@ttsc/playground 0.19.2 → 0.20.0

package/src/npm/collectExternalPackageNames.ts

@@ -21,7 +21,7 @@ export function collectExternalPackageNames(
 /**
  * Collect the string specifiers of every executable module-loading construct in
  * `source`: `import`/`export ... from`, side-effect `import "x"`, dynamic
- * `import("x")`, and `require("x")` calls.
+ * `import("x")`, `require("x")`, and `require?.("x")` calls.
  *
  * The scan tokenizes `source` first so import/export/require lookalikes that
  * live inside comments, string or template contents, or regular-expression
@@ -38,7 +38,11 @@ function collectModuleSpecifiers(source: string): string[] {
   const isOpenParen = (token: Token | undefined): boolean =>
     token !== undefined && token.kind === "punct" && token.value === "(";
   const isMemberAccess = (token: Token | undefined): boolean =>
-    token !== undefined && token.kind === "punct" && token.value === ".";
+    token !== undefined &&
+    token.kind === "punct" &&
+    (token.value === "." || token.value === "?.");
+  const isOptionalChain = (token: Token | undefined): boolean =>
+    token !== undefined && token.kind === "punct" && token.value === "?.";
 
   for (let i = 0; i < tokens.length; i++) {
     const token = tokens[i];
@@ -47,8 +51,10 @@ function collectModuleSpecifiers(source: string): string[] {
     if (token.value === "require") {
       // `obj.require(...)` is an unrelated method call, not CommonJS require.
       if (isMemberAccess(tokens[i - 1])) continue;
-      if (isOpenParen(tokens[i + 1])) {
-        const spec = asString(tokens[i + 2]);
+      const optional =
+        isOptionalChain(tokens[i + 1]) && isOpenParen(tokens[i + 2]);
+      if (isOpenParen(tokens[i + 1]) || optional) {
+        const spec = asString(tokens[i + (optional ? 3 : 2)]);
         if (spec !== null) out.push(spec);
       }
       continue;
@@ -109,8 +115,16 @@ type Token =
   // A single- or double-quoted string literal, with escapes decoded to their
   // literal characters so a specifier survives unchanged.
   | { kind: "string"; value: string }
-  // A single punctuation character.
-  | { kind: "punct"; value: string }
+  // A punctuation token; compound forms are retained where lexical state or
+  // module-call recognition depends on them.
+  | {
+      kind: "punct";
+      value: string;
+      /** Whether a slash after this closing delimiter begins a regex literal. */
+      regexAllowedAfter?: boolean;
+      /** Whether this closes a function parameter list for an expression. */
+      functionBodyIsExpression?: boolean;
+    }
   // An opaque value token — number, template literal, or regular-expression
   // literal — whose contents can never be a static specifier.
   | { kind: "other" };
@@ -129,12 +143,189 @@ const REGEX_PRECEDING_KEYWORDS = new Set([
   "void",
   "do",
   "else",
+  "extends",
   "yield",
   "await",
   "case",
   "throw",
 ]);
 
+const CONTROL_HEADER_KEYWORDS = new Set([
+  "if",
+  "while",
+  "for",
+  "with",
+  "switch",
+  "catch",
+]);
+const BLOCK_PRECEDING_KEYWORDS = new Set(["else", "try", "finally", "do"]);
+
+interface IParenContext {
+  kind: "control" | "function" | "normal";
+  functionIsExpression?: boolean;
+}
+
+/**
+ * Retain just enough delimiter context to distinguish a regex statement after a
+ * control header or block from division after an expression value. The
+ * collector is not a parser: this state only decides whether a following slash
+ * is opaque regex text or executable code worth scanning for module calls.
+ */
+class LexicalContext {
+  public readonly tokens: Token[] = [];
+
+  private readonly parens: IParenContext[] = [];
+  private readonly braces: boolean[] = [];
+  private pendingFunctionExpression: boolean | undefined;
+  private readonly pendingClassExpressions: boolean[] = [];
+
+  public isRegexAllowed(): boolean {
+    return isRegexAllowedAfter(this.tokens[this.tokens.length - 1]);
+  }
+
+  public pushWord(value: string): void {
+    if (
+      value === "function" &&
+      !isMemberAccess(this.tokens[this.tokens.length - 1])
+    )
+      this.pendingFunctionExpression = isExpressionPosition(this.tokens);
+    else if (
+      value === "class" &&
+      !isMemberAccess(this.tokens[this.tokens.length - 1])
+    )
+      this.pendingClassExpressions.push(isExpressionPosition(this.tokens));
+    this.tokens.push({ kind: "word", value });
+  }
+
+  public pushOther(): void {
+    this.tokens.push({ kind: "other" });
+  }
+
+  public pushPunct(value: string): void {
+    if (value === ":") {
+      const previous = this.tokens[this.tokens.length - 1];
+      if (
+        previous?.kind === "word" &&
+        !isMemberAccess(this.tokens[this.tokens.length - 2])
+      ) {
+        if (previous.value === "function")
+          this.pendingFunctionExpression = undefined;
+        else if (previous.value === "class") this.pendingClassExpressions.pop();
+      }
+    }
+    if (value === "(") {
+      const functionIsExpression = this.pendingFunctionExpression;
+      this.pendingFunctionExpression = undefined;
+      this.parens.push(
+        functionIsExpression === undefined
+          ? {
+              kind: isControlHeader(this.tokens) ? "control" : "normal",
+            }
+          : { kind: "function", functionIsExpression },
+      );
+      this.tokens.push({ kind: "punct", value });
+      return;
+    }
+    if (value === ")") {
+      const context = this.parens.pop();
+      this.tokens.push({
+        kind: "punct",
+        value,
+        regexAllowedAfter: context?.kind === "control",
+        functionBodyIsExpression:
+          context?.kind === "function"
+            ? context.functionIsExpression
+            : undefined,
+      });
+      return;
+    }
+    if (value === "{") {
+      this.braces.push(this.braceAllowsRegexAfter());
+      this.tokens.push({ kind: "punct", value });
+      return;
+    }
+    if (value === "}") {
+      this.tokens.push({
+        kind: "punct",
+        value,
+        regexAllowedAfter: this.braces.pop() ?? false,
+      });
+      return;
+    }
+    this.tokens.push({ kind: "punct", value });
+  }
+
+  private braceAllowsRegexAfter(): boolean {
+    const previous = this.tokens[this.tokens.length - 1];
+    if (
+      previous?.kind === "punct" &&
+      previous.value === ")" &&
+      previous.functionBodyIsExpression !== undefined
+    )
+      return !previous.functionBodyIsExpression;
+    if (this.classBodyPrecedes(previous))
+      return !this.pendingClassExpressions.pop()!;
+    if (!previous) return true;
+    if (previous.kind === "word")
+      return BLOCK_PRECEDING_KEYWORDS.has(previous.value);
+    if (previous.kind !== "punct") return false;
+    if (previous.value === ")") return previous.regexAllowedAfter === true;
+    return (
+      previous.value === ";" || previous.value === "{" || previous.value === "}"
+    );
+  }
+
+  private classBodyPrecedes(previous: Token | undefined): boolean {
+    if (this.pendingClassExpressions.length === 0 || this.parens.length !== 0)
+      return false;
+    if (!previous) return false;
+    if (previous.kind === "word" || previous.kind === "other") return true;
+    return (
+      previous.kind === "punct" &&
+      (previous.value === ")" ||
+        previous.value === "]" ||
+        previous.value === "}")
+    );
+  }
+}
+
+function isControlHeader(tokens: readonly Token[]): boolean {
+  const previous = tokens[tokens.length - 1];
+  const beforePrevious = tokens[tokens.length - 2];
+  if (
+    previous?.kind === "word" &&
+    CONTROL_HEADER_KEYWORDS.has(previous.value) &&
+    !isMemberAccess(beforePrevious)
+  )
+    return true;
+  return (
+    previous?.kind === "word" &&
+    previous.value === "await" &&
+    beforePrevious?.kind === "word" &&
+    beforePrevious.value === "for"
+  );
+}
+
+function isMemberAccess(token: Token | undefined): boolean {
+  return (
+    token?.kind === "punct" && (token.value === "." || token.value === "?.")
+  );
+}
+
+function isExpressionPosition(tokens: readonly Token[]): boolean {
+  let index = tokens.length - 1;
+  // `async function` inherits the context before `async`: it is a declaration
+  // at a statement boundary and an expression after an assignment or return.
+  const latest = tokens[index];
+  if (latest?.kind === "word" && latest.value === "async") index--;
+  const previous = tokens[index];
+  if (!previous) return false;
+  if (previous.kind === "word")
+    return REGEX_PRECEDING_KEYWORDS.has(previous.value);
+  if (previous.kind !== "punct") return false;
+  return ![")", "]", "}", ";", "{", "++", "--"].includes(previous.value);
+}
+
 /**
  * Lexically tokenize `source` into the coarse token stream the specifier
  * collector needs. Comments are dropped; strings, templates, regex literals,
@@ -142,7 +333,8 @@ const REGEX_PRECEDING_KEYWORDS = new Set([
  * grammar match.
  */
 function tokenize(source: string): Token[] {
-  const tokens: Token[] = [];
+  const context = new LexicalContext();
+  const tokens = context.tokens;
   const n = source.length;
   const isIdStart = (c: string): boolean =>
     (c >= "a" && c <= "z") || (c >= "A" && c <= "Z") || c === "_" || c === "$";
@@ -152,14 +344,6 @@ function tokenize(source: string): Token[] {
 
   // A `/` opens a regex only in operator/statement position — never right after
   // a value (identifier, number, string, template, regex, `)` or `]`).
-  const regexAllowed = (): boolean => {
-    const prev = tokens[tokens.length - 1];
-    if (!prev) return true;
-    if (prev.kind === "string" || prev.kind === "other") return false;
-    if (prev.kind === "word") return REGEX_PRECEDING_KEYWORDS.has(prev.value);
-    return prev.value !== ")" && prev.value !== "]";
-  };
-
   let i = 0;
   while (i < n) {
     const c = source[i]!;
@@ -189,85 +373,52 @@ function tokenize(source: string): Token[] {
       continue;
     }
     // Regular-expression literal.
-    if (c === "/" && regexAllowed()) {
-      i++;
-      let inClass = false;
-      while (i < n) {
-        const d = source[i];
-        if (d === "\\") {
-          i += 2;
-          continue;
-        }
-        if (d === "\n") break;
-        if (d === "[") inClass = true;
-        else if (d === "]") inClass = false;
-        else if (d === "/" && !inClass) {
-          i++;
-          break;
-        }
-        i++;
-      }
-      while (i < n && isIdPart(source[i]!)) i++; // flags
-      tokens.push({ kind: "other" });
+    if (c === "/" && context.isRegexAllowed()) {
+      i = skipRegularExpression(source, i, isIdPart);
+      context.pushOther();
       continue;
     }
     // String literal.
     if (c === '"' || c === "'") {
-      i++;
-      let value = "";
-      while (i < n) {
-        const d = source[i]!;
-        if (d === "\\") {
-          value += source[i + 1] ?? "";
-          i += 2;
-          continue;
-        }
-        if (d === c) {
-          i++;
-          break;
-        }
-        if (d === "\n") break; // unterminated single-line string
-        value += d;
-        i++;
-      }
-      tokens.push({ kind: "string", value });
+      const quoted = readQuotedString(source, i, c);
+      i = quoted.end;
+      if (quoted.value === null) context.pushOther();
+      else tokens.push({ kind: "string", value: quoted.value });
       continue;
     }
-    // Template literal. Contents (including `${...}` expressions) are treated as
-    // opaque: a template specifier is not statically resolvable.
+    // Template quasis are opaque, but `${...}` substitutions are executable
+    // JavaScript and must receive the same lexical treatment as top-level code.
     if (c === "`") {
       i++;
-      let depth = 0;
       while (i < n) {
         const d = source[i];
         if (d === "\\") {
           i += 2;
           continue;
         }
-        if (depth === 0 && d === "`") {
+        if (d === "`") {
           i++;
           break;
         }
         if (d === "$" && source[i + 1] === "{") {
-          depth++;
-          i += 2;
-          continue;
-        }
-        if (depth > 0 && d === "}") {
-          depth--;
-          i++;
+          const start = i + 2;
+          const end = findTemplateSubstitutionEnd(source, start);
+          context.pushOther();
+          tokens.push(...tokenize(source.slice(start, end)));
+          context.pushOther();
+          i = end < n ? end + 1 : end;
           continue;
         }
         i++;
       }
-      tokens.push({ kind: "other" });
+      context.pushOther();
       continue;
     }
     // Identifier / keyword.
     if (isIdStart(c)) {
       let j = i + 1;
       while (j < n && isIdPart(source[j]!)) j++;
-      tokens.push({ kind: "word", value: source.slice(i, j) });
+      context.pushWord(source.slice(i, j));
       i = j;
       continue;
     }
@@ -275,13 +426,297 @@ function tokenize(source: string): Token[] {
     if (isDigit(c) || (c === "." && isDigit(source[i + 1] ?? ""))) {
       let j = i + 1;
       while (j < n && /[0-9a-fA-FxXoObBeE._]/.test(source[j]!)) j++;
-      tokens.push({ kind: "other" });
+      context.pushOther();
       i = j;
       continue;
     }
+    // Compound punctuation whose identity affects the following slash or
+    // direct optional-call recognition.
+    if ((c === "+" || c === "-") && source[i + 1] === c) {
+      context.pushPunct(c + c);
+      i += 2;
+      continue;
+    }
+    if (c === "?" && source[i + 1] === "." && !isDigit(source[i + 2] ?? "")) {
+      context.pushPunct("?.");
+      i += 2;
+      continue;
+    }
     // Single punctuation character.
-    tokens.push({ kind: "punct", value: c });
+    context.pushPunct(c);
     i++;
   }
   return tokens;
 }
+
+function findTemplateSubstitutionEnd(source: string, start: number): number {
+  let depth = 1;
+  const context = new LexicalContext();
+  const isIdentifierStart = (character: string): boolean =>
+    /[A-Za-z_$]/.test(character);
+  const isIdentifierPart = (character: string): boolean =>
+    /[A-Za-z0-9_$]/.test(character);
+  for (let i = start; i < source.length; i++) {
+    const c = source[i]!;
+    if (/\s/.test(c)) continue;
+    if (c === "\\") {
+      i++;
+      continue;
+    }
+    if (c === "'" || c === '"') {
+      i = skipQuoted(source, i, c);
+      context.tokens.push({ kind: "string", value: "" });
+      continue;
+    }
+    if (c === "`") {
+      i = skipTemplate(source, i);
+      context.pushOther();
+      continue;
+    }
+    if (c === "/" && source[i + 1] === "/") {
+      i += 2;
+      while (i < source.length && source[i] !== "\n") i++;
+      continue;
+    }
+    if (c === "/" && source[i + 1] === "*") {
+      i += 2;
+      while (i < source.length && !(source[i] === "*" && source[i + 1] === "/"))
+        i++;
+      i++;
+      continue;
+    }
+    if (c === "/" && context.isRegexAllowed()) {
+      i = skipRegularExpression(source, i, (character) =>
+        /[A-Za-z0-9_$]/.test(character),
+      );
+      i--;
+      context.pushOther();
+      continue;
+    }
+    if (isIdentifierStart(c)) {
+      let end = i + 1;
+      while (end < source.length && isIdentifierPart(source[end]!)) end++;
+      context.pushWord(source.slice(i, end));
+      i = end - 1;
+      continue;
+    }
+    if (c >= "0" && c <= "9") {
+      let end = i + 1;
+      while (end < source.length && /[0-9a-fA-FxXoObBeE._]/.test(source[end]!))
+        end++;
+      context.pushOther();
+      i = end - 1;
+      continue;
+    }
+    if ((c === "+" || c === "-") && source[i + 1] === c) {
+      context.pushPunct(c + c);
+      i++;
+      continue;
+    }
+    if (
+      c === "?" &&
+      source[i + 1] === "." &&
+      !/[0-9]/.test(source[i + 2] ?? "")
+    ) {
+      context.pushPunct("?.");
+      i++;
+      continue;
+    }
+    if (c === "{") depth++;
+    else if (c === "}" && --depth === 0) return i;
+    context.pushPunct(c);
+  }
+  return source.length;
+}
+
+function isRegexAllowedAfter(previous: Token | undefined): boolean {
+  if (!previous) return true;
+  if (previous.kind === "string" || previous.kind === "other") return false;
+  if (previous.kind === "word")
+    return REGEX_PRECEDING_KEYWORDS.has(previous.value);
+  if (previous.value === ")" || previous.value === "}")
+    return previous.regexAllowedAfter === true;
+  return (
+    previous.value !== "]" && previous.value !== "++" && previous.value !== "--"
+  );
+}
+
+interface IQuotedStringResult {
+  end: number;
+  value: string | null;
+}
+
+/** Read one quoted literal and compute its JavaScript StringValue. */
+function readQuotedString(
+  source: string,
+  start: number,
+  quote: string,
+): IQuotedStringResult {
+  let value = "";
+  for (let index = start + 1; index < source.length; ) {
+    const character = source[index]!;
+    if (character === quote) {
+      return { end: index + 1, value };
+    }
+    if (isLineTerminator(character)) {
+      return { end: index, value: null };
+    }
+    if (character !== "\\") {
+      value += character;
+      index++;
+      continue;
+    }
+
+    const escaped = source[index + 1];
+    if (escaped === undefined) {
+      return { end: source.length, value: null };
+    }
+    if (isLineTerminator(escaped)) {
+      index += escaped === "\r" && source[index + 2] === "\n" ? 3 : 2;
+      continue;
+    }
+
+    const simple = SIMPLE_STRING_ESCAPES[escaped];
+    if (simple !== undefined) {
+      value += simple;
+      index += 2;
+      continue;
+    }
+    if (escaped === "0") {
+      if (/[0-9]/.test(source[index + 2] ?? "")) {
+        return invalidQuotedString(source, start, quote);
+      }
+      value += "\0";
+      index += 2;
+      continue;
+    }
+    if (escaped >= "1" && escaped <= "9") {
+      return invalidQuotedString(source, start, quote);
+    }
+    if (escaped === "x") {
+      const digits = source.slice(index + 2, index + 4);
+      if (digits.length !== 2 || !/^[0-9a-fA-F]{2}$/.test(digits)) {
+        return invalidQuotedString(source, start, quote);
+      }
+      value += String.fromCharCode(Number.parseInt(digits, 16));
+      index += 4;
+      continue;
+    }
+    if (escaped === "u") {
+      if (source[index + 2] === "{") {
+        const close = source.indexOf("}", index + 3);
+        const digits = close < 0 ? "" : source.slice(index + 3, close);
+        const significantDigits = digits.replace(/^0+/, "") || "0";
+        const codePoint = Number.parseInt(digits, 16);
+        if (
+          close < 0 ||
+          !/^[0-9a-fA-F]+$/.test(digits) ||
+          significantDigits.length > 6 ||
+          codePoint > 0x10ffff
+        ) {
+          return invalidQuotedString(source, start, quote);
+        }
+        value += String.fromCodePoint(codePoint);
+        index = close + 1;
+        continue;
+      }
+      const digits = source.slice(index + 2, index + 6);
+      if (digits.length !== 4 || !/^[0-9a-fA-F]{4}$/.test(digits)) {
+        return invalidQuotedString(source, start, quote);
+      }
+      value += String.fromCharCode(Number.parseInt(digits, 16));
+      index += 6;
+      continue;
+    }
+
+    // IdentityEscape / NonEscapeCharacter: the slash is discarded and the
+    // escaped source character contributes directly to the StringValue.
+    value += escaped;
+    index += 2;
+  }
+  return { end: source.length, value: null };
+}
+
+const SIMPLE_STRING_ESCAPES: Readonly<Record<string, string>> = {
+  "'": "'",
+  '"': '"',
+  "\\": "\\",
+  b: "\b",
+  f: "\f",
+  n: "\n",
+  r: "\r",
+  t: "\t",
+  v: "\v",
+};
+
+function invalidQuotedString(
+  source: string,
+  start: number,
+  quote: string,
+): IQuotedStringResult {
+  const end = skipQuoted(source, start, quote);
+  return {
+    end: end < source.length && source[end] === quote ? end + 1 : end,
+    value: null,
+  };
+}
+
+function isLineTerminator(character: string): boolean {
+  return (
+    character === "\n" ||
+    character === "\r" ||
+    character === "\u2028" ||
+    character === "\u2029"
+  );
+}
+
+function skipRegularExpression(
+  source: string,
+  start: number,
+  isIdentifierPart: (character: string) => boolean,
+): number {
+  let index = start + 1;
+  let inClass = false;
+  while (index < source.length) {
+    const character = source[index];
+    if (character === "\\") {
+      index += 2;
+      continue;
+    }
+    if (character === "\n") break;
+    if (character === "[") inClass = true;
+    else if (character === "]") inClass = false;
+    else if (character === "/" && !inClass) {
+      index++;
+      break;
+    }
+    index++;
+  }
+  while (index < source.length && isIdentifierPart(source[index]!)) index++;
+  return index;
+}
+
+function skipQuoted(source: string, start: number, quote: string): number {
+  for (let i = start + 1; i < source.length; i++) {
+    if (source[i] === "\\") {
+      i++;
+      continue;
+    }
+    if (source[i] === quote || isLineTerminator(source[i]!)) return i;
+  }
+  return source.length;
+}
+
+function skipTemplate(source: string, start: number): number {
+  for (let i = start + 1; i < source.length; i++) {
+    if (source[i] === "\\") {
+      i++;
+      continue;
+    }
+    if (source[i] === "`") return i;
+    if (source[i] === "$" && source[i + 1] === "{") {
+      i = findTemplateSubstitutionEnd(source, i + 2);
+    }
+  }
+  return source.length;
+}