@triflare/fx 2.0.0 → 2.1.0-beta.1

package/src/formatter.js

@@ -0,0 +1,499 @@
+/**
+ * Triflare Fx source formatter.
+ *
+ * The formatter walks the parsed AST to decide structure and spacing, and
+ * replays the lexer's comment log to keep comments attached to the code around
+ * them. Formatting never compiles a project, so it cannot change generated
+ * project behavior, and running it twice produces no further changes.
+ *
+ * Canonical style:
+ * - four-space indentation, one statement per line;
+ * - `{` on the same line as its header, `} else {` for alternates;
+ * - single spaces around operators, after commas, and after colons;
+ * - double-quoted strings with minimal escapes;
+ * - unchanged declaration order and project data.
+ */
+import { tokenizeWithComments } from "./lexer.js";
+import { parseTokens } from "./parser.js";
+
+const INDENT = "    ";
+
+/** Binding strength of binary operators; larger binds tighter. */
+const PRECEDENCE = {
+  or: 1,
+  and: 2,
+  "==": 3,
+  "!=": 3,
+  "<": 3,
+  "<=": 3,
+  ">": 3,
+  ">=": 3,
+  "+": 4,
+  "-": 4,
+  "*": 5,
+  "/": 5,
+  "%": 5,
+};
+
+const UNARY_PRECEDENCE = 6;
+
+function escapeString(value) {
+  return String(value)
+    .replace(/\\/g, "\\\\")
+    .replace(/"/g, '\\"')
+    .replace(/\n/g, "\\n")
+    .replace(/\r/g, "\\r")
+    .replace(/\t/g, "\\t");
+}
+
+function stringLiteral(value) {
+  return `"${escapeString(value)}"`;
+}
+
+function isIdentifier(name) {
+  return /^[A-Za-z_$][A-Za-z0-9_$]*$/.test(name);
+}
+
+function mapKey(key) {
+  return isIdentifier(key) ? key : stringLiteral(key);
+}
+
+/**
+ * Render an expression, adding parentheses only where the AST needs them to
+ * survive a round trip through the parser.
+ *
+ * @param {object} node
+ * @param {number} [minimum] lowest precedence allowed without parentheses
+ * @param {string} [parentOperator] operator receiving this expression
+ * @returns {string}
+ */
+function renderExpression(node, minimum = 0, parentOperator) {
+  switch (node.type) {
+    case "Number":
+      return String(node.value);
+    case "String":
+      return stringLiteral(node.value);
+    case "Boolean":
+      return node.value ? "true" : "false";
+    case "Identifier":
+      return node.name;
+    case "Call":
+      return `${node.name}(${node.args
+        .map((arg) => renderExpression(arg))
+        .join(", ")})`;
+    case "ArrayLiteral":
+      return `[${node.items.map((item) => renderExpression(item)).join(", ")}]`;
+    case "Unary": {
+      const operand = renderExpression(node.operand, UNARY_PRECEDENCE, node.op);
+      if (node.op === "not") {
+        const text = `not ${operand}`;
+        return parentOperator && !["and", "or", "not"].includes(parentOperator)
+          ? `(${text})`
+          : text;
+      }
+      // `--x` is legal but reads like a decrement waiting to happen.
+      const nested = node.operand.type === "Unary" && node.operand.op === "-";
+      return `-${nested ? `(${operand})` : operand}`;
+    }
+    case "Binary": {
+      const precedence = PRECEDENCE[node.op] ?? 0;
+      const left = renderExpression(node.left, precedence, node.op);
+      // Right operands need parentheses at equal precedence to keep left
+      // associativity from regrouping them, e.g. `a - (b - c)`.
+      const right = renderExpression(node.right, precedence + 1, node.op);
+      const text = `${left} ${node.op} ${right}`;
+      return precedence < minimum ? `(${text})` : text;
+    }
+    default:
+      throw new Error(`Cannot format expression node ${node.type}`);
+  }
+}
+
+/** Render a hat event: a bare name when it takes no arguments. */
+function renderEvent(event) {
+  if (event.args.length === 0) return event.segments.join(".");
+  return `${event.segments.join(".")}(${event.args
+    .map((arg) => renderExpression(arg))
+    .join(", ")})`;
+}
+
+/** Group statements so the formatter can insert stable blank lines. */
+function category(node) {
+  switch (node.type) {
+    case "StageDecl":
+    case "SpriteDecl":
+      return "declaration";
+    case "PropertyDecl":
+      return "property";
+    case "VarDecl":
+    case "ListDecl":
+      return "data";
+    case "FunctionDecl":
+      return "function";
+    case "HatDecl":
+      return "script";
+    default:
+      return "statement";
+  }
+}
+
+/** Whether two adjacent statements belong in separate visual groups. */
+function needsBlankLine(previous, next) {
+  const before = category(previous);
+  const after = category(next);
+  if (before !== after) return true;
+  return (
+    before === "declaration" || before === "function" || before === "script"
+  );
+}
+
+class Formatter {
+  /**
+   * @param {string} source
+   * @param {string} file
+   */
+  constructor(source, file) {
+    const { tokens, comments } = tokenizeWithComments(source, file);
+    this.sourceLines = source.split(/\r?\n/);
+    this.tokens = tokens;
+    this.comments = comments;
+    this.commentIndex = 0;
+    this.tokenIndexes = new Map();
+    tokens.forEach((token, index) => this.tokenIndexes.set(token, index));
+    this.out = [];
+    this.program = parseTokens(tokens, source, file);
+  }
+
+  /** Token index for a loc that the parser took straight from the lexer. */
+  tokenIndexOf(token) {
+    return this.tokenIndexes.get(token) ?? this.tokens.length - 1;
+  }
+
+  /** First token with `value` at or after `from`, or -1. */
+  findToken(from, value) {
+    for (let index = Math.max(0, from); index < this.tokens.length; index += 1)
+      if (this.tokens[index].value === value) return index;
+    return -1;
+  }
+
+  /** Index of the `}` matching the `{` at `open`. */
+  matchBrace(open) {
+    let depth = 0;
+    for (let index = open; index < this.tokens.length; index += 1) {
+      const token = this.tokens[index];
+      if (token.type !== "punct") continue;
+      if (token.value === "{") depth += 1;
+      else if (token.value === "}") {
+        depth -= 1;
+        if (depth === 0) return index;
+      }
+    }
+    return this.tokens.length - 1;
+  }
+
+  /** Token span of a node's block body. */
+  bodySpan(node) {
+    const open = this.findToken(this.tokenIndexOf(node.loc), "{");
+    return { open, close: this.matchBrace(open) };
+  }
+
+  /** Whether a comment can share the previous output line. */
+  canInline(comment) {
+    if (/\r(?!$)|\n/.test(comment.text)) return false;
+    if (this.out.length === 0) return false;
+    if (this.out[this.out.length - 1].trim() === "") return false;
+    const previous = this.tokens[comment.after - 1];
+    return Boolean(previous && previous.line === comment.line);
+  }
+
+  /** Attach only the trailing comments that belong on the previous line. */
+  flushInline(limit) {
+    while (this.commentIndex < this.comments.length) {
+      const comment = this.comments[this.commentIndex];
+      if (comment.after > limit || !this.canInline(comment)) return;
+      this.commentIndex += 1;
+      this.out[this.out.length - 1] += ` ${comment.text.trim()}`;
+    }
+  }
+
+  /**
+   * Emit every comment that appears before token `limit`.
+   *
+   * @param {number} limit
+   * @param {number} indent
+   */
+  flush(limit, indent) {
+    const pad = INDENT.repeat(indent);
+    while (this.commentIndex < this.comments.length) {
+      const comment = this.comments[this.commentIndex];
+      if (comment.after > limit) break;
+      this.commentIndex += 1;
+      // The lexer includes the CR of a CRLF terminator in line comments.
+      const text = comment.text.startsWith("//")
+        ? comment.text.replace(/\r$/, "")
+        : comment.text;
+      const lines = text
+        .replace(/\r\n?/g, "\n")
+        .replace(/[ \t]+$/gm, "")
+        .split("\n");
+      if (lines.length === 1 && this.canInline(comment)) {
+        this.out[this.out.length - 1] += ` ${lines[0].trim()}`;
+        continue;
+      }
+      this.out.push(`${pad}${lines[0]}`);
+      for (let index = 1; index < lines.length; index += 1)
+        this.out.push(lines[index]);
+    }
+  }
+
+  pushBlank() {
+    if (this.out.length === 0) return;
+    if (this.out[this.out.length - 1] === "") return;
+    this.out.push("");
+  }
+
+  /** Whether the source has a blank line strictly between two lines. */
+  hasBlankBetween(fromLine, toLine) {
+    for (let line = fromLine + 1; line < toLine; line += 1)
+      if ((this.sourceLines[line - 1] ?? "").trim() === "") return true;
+    return false;
+  }
+
+  /**
+   * Render a statement list, flushing comments between its statements.
+   *
+   * Blank lines are canonical: declarations, functions and scripts are always
+   * separated, an intentional single blank line in the source survives, and
+   * runs of blank lines collapse to one.
+   */
+  renderBody(items, indent, closeIndex) {
+    items.forEach((item, index) => {
+      const start = this.tokenIndexOf(item.loc);
+      if (index > 0) {
+        this.flushInline(start);
+        const previousLine = this.tokens[start - 1]?.line ?? 0;
+        if (
+          needsBlankLine(items[index - 1], item) ||
+          this.hasBlankBetween(previousLine, item.loc.line)
+        )
+          this.pushBlank();
+      }
+      const pending = this.comments[this.commentIndex];
+      this.flush(start, indent);
+      // Keep a file header comment separated from the first declaration, or a
+      // leading comment inside a block separated from the code it introduces.
+      if (index === 0 && pending && pending.after <= start) {
+        const last = this.comments[this.commentIndex - 1];
+        const endLine = last.line + (last.text.match(/\n/g)?.length ?? 0);
+        if (this.hasBlankBetween(endLine, item.loc.line)) this.pushBlank();
+      }
+      this.renderItem(item, indent);
+    });
+    this.flush(closeIndex, indent);
+  }
+
+  /** Render the block body of a node that owns braces. */
+  renderNodeBody(node, indent) {
+    this.renderBody(node.body, indent + 1, this.bodySpan(node).close);
+  }
+
+  renderItem(node, indent) {
+    const pad = INDENT.repeat(indent);
+    switch (node.type) {
+      case "StageDecl":
+        this.out.push(`${pad}stage {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      case "SpriteDecl": {
+        const modifier = node.partial ? "partial " : "";
+        this.out.push(`${pad}${modifier}sprite ${stringLiteral(node.name)} {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      }
+      case "PropertyDecl":
+        this.renderProperty(node, indent);
+        return;
+      case "VarDecl":
+        this.out.push(
+          `${pad}${node.cloud ? "cloud " : ""}let ${node.name}` +
+            `${node.init ? ` = ${renderExpression(node.init)}` : ""};`,
+        );
+        return;
+      case "TempVarDecl":
+        this.out.push(
+          `${pad}${node.kind} let ${node.name}` +
+            `${node.init ? ` = ${renderExpression(node.init)}` : ""};`,
+        );
+        return;
+      case "ListDecl":
+        this.out.push(
+          `${pad}list ${node.name} = [${node.items
+            .map((item) => renderExpression(item))
+            .join(", ")}];`,
+        );
+        return;
+      case "FunctionDecl": {
+        const modifiers = `${node.returns ? "returning " : ""}${
+          node.warp ? "warp " : ""
+        }`;
+        const params = node.params
+          .map((param) =>
+            param.type ? `${param.name}: ${param.type}` : param.name,
+          )
+          .join(", ");
+        this.out.push(`${pad}${modifiers}function ${node.name}(${params}) {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      }
+      case "HatDecl":
+        this.out.push(`${pad}when ${renderEvent(node.event)} {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      case "If":
+        this.renderIf(node, indent);
+        return;
+      case "Repeat":
+        this.out.push(`${pad}repeat (${renderExpression(node.count)}) {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      case "While":
+        this.out.push(`${pad}while (${renderExpression(node.cond)}) {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      case "Forever":
+        this.out.push(`${pad}forever {`);
+        this.renderNodeBody(node, indent);
+        this.out.push(`${pad}}`);
+        return;
+      case "Return":
+        this.out.push(
+          `${pad}return${node.value ? ` ${renderExpression(node.value)}` : ""};`,
+        );
+        return;
+      case "Assign":
+        this.out.push(
+          `${pad}${node.name} ${node.op} ${renderExpression(node.value)};`,
+        );
+        return;
+      case "BlockCall":
+        this.renderBlockCall(node, indent);
+        return;
+      default:
+        throw new Error(`Cannot format statement node ${node.type}`);
+    }
+  }
+
+  renderProperty(node, indent) {
+    const pad = INDENT.repeat(indent);
+    if (!Array.isArray(node.value)) {
+      this.out.push(`${pad}${node.key}: ${renderExpression(node.value)};`);
+      return;
+    }
+    if (node.value.length === 0) {
+      this.out.push(`${pad}${node.key}: {};`);
+      return;
+    }
+    const span = this.bodySpan(node);
+    const inner = INDENT.repeat(indent + 1);
+    this.out.push(`${pad}${node.key}: {`);
+    node.value.forEach((entry, index) => {
+      this.flush(this.tokenIndexOf(entry.loc), indent + 1);
+      const comma = index === node.value.length - 1 ? "" : ",";
+      this.out.push(
+        `${inner}${mapKey(entry.key)}: ${stringLiteral(entry.value)}${comma}`,
+      );
+    });
+    this.flush(span.close, indent + 1);
+    this.out.push(`${pad}};`);
+  }
+
+  /** Find the `else` token that follows a block's closing brace. */
+  elseIndexAfter(close) {
+    const index = this.findToken(close + 1, "else");
+    return index === close + 1 ? index : -1;
+  }
+
+  renderIf(node, indent, prefix = "") {
+    const pad = INDENT.repeat(indent);
+    const span = this.bodySpan(node);
+    this.out.push(`${pad}${prefix}if (${renderExpression(node.cond)}) {`);
+    this.renderBody(node.then, indent + 1, span.close);
+    if (!node.else) {
+      this.out.push(`${pad}}`);
+      return;
+    }
+    const elseIndex = this.elseIndexAfter(span.close);
+    // Comments between `}` and `else` cannot sit on their own line without
+    // breaking the statement, so keep them inside the preceding block.
+    if (elseIndex !== -1) this.flush(elseIndex, indent + 1);
+    if (node.else.length === 1 && node.else[0].type === "If") {
+      this.renderIf(node.else[0], indent, "} else ");
+      return;
+    }
+    const elseOpen = elseIndex === -1 ? span.close + 1 : elseIndex + 1;
+    const elseClose = this.matchBrace(elseOpen);
+    this.out.push(`${pad}} else {`);
+    this.renderBody(node.else, indent + 1, elseClose);
+    this.out.push(`${pad}}`);
+  }
+
+  renderBlockCall(node, indent) {
+    const pad = INDENT.repeat(indent);
+    const args = node.args.map((arg) => renderExpression(arg)).join(", ");
+    // `control.forever { }` reads better than `control.forever() { }`, and the
+    // parser treats them identically.
+    const head =
+      node.body && node.args.length === 0
+        ? `${pad}${node.name}`
+        : `${pad}${node.name}(${args})`;
+    if (!node.body) {
+      this.out.push(`${head};`);
+      return;
+    }
+    const span = this.bodySpan(node);
+    this.out.push(`${head} {`);
+    this.renderBody(node.body, indent + 1, span.close);
+    if (!node.else) {
+      this.out.push(`${pad}}`);
+      return;
+    }
+    const elseIndex = this.elseIndexAfter(span.close);
+    if (elseIndex !== -1) this.flush(elseIndex, indent + 1);
+    const elseOpen = elseIndex === -1 ? span.close + 1 : elseIndex + 1;
+    const elseClose = this.matchBrace(elseOpen);
+    this.out.push(`${pad}} else {`);
+    this.renderBody(node.else, indent + 1, elseClose);
+    this.out.push(`${pad}}`);
+  }
+
+  format() {
+    const units = [];
+    if (this.program.stage) units.push(this.program.stage);
+    units.push(...this.program.sprites, ...this.program.topItems);
+    // Preserve the order declarations had in this file; only separate them.
+    units.sort((a, b) => this.tokenIndexOf(a.loc) - this.tokenIndexOf(b.loc));
+    this.renderBody(units, 0, this.tokens.length - 1);
+    if (this.out.length === 0) return "";
+    return `${this.out.join("\n")}\n`;
+  }
+}
+
+/**
+ * Format Triflare Fx source into the canonical style.
+ *
+ * Throws `LexError`/`ParseError` with source locations when the input cannot
+ * be parsed; callers must leave such files untouched.
+ *
+ * @param {string} source
+ * @param {string} [file]
+ * @returns {string}
+ */
+export function formatSource(source, file = "<input>") {
+  return new Formatter(source, file).format();
+}

package/src/index.js

@@ -8,6 +8,16 @@
  */
 export { tokenize } from "./lexer.js";
 export { parse } from "./parser.js";
+export { formatSource } from "./formatter.js";
+export { applyFixes, lintSources, sortDiagnostics } from "./linter.js";
+export {
+  DEFAULT_LINT,
+  effectivePreset,
+  normalizeLintConfig,
+  resolveRuleSeverities,
+} from "./lint/config.js";
+export { parseProjectConfigText } from "./lint/config-file.js";
+export { CATEGORIES, PRESETS, RULES } from "./lint/rules.js";
 export { Compiler, compile } from "./compiler.js";
 export { build, writeSb3 } from "./build.js";
 export { discoverSources, mergePrograms } from "./project.js";

package/src/lexer.js

@@ -52,13 +52,28 @@ const isDigit = (ch) => ch >= "0" && ch <= "9";
  */
 
 /**
- * Tokenize a Triflare Fx source string.
+ * @typedef {object} Comment
+ * @property {string} text the raw comment, including its delimiters
+ * @property {number} line
+ * @property {number} column
+ * @property {number} after index of the token that follows the comment
+ */
+
+/**
+ * Tokenize a Triflare Fx source string and record its comments.
+ *
+ * Comments stay out of the token stream, but the formatter needs them to put
+ * source comments back into a canonical layout. `after` is the index of the
+ * next token, so a comment can be re-attached between tokens even after the
+ * whitespace around it changes.
+ *
  * @param {string} source
  * @param {string} [file]
- * @returns {Token[]}
+ * @returns {{tokens: Token[], comments: Comment[]}}
  */
-export function tokenize(source, file = "<input>") {
+export function tokenizeWithComments(source, file = "<input>") {
   const tokens = [];
+  const comments = [];
   let i = 0;
   let line = 1;
   let column = 1;
@@ -99,10 +114,20 @@ export function tokenize(source, file = "<input>") {
 
     // --- comments -------------------------------------------------------
     if (ch === "/" && source[i + 1] === "/") {
+      const startIndex = i;
+      const startLine = line;
+      const startColumn = column;
       while (i < source.length && source[i] !== "\n") advance();
+      comments.push({
+        text: source.slice(startIndex, i),
+        line: startLine,
+        column: startColumn,
+        after: tokens.length,
+      });
       continue;
     }
     if (ch === "/" && source[i + 1] === "*") {
+      const startIndex = i;
       const startLine = line;
       const startColumn = column;
       advance(2);
@@ -111,6 +136,12 @@ export function tokenize(source, file = "<input>") {
       if (i >= source.length)
         fail("Unterminated block comment", startLine, startColumn);
       advance(2);
+      comments.push({
+        text: source.slice(startIndex, i),
+        line: startLine,
+        column: startColumn,
+        after: tokens.length,
+      });
       continue;
     }
 
@@ -198,5 +229,15 @@ export function tokenize(source, file = "<input>") {
   }
 
   push("eof", "", line, column);
-  return tokens;
+  return { tokens, comments };
+}
+
+/**
+ * Tokenize a Triflare Fx source string.
+ * @param {string} source
+ * @param {string} [file]
+ * @returns {Token[]}
+ */
+export function tokenize(source, file = "<input>") {
+  return tokenizeWithComments(source, file).tokens;
 }