feat: add AL (.al) language support

[PlaySkyHD]

feat: add AL (.al) language support

What

Adds extraction support for Microsoft Dynamics 365 Business Central AL (.al) source files, including production applications and dedicated AL test applications.

On Python 3.12+, the optional tree-sitter-al parser provides full structural and semantic extraction. On other supported Python versions, or when the parser is absent, graphify falls back to a built-in structural scanner for files, application objects, procedures, and triggers.

Why

Business Central extensions are currently invisible to graphify because .al is not treated as source code. This PR makes AL a first-class language while preserving graphify's Python 3.10+ compatibility and offline extraction behavior.

What it extracts

• Business Central application objects, extensions, and permission sets
• Procedures, overloads, triggers, parameters, fields, and enum values
• Distinct field/action triggers even when names such as OnValidate repeat
• Namespaces, using declarations, attributes, and visibility
• Extension and interface implementation relationships
• Typed cross-file calls and object references
• Event subscriber and enum implementation relationships
• In-corpus app.json application dependencies
• TestPage targets and HandlerFunctions bindings in AL test apps

Test application semantics

• TestPage "X" declarations resolve to visible production page objects through references edges with test_target context.
• [HandlerFunctions('X')] entries resolve to unique procedures in the same test codeunit through references edges with test_handler context.
• TestPage runtime APIs such as OpenView, SetValue, Invoke, and AssertEquals are not fabricated as calls to undeclared production procedures.
• Test source remains ordinary code in the graph, preserving [Test], handler, and event attributes.

Details

• Preserves AL declaration spelling while resolving identifiers case-insensitively with casefold().
• Uses application, namespace, object kind, and containment context to avoid ambiguous cross-file matches.
• Resolves overloaded procedures conservatively by signature identity and call arity; same-arity ambiguity remains unresolved.
• Reads optional app.json metadata for application and dependency boundaries without requiring a valid manifest.
• Distinguishes missing parser dependencies from parser load failures in diagnostics.
• Keeps fallback extraction deliberately structural and does not fabricate semantic edges.

Validation corpora

AcmeComments

Tested against the 11-file Business Central extension:

• 60 nodes
• 89 edges
• No failed sources or syntax errors

InterestMgt and InterestMgtTest

Tested against both applications together:

• 31 detected code/manifest files
• 211 nodes and 250 edges
• 25/25 AL application objects represented, including the permission set
• 48/48 production procedures and 34/34 triggers represented
• Both OpenInterestCalculator overloads preserved; the one-argument call resolves correctly
• 13 resolved TestPage target edges
• 1 resolved test handler binding
• 1 resolved test-app-to-production-app dependency
• No fabricated TestPage runtime calls

Changes

• graphify/extractors/al.py - dedicated Tree-sitter and fallback AL extractor, permission sets, scoped members, overload identities, and test facts
• graphify/al_resolution.py - ambiguity-safe cross-file, test target, handler, overload, and application dependency resolution
• graphify/detect.py / graphify/extract.py - .al detection, dispatch, diagnostics, and resolver registration
• pyproject.toml / uv.lock - optional [al] extra using tree-sitter-al on Python 3.12+
• tests/test_al.py / AL fixtures - focused extraction, fallback, diagnostics, resolution, test-app, and overload coverage
• README.md / CHANGELOG.md - language support and installation documentation

Tests

• uv run pytest tests/test_al.py -q - 17 passed
• uv run pytest tests/test_detect.py -q - 243 passed, 11 skipped
• Affected extraction/language regression modules - 629 passed, 21 skipped
• Windows long-path regression rerun with a short temp root - 1 passed
• uv run ruff check graphify/extractors/al.py graphify/al_resolution.py tests/test_al.py

Notes / possible follow-ups

• Full Tree-sitter extraction requires Python 3.12+ because that is the minimum supported version of the parser package; structural fallback remains available on graphify's older supported Python versions.
• External Microsoft test-library objects remain unresolved unless their source applications are included in the corpus.
• .app package binary parsing remains intentionally out of scope.

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 5 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) as a supported language: introduces graphify/al_resolution.py for application-level, case-insensitive cross-file symbol resolution (objects, procedures, triggers, fields, enum values, extensions, interfaces, typed calls, event subscriptions) plus AL extractor/detector wiring and a new [al] extra with a Tree-sitter path on Python 3.12+ and a structural fallback elsewhere. Documents the new grammar and extra in the README (now 38 grammars) and CHANGELOG, and adds a test covering AL corpus extraction continuing past a failing file.

Worth a look

• Source-file comparison only handles one absolute/relative direction — graphify/al_resolution.py:19 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Uppercase AL files are not classified as code — graphify/detect.py:45 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• AL fallback omits al_facts from extract_al return schema — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• AL fallback returns callable parameters as string instead of metadata list — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Fallback callable metadata uses string parameters vs list in tree-sitter path — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2267 functions depend on the 596 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 481 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 40 more — each is listed as a finding

Verification — 2267 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2072 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 3 grounded finding(s) anchored inline below; 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 5 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) extraction as the 38th tree-sitter grammar, with a new graphify/al_resolution.py module that does application-level, case-insensitive cross-file symbol resolution (objects, permission sets, procedures/overloads, triggers, fields, enums, extensions, interfaces, typed calls, event subscriptions, app dependencies, and TestPage/test-handler bindings). Wires up the optional [al] extra for full Tree-sitter extraction on Python 3.12+ and a structural object/procedure/trigger fallback elsewhere, plus tests covering the resolver, invalid/missing manifests, and the optional-extra parser path. Updates README (supported types table, prerequisites) and CHANGELOG accordingly.

Worth a look

• _same_source suffix match can produce false positives across distinct files — graphify/al_resolution.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• AL fact ids collide across files — graphify/al_resolution.py:106 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Fallback AL callable IDs collide for repeated triggers — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• masked.find with limited end may miss object opening brace — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Fallback AL object regex omits permissionset objects — graphify/extractors/al.py:10 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2270 functions depend on the 599 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 483 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 40 more — each is listed as a finding

Verification — 2270 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2075 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 3 grounded finding(s) anchored inline below; 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 1 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) extraction: a new graphify/al_resolution.py performs application-level cross-file symbol resolution (objects, permission sets, procedures/overloads, triggers, fields, enums, extensions, interfaces, typed calls, event subscriptions, app dependencies, and TestPage/handler bindings) with case-insensitive matching and app.json manifest awareness, backed by new detectors and extractors (graphify_detect_*, graphify_extractors_al_*). Registers .al as the 38th grammar with an optional [al] extra and a structural fallback for non-3.12 Python, updating README supported-types and the CHANGELOG.

Worth a look

• Fallback AL extractor collapses duplicate trigger members — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2326 functions depend on the 655 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 483 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2326 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2131 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 5 advisory finding(s) below merit a look before merge.

---

Graphify review — findings

Adds Business Central AL (.al) extraction as the 38th grammar, spanning a new graphify/al_resolution.py application-level resolver (_ALSymbolResolver with manifest/dependency, object, member, and event-subscriber mapping) and AL extractor helpers (Tree-sitter parsing plus a structural fallback for non-3.12 Pythons). Registers the [al] optional extra and documents it in the README supported-types table and CHANGELOG. Includes AL tree-sitter tests asserting callable/field metadata preservation.

Worth a look

• target selected as candidates[0] after set() dedup on unhashable dicts — graphify/al_resolution.py · Escalate · high
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• opening brace search fails when '{' is outside match span — graphify/extractors/al.py · Escalate · high
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Whitespace not stripped from split handler names — graphify/al_resolution.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• _same_source suffix match can produce false positives across files — graphify/al_resolution.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Aggregate extractors package now requires optional AL extractor — graphify/extractors/__init__.py:13 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2327 functions depend on the 656 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 483 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2327 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2132 function(s) in the blast radius were not formally verified this run

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 3 advisory finding(s) below merit a look before merge.

---

Graphify review — findings

Adds Business Central AL (.al) extraction with a new al_resolution.py application-level resolver and extractors/al.py, wiring the extractor into extractors/__init__.py and the language-family/detection paths. Models objects, permission sets, procedures/overloads, triggers, fields, enums, extensions, interfaces, typed calls, event subscriptions, and test-app relationships with case-insensitive cross-file resolution and app.json-driven dependency edges; full Tree-sitter parsing requires the optional [al] extra on Python 3.12+ with a structural fallback otherwise. Documents the new grammar in the README and CHANGELOG and adds AL parser fallback/read-error tests.

Worth a look

• ImportError from tree_sitter (not tree_sitter_al) misreported as fallback path or wrong error — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Comment masker treats // inside quoted AL identifiers as a real comment — graphify/extractors/al.py:57 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Only the first implemented interface is recorded — graphify/extractors/al.py:268 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2330 functions depend on the 659 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 483 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2330 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2135 function(s) in the blast radius were not formally verified this run

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 4 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) extraction as the 38th tree-sitter grammar, introducing graphify/al_resolution.py for application-level cross-file symbol resolution (objects, procedures, fields, enums, extensions, events, app dependencies, and test-app bindings) plus new AL extractors. Documents the optional [al] extra in the README support table and language list, with a structural fallback for non-3.12 Python. Records the feature in the 0.9.48 changelog.

Worth a look

• _matching_brace ignores braces inside masked/unmasked strings and comments — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Fallback skipped when tree_sitter core import fails — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Regex fallback omits permissionset objects — graphify/extractors/al.py:10 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• AL optional-extra guidance is codified with misspelled package name — tests/test_al.py:50 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2332 functions depend on the 661 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 484 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2332 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2137 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 3 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) extraction as the 38th tree-sitter grammar, introducing graphify/al_resolution.py with an _ALSymbolResolver that maps object/member facts and resolves cross-file references (calls, events, extensions, app dependencies, TestPage/handler bindings) case-insensitively. Registers the optional [al] extra with a structural object/procedure/trigger fallback for Python versions without the Tree-sitter grammar, including UTF-8 BOM handling. Updates README extension tables and adds the 0.9.48 changelog entry.

Worth a look

• Malformed non-object app.json crashes AL resolution — graphify/al_resolution.py:52 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Fallback AL object regex omits permissionset objects — graphify/extractors/al.py:11 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• _matching_brace ignores masking and counts braces inside strings/comments — graphify/extractors/al.py:128 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2333 functions depend on the 662 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 484 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2333 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2138 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 2 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) extraction as a new tree-sitter grammar with an optional [al] extra, wiring it into graphify.detect (filetype detection, office/import paths) and the extractor registry, with a structural object/procedure/trigger fallback on Python versions lacking the Tree-sitter parser. Introduces graphify/al_resolution.py (_ALSymbolResolver) for application-level, case-insensitive cross-file symbol resolution — objects, members, extensions, typed calls, event subscriptions, TestPage/handler bindings, and app.json dependency edges. Documents the new grammar and extra in README/CHANGELOG and adds AL extraction tests.

Worth a look

• extract_al returns an error instead of falling back when the installed AL parser fails — graphify/extractors/al.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Valid AL permission set extensions are not recognized — graphify/extractors/al.py:10 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2339 functions depend on the 668 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 485 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2339 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2144 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 2 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL (.al) as a supported language, with full Tree-sitter extraction via the optional [al] extra and a structural fallback extractor for other Python versions. Introduces graphify/al_resolution.py for application-level, case-insensitive cross-file symbol resolution (objects, members/overloads, extensions, app-manifest dependencies) and adds AL extractor/resolver tests. Updates README (grammar count 37→38, extras table, extension list) and the CHANGELOG.

Worth a look

• Fallback flag overwritten across files of same extension — graphify/extract.py:5704 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Drops all but the first interface in implements clauses — graphify/extractors/al.py:323 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2342 functions depend on the 671 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 485 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2342 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2147 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 2 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Microsoft Dynamics 365 Business Central AL (.al) extraction with a new al_resolution.py application-level resolver and al extractor covering objects, permission sets, extensions, procedures/overloads, triggers, fields, enums, interfaces, typed calls, event subscriptions, app dependencies, and TestPage/handler bindings. Registers the optional [al] extra (full Tree-sitter on Python 3.12+, structural fallback otherwise) and wires app.json manifest resolution with case-insensitive cross-file matching. Updates the README grammar count to 38, adds the extension/extra rows, and adds a 0.9.48 CHANGELOG entry.

Worth a look

• Comment masking starts inside quoted AL identifiers — graphify/extractors/al.py:65 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Optional tree_sitter dependency is required by fallback test — tests/test_al.py:127 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2564 functions depend on the 1033 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 487 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2564 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2369 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds first-class Business Central AL (.al) extraction: a new graphify/al_resolution.py with _ALSymbolResolver for application-level, case-insensitive cross-file symbol resolution (objects, members, app.json manifest dependencies), plus AL extractor internals (altreecontext, al_extract_member, mask_al_code) and the [al] optional extra with a Tree-sitter path on Python 3.12+ and a structural fallback elsewhere. Registers .al in the supported-language surface (README extension table now 38 grammars, environment-variable/extras docs) and adds the 0.9.48 changelog entry. Includes extensive AL parser and extraction tests covering fallback specs, optional-extra reporting, case-insensitive detection, and calls/handler edges.

No blocking issues surfaced. 21 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2567 functions depend on the 1036 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 487 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2567 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2372 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds a Business Central AL (.al) extractor with full Tree-sitter extraction under the optional [al] extra on Python 3.12+ and a structural object/procedure/trigger fallback otherwise, plus the new graphify/al_resolution.py for case-insensitive application-level symbol resolution (objects, members, extensions, ControlAddIns, event bindings, and app-manifest dependencies). Registers .al as the 38th grammar and documents the extra in the README and CHANGELOG. Includes extensive tests covering fallback paths, namespace/manifest resolution, and test-dispatch handling.

No blocking issues surfaced. 22 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2572 functions depend on the 1041 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 488 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 39 more — each is listed as a finding

Verification — 2572 functions in the blast radius were not formally verified this run (proofs are advisory here).

Health delta baseline: last indexed commit 282976b (diverged from this PR's base — delta is approximate).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2377 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 2 grounded finding(s) anchored inline below; 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 1 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Microsoft Dynamics 365 Business Central AL (.al) extraction as the 38th tree-sitter grammar behind an optional [al] extra, modeling objects, permission sets/extensions, ControlAddIns and UserControls, procedures/overloads, scoped triggers, fields, enums, typed calls, event bindings/subscriptions, app dependencies, and test-app relationships via _ALSymbolResolver with case-insensitive cross-file resolution. Falls back to structural object/procedure/trigger extraction through ALFallbackExtractor when the Tree-sitter grammar is unavailable (non-3.12+ or extra not installed). Wires .al into detection via is_graphable_source and friends and updates the README/CHANGELOG accordingly.

Worth a look

• Missing-parser diagnostic test expects misspelled package extra — tests/test_al.py:94 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2576 functions depend on the 1045 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 488 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2576 functions in the blast radius were not formally verified this run (proofs are advisory here).

Health delta baseline: last indexed commit 282976b (diverged from this PR's base — delta is approximate).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2381 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 3 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds full Business Central AL (.al) support as the 38th tree-sitter grammar, with a new al optional extra and ALFallbackExtractor that provides structural object/procedure/trigger extraction when Tree-sitter isn't available (Python < 3.12 or extra not installed). Extracts and emits application objects, permission sets/extensions, ControlAddIns, UserControls, procedures/overloads, scoped triggers, fields, enum values, interfaces, typed calls, event bindings/subscriptions, and TestPage/handler relationships, then resolves them across files case-insensitively via _ALSymbolResolver, using app.json manifests for application identity and dependency edges. Member resolution requires a unique candidate (_unique_member_id) matched by parent, signature, and source location, so ambiguous string ids or duplicated members are left unlinked rather than guessed.

Worth a look

• AL resolution is order-dependent across files — graphify/al_resolution.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Dependency warning is ignored when an extractor also reports an error — graphify/extract.py:5696 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Quoted-identifier masking corrupts identifier characters via wrong handler dispatch — graphify/extractors/al.py:60 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2576 functions depend on the 1045 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 488 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2576 functions in the blast radius were not formally verified this run (proofs are advisory here).

Health delta baseline: last indexed commit 282976b (diverged from this PR's base — delta is approximate).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2381 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds Business Central AL as a first-class language: .al files now extract application objects, permission sets/extensions, ControlAddIns and hosted UserControls, procedures/overloads, scoped triggers, fields, enums, interfaces, and their typed calls, event bindings/subscriptions, app dependencies, and test-app relationships, with case-insensitive cross-file resolution via _ALSymbolResolver. Wires up manifest-aware application context by walking app.json up the tree to attach app ids/names and emit dependency edges. Full Tree-sitter extraction requires the optional [al] extra on Python 3.12+; other versions fall back to structural object/procedure/trigger extraction.

No blocking issues surfaced. 11 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2579 functions depend on the 1048 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 490 callers, 42 callees
• new: _rebuild_code() — 98 callers, 50 callees
• new: detect() — 108 callers, 15 callees
• new: save_manifest() — 39 callers, 11 callees
• new: extract_files_direct() — 17 callers, 20 callees
• new: extract_xaml() — 19 callers, 17 callees
• new: extract_corpus_parallel() — 26 callers, 11 callees
• new: extract_js() — 80 callers, 3 callees
• …and 37 more — each is listed as a finding

Verification — 2579 functions in the blast radius were not formally verified this run (proofs are advisory here).

Health delta baseline: last indexed commit 282976b (diverged from this PR's base — delta is approximate).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 2384 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract.

The verifier did not have enough to check extract, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `cache_root` is annotated `Path | None` — outside the synthesizable primitive/collection set

· 45 more finding(s) on lines outside this diff (see the check run).