector_methods.py, tests/unit/test_graph_selection.py)
• Manually reproduced against a live project (jaffle-shop) with and without the fix
• Changelog entry added via changie new

[Copilot]

🔵 Needs a closer look

Nested versioned models remain unfixed; add matching support and regression coverage.

Pull request overview

Fixes bare wildcard FQN selectors so they match nested model leaf names.

Changes:

• Adds leaf-level fallback matching for bare wildcards.
• Adds regression tests for root and nested models.
• Adds a changelog entry.

File summaries

File	Description
tests/unit/graph/test_selector_methods.py	Adds wildcard selector regression tests.
core/dbt/graph/selector_methods.py	Implements nested leaf matching.
.changes/unreleased/Fixes-20260917-224607.yaml	Documents the fix.

Review details

Suppressed comments (1)

core/dbt/graph/selector_methods.py:119

• This guard leaves nested versioned models unfixed: for an FQN such as ["my_pkg", "marts", "f_model", "v1"], f_* still fails the whole-FQN fnmatch, and the leaf fallback is skipped because is_versioned is true. The fallback should match the model-name segment (fqn[-2]) for versioned nodes, or otherwise explicitly preserve the version suffix while matching the bare pattern; please add a regression test for this case.

        if not is_versioned and "." not in node_selector:
            return fnmatch(fqn[-1], node_selector)

• Files reviewed: 3/3 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🟡 Changes recommended

The fallback still excludes nested versioned models from bare wildcard matching.

Get a fresh assessment by requesting another Copilot review.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 1
• Review effort level: Balanced

[Copilot]

🟢 Approval recommended

The focused implementation preserves existing matching behavior and covers the affected cases with regression tests.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 0 new
• Review effort level: Balanced