V0.2.2 - Better Diagnostics

[DaemonF0rge]

This pull request introduces several improvements to the Enfusion Script language server, focusing on enhanced parsing and diagnostics, better handling of file events, and support for additional language features. The most significant changes include improved detection of symbols and types within function bodies for more accurate diagnostics, robust handling of file closures and deletions, and parser enhancements for variable declarations and enum member separation.

Parser and Symbol Analysis Improvements:

• Added tracking of type references (bodyTypeRefs) and standalone identifier references (bodyIdentifierRefs) within function bodies in the AST, enabling more precise unknown-symbol diagnostics and cross-module checks. [1] [2] [3] [4] [5] [6] [7] [8]
• Enhanced parser logic to correctly handle comma-separated variable declarations (e.g., float x, y, z;), including initializer expressions, and to avoid false positives inside function call arguments. [1] [2] [3]

Diagnostics and File Event Handling:

• Improved handling of file closure and deletion events: when a file is closed and no longer exists on disk, its symbols are immediately removed from the index and diagnostics are cleared, preventing duplicate warnings. [1] [2]
• When re-indexing or revalidating open files, diagnostics are now properly cleared for files that are no longer part of the workspace, ensuring users do not see stale warnings. [1] [2]

Parser Robustness and Language Support:

• Added support for the thread keyword in the lexer’s keyword set, improving language coverage.
• Updated enum member parsing to flag and recover from semicolon-separated members, enforcing correct comma separation per language rules.

Miscellaneous:

• Bumped extension version to 0.2.2 in package.json.
• Added debug output for server startup and diagnostics in debug-output.txt.

[Copilot]

Pull request overview

This PR updates the Enfusion Script language server to improve parsing fidelity and diagnostic accuracy, especially for symbol/type usage within function bodies and for handling workspace file lifecycle events.

Changes:

• Extended the parser AST to track bodyTypeRefs (static access targets) and bodyIdentifierRefs (standalone identifiers) inside function bodies, enabling new/stronger unknown-symbol and cross-module checks.
• Improved local variable detection in function bodies (comma-separated declarations, constructor-style locals) and added enum-member separator diagnostics (semicolon vs comma).
• Hardened diagnostics behavior around file close/delete and workspace revalidation to avoid stale/duplicate diagnostics.

Reviewed changes

Copilot reviewed 12 out of 13 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
server/src/analysis/ast/parser.ts	Adds function-body reference tracking and improves local/enum parsing behavior.
server/src/analysis/project/graph.ts	Uses new body refs for unknown-symbol and cross-module checks; adds access-modifier diagnostics.
server/src/lsp/handlers/diagnostics.ts	Cleans up index + diagnostics on close when the backing file no longer exists.
server/src/index.ts	Clears diagnostics for open docs that are no longer workspace files during reindex/revalidate.
server/src/analysis/lexer/rules.ts	Adds thread to keyword set.
test/parser.test.ts	Adds broad parser coverage for comma-chain locals, enum separators, and bodyIdentifierRefs behavior.
test/module-visibility.test.ts	Adds cross-module diagnostics coverage for static call targets.
test/features.test.ts	Adds regression coverage for diagnostics + modded class resolution + access checks.
test/test_enscript.c	Removes old test fixture file from root.
test/3_game/test_enscript.c	Adds a larger script fixture under module folder.
package.json	Bumps extension version to 0.2.2.
debug-output.txt	Adds captured debug output from a local run.
.gitignore	Adds /docs ignore entry.

Review notes (issues found):

• server/src/analysis/project/graph.ts:
  • checkBodyIdentifierRefs does a full docCache scan per identifier ref (for (const [, fileAst] of this.docCache) { ... } around lines ~7146–7167). This is redundant given the existing fast indexes (globalSymbolIndex, classIndex, enumIndex, functionIndex) and can become a major performance bottleneck once the 500-file threshold is reached. Consider removing the scan and relying on the indexes, or building a single Set of top-level names once per run and querying that.
  • NATIVE_ENGINE_CONSTANTS contains multiple duplicates (e.g., UAGear, UAUIGesturesOpen, UAUIQuickbarToggle, DT_CUSTOM in the ~54–93 region). Duplicates won’t break correctness (they’re later put into a Set), but they add maintenance noise—dedupe the list to keep it authoritative.
• server/src/analysis/ast/parser.ts:
  • The “STANDALONE IDENTIFIER REFERENCE DETECTION” comment claims identifiers are excluded when “followed by '.' (static access…)”, but the implementation still records the identifier when t.value === '.' (see the block around ~964–987). Either adjust the condition to match the comment, or update the comment to reflect the actual behavior (especially since tests expect some dot-prefixed cases like SomeClass to be captured).
• debug-output.txt:
  • This looks like a local runtime log (timestamps, local drive paths). Keeping it in-repo can create churn and isn’t stable test data; consider removing it from version control and adding it to .gitignore if it’s meant for local troubleshooting only.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.