.gitignore

@@ -35,5 +35,3 @@ report.[0-9]_.[0-9]_.[0-9]_.[0-9]_.json
 
 /tmp
 /docs
-
-*.vsix

CLAUDE.md

@@ -200,7 +200,7 @@ function parseExpression(input: string) {
 - **Not in scope** → Parses as `Word("obj.prop")` → compiles to `PUSH 'obj.prop'` (treated as file path/string)
 
 Implementation files:
-- **src/parser/parserScopeContext.ts**: ContextTracker that maintains immutable scope chain
+- **src/parser/scopeTracker.ts**: ContextTracker that maintains immutable scope chain
 - **src/parser/tokenizer.ts**: External tokenizer checks `stack.context` to decide if dot creates DotGet or Word
 - Scope tracking: Captures variables from assignments (`x = 5`) and function parameters (`fn x:`)
 - See `src/parser/tests/dot-get.test.ts` for comprehensive examples

src/parser/shrimp.grammar

@@ -1,6 +1,6 @@
 @external propSource highlighting from "./highlight"
 
-@context trackScope from "./parserScopeContext"
+@context trackScope from "./scopeTracker"
 
 @skip { space | Comment }
 

src/parser/shrimp.ts

@@ -2,7 +2,7 @@
 import {LRParser, LocalTokenGroup} from "@lezer/lr"
 import {operatorTokenizer} from "./operatorTokenizer"
 import {tokenizer, specializeKeyword} from "./tokenizer"
-import {trackScope} from "./parserScopeContext"
+import {trackScope} from "./scopeTracker"
 import {highlighting} from "./highlight"
 const spec_Identifier = {__proto__:null,if:66, null:94, catch:100, finally:106, end:108, else:116, while:130, try:136, throw:140}
 export const parser = LRParser.deserialize({

vscode-extension/.gitignore

@@ -0,0 +1,4 @@
+node_modules
+client/dist
+server/dist
+*.vsix

vscode-extension/server/src/diagnostics.ts

@@ -1,12 +1,12 @@
 import { TextDocument, Position } from 'vscode-languageserver-textdocument'
 import { Diagnostic, DiagnosticSeverity } from 'vscode-languageserver/node'
-import { Tree } from '@lezer/common'
+import { parser } from '../../../src/parser/shrimp'
 import { Compiler } from '../../../src/compiler/compiler'
 import { CompilerError } from '../../../src/compiler/compilerError'
 
-export const buildDiagnostics = (textDocument: TextDocument, tree: Tree): Diagnostic[] => {
+export const buildDiagnostics = (textDocument: TextDocument): Diagnostic[] => {
   const text = textDocument.getText()
-  const diagnostics = getParseErrors(textDocument, tree)
+  const diagnostics = getParseErrors(textDocument)
 
   if (diagnostics.length > 0) {
     return diagnostics
@@ -59,7 +59,9 @@ const unknownDiagnostic = (message: string): Diagnostic => {
   return diagnostic
 }
 
-const getParseErrors = (textDocument: TextDocument, tree: Tree): Diagnostic[] => {
+const getParseErrors = (textDocument: TextDocument): Diagnostic[] => {
+  const tree = parser.parse(textDocument.getText())
+
   const ranges: { start: Position; end: Position }[] = []
   tree.iterate({
     enter(n) {

vscode-extension/server/src/scopeTracker.test.ts

@@ -1,10 +1,10 @@
 import { test, expect, describe } from 'bun:test'
-import { EditorScopeAnalyzer } from './editorScopeAnalyzer'
+import { ScopeTracker } from './scopeTracker'
 import { TextDocument } from 'vscode-languageserver-textdocument'
 import { parser } from '../../../src/parser/shrimp'
 import * as Terms from '../../../src/parser/shrimp.terms'
 
-describe('EditorScopeAnalyzer', () => {
+describe('ScopeTracker', () => {
   test('top-level assignment is in scope', () => {
     const code = 'x = 5\necho x'
     const { tree, tracker } = parseAndGetScope(code)
@@ -146,6 +146,6 @@ end`
 const parseAndGetScope = (code: string) => {
   const document = TextDocument.create('test://test.sh', 'shrimp', 1, code)
   const tree = parser.parse(code)
-  const tracker = new EditorScopeAnalyzer(document)
+  const tracker = new ScopeTracker(document)
   return { document, tree, tracker }
 }

vscode-extension/server/src/scopeTracker.ts

@@ -7,7 +7,7 @@ import { globals } from '../../../src/prelude'
  * Tracks variables in scope at a given position in the parse tree.
  * Used to distinguish identifiers (in scope) from words (not in scope).
  */
-export class EditorScopeAnalyzer {
+export class ScopeTracker {
   private document: TextDocument
   private scopeCache = new Map<number, Set<string>>()
 

vscode-extension/server/src/semanticTokens.ts

@@ -1,13 +1,13 @@
 import { parser } from '../../../src/parser/shrimp'
 import * as Terms from '../../../src/parser/shrimp.terms'
-import { SyntaxNode, Tree } from '@lezer/common'
+import { SyntaxNode } from '@lezer/common'
 import { TextDocument } from 'vscode-languageserver-textdocument'
 import {
   SemanticTokensBuilder,
   SemanticTokenTypes,
   SemanticTokenModifiers,
 } from 'vscode-languageserver/node'
-import { EditorScopeAnalyzer } from './editorScopeAnalyzer'
+import { ScopeTracker } from './scopeTracker'
 
 export const TOKEN_TYPES = [
   SemanticTokenTypes.function,
@@ -28,9 +28,11 @@ export const TOKEN_MODIFIERS = [
   SemanticTokenModifiers.readonly,
 ]
 
-export function buildSemanticTokens(document: TextDocument, tree: Tree): number[] {
+export function buildSemanticTokens(document: TextDocument): number[] {
+  const text = document.getText()
+  const tree = parser.parse(text)
   const builder = new SemanticTokensBuilder()
-  const scopeTracker = new EditorScopeAnalyzer(document)
+  const scopeTracker = new ScopeTracker(document)
 
   walkTree(tree.topNode, document, builder, scopeTracker)
 
@@ -75,7 +77,7 @@ function walkTree(
   node: SyntaxNode,
   document: TextDocument,
   builder: SemanticTokensBuilder,
-  scopeTracker: EditorScopeAnalyzer
+  scopeTracker: ScopeTracker
 ) {
   // Special handling for NamedArgPrefix to split "name=" into two tokens
   if (node.type.id === Terms.NamedArgPrefix) {
@@ -102,7 +104,7 @@ type TokenInfo = { type: number; modifiers: number } | undefined
 function getTokenType(
   node: SyntaxNode,
   document: TextDocument,
-  scopeTracker: EditorScopeAnalyzer
+  scopeTracker: ScopeTracker
 ): TokenInfo {
   const nodeTypeId = node.type.id
   const parentTypeId = node.parent?.type.id

vscode-extension/server/src/server.ts

@@ -3,7 +3,6 @@ import { buildDiagnostics } from './diagnostics'
 import { buildSemanticTokens, TOKEN_MODIFIERS, TOKEN_TYPES } from './semanticTokens'
 import { parser } from '../../../src/parser/shrimp'
 import { Compiler } from '../../../src/compiler/compiler'
-import { Tree } from '@lezer/common'
 import {
   InitializeResult,
   TextDocuments,
@@ -11,23 +10,18 @@ import {
   createConnection,
   ProposedFeatures,
   CompletionItemKind,
-  TextDocumentChangeEvent,
 } from 'vscode-languageserver/node'
 
 const connection = createConnection(ProposedFeatures.all)
 const documents = new TextDocuments(TextDocument)
 documents.listen(connection)
 
-const documentTrees = new Map<string, Tree>()
-
 // Server capabilities
 connection.onInitialize(handleInitialize)
 
 // Language features
 connection.languages.semanticTokens.on(handleSemanticTokens)
-documents.onDidOpen(handleDocumentOpen)
 documents.onDidChangeContent(handleDocumentChange)
-documents.onDidClose(handleDocumentClose)
 connection.onCompletion(handleCompletion)
 
 // Debug commands
@@ -37,7 +31,10 @@ connection.onRequest('shrimp/bytecode', handleBytecode)
 // Start listening
 connection.listen()
 
+// ============================================================================
 // Handler implementations
+// ============================================================================
+
 function handleInitialize(): InitializeResult {
   connection.console.log('🦐 Server initialized with capabilities')
   const result: InitializeResult = {
@@ -59,40 +56,21 @@ function handleInitialize(): InitializeResult {
   return result
 }
 
-function handleDocumentOpen(event: TextDocumentChangeEvent<TextDocument>) {
-  const document = event.document
-  const tree = parser.parse(document.getText())
-  documentTrees.set(document.uri, tree)
-}
-
 function handleSemanticTokens(params: any) {
   const document = documents.get(params.textDocument.uri)
   if (!document) return { data: [] }
 
-  const tree = documentTrees.get(params.textDocument.uri)
+  const data = buildSemanticTokens(document)
-  if (!tree) return { data: [] }
-
-  const data = buildSemanticTokens(document, tree)
   return { data }
 }
 
-function handleDocumentChange(change: TextDocumentChangeEvent<TextDocument>) {
+function handleDocumentChange(change: any) {
-  const document = change.document
+  const textDocument = change.document
-
+  const diagnostics = buildDiagnostics(textDocument)
-  // Parse and cache
+  connection.sendDiagnostics({ uri: textDocument.uri, diagnostics })
-  const tree = parser.parse(document.getText())
-  documentTrees.set(document.uri, tree)
-
-  // Build diagnostics using cached tree
-  const diagnostics = buildDiagnostics(document, tree)
-  connection.sendDiagnostics({ uri: document.uri, diagnostics })
 }
 
-function handleDocumentClose(event: TextDocumentChangeEvent<TextDocument>) {
+function handleCompletion(params: any) {
-  documentTrees.delete(event.document.uri)
-}
-
-function handleCompletion() {
   const keywords = ['if', 'else', 'do', 'end', 'and', 'or', 'true', 'false', 'null']
 
   return keywords.map((keyword) => ({
@@ -106,13 +84,8 @@ function handleParseTree(params: { uri: string }) {
   const document = documents.get(params.uri)
   if (!document) return 'Document not found'
 
-  const tree = documentTrees.get(params.uri)
-  if (!tree) {
-    connection.console.error(`🦐 No cached tree for ${params.uri}`)
-    return 'No cached parse tree available'
-  }
-
   const text = document.getText()
+  const tree = parser.parse(text)
   const cursor = tree.cursor()
 
   let formatted = ''