@cplieger/deadset-ts 8.0.0-dev.6 → 8.0.0-dev.7

package/src/serialization-contract.ts

@@ -53,6 +53,7 @@ import {
   type SourceFile,
   type SpreadAssignment,
 } from "@typescript/native/ast";
+import { parenthesesRemoved } from "./unwrap.ts";
 import {
   SignatureKind,
   SymbolFlags,
@@ -227,12 +228,6 @@ interface MethodForm {
   readonly invocation: Invocation;
 }
 
-function unparenthesized(expression: Expression): Expression {
-  return isParenthesizedExpression(expression)
-    ? unparenthesized(expression.expression)
-    : expression;
-}
-
 /**
  * The invocation a call writes through a method of `Function.prototype`, read from
  * the syntax alone: `f.call(self, a)`, `f.apply(self, [a])`, `f.bind(self, a)`, which
@@ -241,12 +236,12 @@ function unparenthesized(expression: Expression): Expression {
  * carries. Whether the method is the prototype's is resolution's to say.
  */
 function throughMethod(call: CallExpression): MethodForm | undefined {
-  const expression = unparenthesized(call.expression);
+  const expression = parenthesesRemoved(call.expression);
   if (isPropertyAccessExpression(expression) && FUNCTION_METHODS.has(expression.name.text)) {
     const [, ...rest] = call.arguments;
     let passed: readonly Expression[] = rest;
     if (expression.name.text === "apply") {
-      const list = rest[0] === undefined ? undefined : unparenthesized(rest[0]);
+      const list = rest[0] === undefined ? undefined : parenthesesRemoved(rest[0]);
       passed = list === undefined ? [] : isArrayLiteralExpression(list) ? list.elements : [list];
     }
     return {
@@ -255,7 +250,7 @@ function throughMethod(call: CallExpression): MethodForm | undefined {
     };
   }
   if (isCallExpression(expression)) {
-    const bound = unparenthesized(expression.expression);
+    const bound = parenthesesRemoved(expression.expression);
     if (isPropertyAccessExpression(bound) && bound.name.text === "bind") {
       return {
         method: bound.name,

package/src/session.ts

@@ -26,6 +26,7 @@ import {
   type Queries,
   type UnansweredQuestion,
 } from "./query.ts";
+import { MANIFEST } from "./fs/manifest.ts";
 
 /**
  * The compiler the analysis reads through: a client over a protocol to a separate
@@ -115,7 +116,7 @@ export function openEngine(options: EngineOptions): Engine {
   if (components !== undefined && components.extensions.length > 0) {
     const made = components.host.temporaryDirectory();
     try {
-      components.host.writeDocument(joinPath(made.path, "package.json"), [
+      components.host.writeDocument(joinPath(made.path, MANIFEST), [
         mapperManifest(components.host.componentMapperCommand()),
       ]);
     } catch (error: unknown) {

package/src/suppress-file.ts

@@ -10,7 +10,7 @@ import type { Host } from "./host.ts";
 import type { InventorySymbol } from "./inventory.ts";
 import { jsonChunks } from "./json-chunks.ts";
 import { joinPath } from "./paths.ts";
-import { positionAt, walkDocument } from "./json-document.ts";
+import { isRecord, positionAt, walkDocument } from "./json-document.ts";
 import type { Position } from "./position.ts";
 import { declarationKey } from "./ref.ts";
 import {
@@ -26,6 +26,7 @@ import {
   type SuppressionRecord,
   type Suppressions,
 } from "./suppress.ts";
+import { message } from "./product.ts";
 
 /** The name and the location the grammar fixes for the ignore file. */
 export const IGNORE_FILE = "deadset-ignore.json";
@@ -79,10 +80,6 @@ function documentSite(shape: Shape): Position {
   return { path: shape.file, line: 0, column: 0 };
 }
 
-function isObject(value: unknown): value is Record<string, unknown> {
-  return typeof value === "object" && value !== null && !Array.isArray(value);
-}
-
 /** What one entry or row resolves to. */
 interface Read {
   readonly records: readonly SuppressionRecord[];
@@ -100,7 +97,7 @@ function readRecord(value: unknown, site: Position, byReference: ByReference, sh
   const refuse = (text: string, want: string): never => {
     throw new SuppressionError(shape.mechanism, siteText(site), text, want);
   };
-  if (!isObject(value)) {
+  if (!isRecord(value)) {
     return refuse(JSON.stringify(value), shape.recordWant);
   }
   for (const [key, held] of Object.entries(value)) {
@@ -200,9 +197,9 @@ function readDocument(
   try {
     parsed = JSON.parse(text);
   } catch (error: unknown) {
-    return malformed(`${shape.file}: ${error instanceof Error ? error.message : String(error)}`);
+    return malformed(`${shape.file}: ${message(error)}`);
   }
-  if (!isObject(parsed)) {
+  if (!isRecord(parsed)) {
     return malformed(shape.file);
   }
   const walk = walkDocument(text, shape.array);
@@ -328,9 +325,9 @@ export function otherSchemeRows(text: string, own: string = OWN_SCHEME): readonl
   try {
     parsed = JSON.parse(text);
   } catch (error: unknown) {
-    return refuse(error instanceof Error ? error.message : String(error));
+    return refuse(message(error));
   }
-  const held = isObject(parsed) ? parsed[BASELINE_SHAPE.array] : undefined;
+  const held = isRecord(parsed) ? parsed[BASELINE_SHAPE.array] : undefined;
   if (!Array.isArray(held)) {
     return refuse(BASELINE_SHAPE.documentWant);
   }
@@ -338,7 +335,7 @@ export function otherSchemeRows(text: string, own: string = OWN_SCHEME): readonl
     try {
       readRecord(value, documentSite(BASELINE_SHAPE), new Map(), BASELINE_SHAPE);
     } catch (error: unknown) {
-      return refuse(error instanceof Error ? error.message : String(error));
+      return refuse(message(error));
     }
     // readRecord has checked every member is a string and the code and symbol are present.
     const { code, symbol, path, reason } = value as Readonly<Record<string, string>>;

package/src/symbol-keys.ts

@@ -12,7 +12,6 @@ import {
   isIdentifier,
   isNoSubstitutionTemplateLiteral,
   isNumericLiteral,
-  isParenthesizedExpression,
   isPropertyAccessExpression,
   isSpreadElement,
   isStringLiteral,
@@ -21,6 +20,7 @@ import {
   type Node,
   type SourceFile,
 } from "@typescript/native/ast";
+import { parenthesesRemoved } from "./unwrap.ts";
 import type { Symbol as TSSymbol, Type } from "@typescript/native/sync";
 import type { AliasChains } from "./alias-chain.ts";
 import { nodeKey, type Inventory, type InventorySymbol } from "./inventory.ts";
@@ -52,10 +52,7 @@ interface KeyedUse {
 
 /** The identifier an element access's argument names its key by, where it is a name. */
 export function keyNameOf(argument: Node): Node | undefined {
-  let at = argument;
-  while (isParenthesizedExpression(at)) {
-    at = at.expression;
-  }
+  const at = parenthesesRemoved(argument);
   if (isIdentifier(at)) {
     return at;
   }

package/src/template-field.ts

@@ -20,6 +20,8 @@ import type { SymbolKind } from "./inventory.ts";
 import { memberNames } from "./member-names.ts";
 import { extensionOf, isAbsolutePath, joinPath, normalizePath, relativePath } from "./paths.ts";
 import { isComponentFile, type Position } from "./position.ts";
+import { bytewise } from "./text/compare.ts";
+import { filesBelow } from "./fs/tree.ts";
 
 /** One template, by its path below the target root. */
 interface TemplateFile {
@@ -50,14 +52,6 @@ const READ_BY_MARKUP: ReadonlySet<SymbolKind> = new Set([...READ_BY_A_TEMPLATE,
 /** One identifier, as the grammar of the analyzed language spells one. */
 const IDENTIFIER = /[\p{ID_Start}$_][\p{ID_Continue}$\u200C\u200D]*/uy;
 
-/** Two strings ordered bytewise. */
-function compare(a: string, b: string): number {
-  if (a === b) {
-    return 0;
-  }
-  return a < b ? -1 : 1;
-}
-
 /**
  * Every template under the configured directories, read once for the run, in path
  * order and each file once. A directory is named relative to the target root, and
@@ -69,15 +63,8 @@ function compare(a: string, b: string): number {
 export function readTemplates(host: Host, targetRoot: string, analysis: Analysis): Templates {
   const files = new Map<string, string>();
   const read = (dir: string): void => {
-    const entries = [...host.readDirectory(joinPath(targetRoot, dir))].sort((a, b) =>
-      compare(a.name, b.name),
-    );
-    for (const entry of entries) {
-      const path = joinPath(dir, entry.name);
-      if (entry.directory) {
-        read(path);
-        continue;
-      }
+    for (const below of filesBelow(host, joinPath(targetRoot, dir), () => true)) {
+      const path = joinPath(dir, below);
       const absolute = joinPath(targetRoot, path);
       if (!files.has(path) && host.kindOf(absolute) === "file") {
         files.set(path, host.readFile(absolute));
@@ -104,7 +91,7 @@ export function readTemplates(host: Host, targetRoot: string, analysis: Analysis
     delimiters: analysis.templateDelimiters,
     files: [...files]
       .map(([path, text]) => ({ path, text }))
-      .sort((a, b) => compare(a.path, b.path)),
+      .sort((a, b) => bytewise(a.path, b.path)),
   };
 }
 

package/src/test-globals.ts

@@ -25,10 +25,12 @@ import {
 } from "@typescript/native/ast";
 import type { ParseFile } from "./configuration-files.ts";
 import type { Host } from "./host.ts";
-import { dirnamePath, joinPath } from "./paths.ts";
+import { ancestors, dirnamePath, joinPath } from "./paths.ts";
 import { UNANSWERED } from "./query.ts";
 import type { Engine, ProjectView } from "./session.ts";
 import { writtenPath, type WorkspaceProgram } from "./workspace-programs.ts";
+import { filesIn } from "./fs/tree.ts";
+import { holdsManifest, INSTALLED_PACKAGES } from "./fs/manifest.ts";
 
 /** The test runner whose configuration injects the globals, and the `types` entry declaring them. */
 export const GLOBALS_RUNNER = "vitest";
@@ -120,16 +122,7 @@ function setsGlobals(file: SourceFile): boolean {
 
 /** Whether the runner's configuration in one manifest's directory injects the globals. */
 function injectsGlobals(host: Host, parse: ParseFile, dir: string): boolean {
-  let names: readonly string[];
-  try {
-    names = host
-      .readDirectory(dir)
-      .filter((entry) => !entry.directory && RUNNER_CONFIG.test(entry.name))
-      .map((entry) => entry.name);
-  } catch {
-    return false;
-  }
-  return names.some((name) => {
+  return filesIn(host, dir, (name) => RUNNER_CONFIG.test(name)).some((name) => {
     const path = joinPath(dir, name);
     try {
       return setsGlobals(parse(path, host.readFile(path)));
@@ -141,26 +134,15 @@ function injectsGlobals(host: Host, parse: ParseFile, dir: string): boolean {
 
 /** Whether a file declaring the globals is installed where the configuration resolves packages from. */
 function globalsInstalled(host: Host, configFile: string): boolean {
-  for (let dir = dirnamePath(configFile); ; dir = dirnamePath(dir)) {
-    if (host.kindOf(joinPath(dir, "node_modules", GLOBALS_RUNNER, "globals.d.ts")) === "file") {
-      return true;
-    }
-    if (dirnamePath(dir) === dir) {
-      return false;
-    }
-  }
+  return ancestors(dirnamePath(configFile)).some(
+    (dir) =>
+      host.kindOf(joinPath(dir, INSTALLED_PACKAGES, GLOBALS_RUNNER, "globals.d.ts")) === "file",
+  );
 }
 
 /** The directory of the manifest nearest one configuration file, at or above its own. */
 function manifestDir(host: Host, configFile: string): string | undefined {
-  for (let dir = dirnamePath(configFile); ; dir = dirnamePath(dir)) {
-    if (host.kindOf(joinPath(dir, "package.json")) === "file") {
-      return dir;
-    }
-    if (dirnamePath(dir) === dir) {
-      return undefined;
-    }
-  }
+  return ancestors(dirnamePath(configFile)).find((dir) => holdsManifest(host, dir));
 }
 
 function typesOf(value: unknown): readonly string[] | undefined {

package/src/test-joins.ts

@@ -8,9 +8,11 @@
 
 import { sourcePaths } from "./file-facts.ts";
 import type { Host } from "./host.ts";
-import { dirnamePath, joinPath, relativePath } from "./paths.ts";
+import { ancestors, dirnamePath, joinPath } from "./paths.ts";
 import type { Engine } from "./session.ts";
 import { writtenPath, type WorkspaceProgram } from "./workspace-programs.ts";
+import { bytewise } from "./text/compare.ts";
+import { holdsManifest } from "./fs/manifest.ts";
 
 const PREFERRED = "/tsconfig.json";
 
@@ -18,7 +20,7 @@ const PREFERRED = "/tsconfig.json";
 function byPreference(a: string, b: string): number {
   const first = a.endsWith(PREFERRED) ? 0 : 1;
   const second = b.endsWith(PREFERRED) ? 0 : 1;
-  return first - second || (a < b ? -1 : a > b ? 1 : 0);
+  return first - second || bytewise(a, b);
 }
 
 /**
@@ -54,17 +56,13 @@ export function joinedTestFiles(
     if (listed.has(file) || !patterns.some((pattern) => pattern.test(path))) {
       continue;
     }
-    for (let dir = dirnamePath(file); ; dir = dirnamePath(dir)) {
+    for (const dir of ancestors(dirnamePath(file), root)) {
       const [nearest] = byDir.get(dir) ?? [];
       if (nearest !== undefined) {
         joins.set(nearest, [...(joins.get(nearest) ?? []), file]);
         break;
       }
-      if (
-        host.kindOf(joinPath(dir, "package.json")) === "file" ||
-        dir === root ||
-        relativePath(root, dir) === undefined
-      ) {
+      if (holdsManifest(host, dir)) {
         break;
       }
     }

package/src/text/compare.ts

@@ -0,0 +1,27 @@
+/**
+ * Two strings in the order of their UTF-8 bytes, the order every set of a run is read in.
+ * That is code point order, which UTF-16 code unit order (`<` on strings) breaks above
+ * U+FFFF: a surrogate unit is below U+E000 while the code point it encodes is above U+FFFF.
+ */
+export function bytewise(a: string, b: string): number {
+  if (a === b) {
+    return 0;
+  }
+  const length = Math.min(a.length, b.length);
+  for (let at = 0; at < length; at += 1) {
+    const x = a.charCodeAt(at);
+    const y = b.charCodeAt(at);
+    if (x !== y) {
+      return ranked(x) < ranked(y) ? -1 : 1;
+    }
+  }
+  return a.length < b.length ? -1 : 1;
+}
+
+/** One UTF-16 unit moved so that a surrogate ranks above every unit of U+E000 to U+FFFF. */
+function ranked(unit: number): number {
+  if (unit >= 0xe000) {
+    return unit - 0x800;
+  }
+  return unit >= 0xd800 ? unit + 0x2000 : unit;
+}

package/src/text/glob.ts

@@ -0,0 +1,210 @@
+/**
+ * Glob patterns over paths in two forms. The configured form (the configuration, convention
+ * rows, bundler entries, workspace member patterns) has `*` and `?` within a segment, `**`
+ * across segments and brace lists. A `.gitignore` line has git's form: bracket expressions
+ * and backslash escapes, no brace lists (https://git-scm.com/docs/gitignore#_pattern_format).
+ * A configured root is not a glob: it names a symbol reference, and is read where it is used.
+ */
+
+type GlobForm = "configured" | "gitignore";
+
+/**
+ * One character of a pattern as the expression syntax reads it literally. The set escaped
+ * is closed: under the unicode flag an escape of any other character is refused outright.
+ */
+function quoted(character: string): string {
+  return /[.*+?^${}()|[\]\\/]/u.test(character) ? `\\${character}` : character;
+}
+
+/**
+ * The ASCII ranges each POSIX class a bracket expression may name holds, as git's own
+ * character table reads them, each range written as its first and last character.
+ */
+const CLASSES: ReadonlyMap<string, string> = new Map([
+  ["alnum", "09AZaz"],
+  ["alpha", "AZaz"],
+  ["blank", "\t\t  "],
+  ["cntrl", "\0\x1f\x7f\x7f"],
+  ["digit", "09"],
+  ["graph", "!~"],
+  ["lower", "az"],
+  ["print", " ~"],
+  ["punct", "!/:@[`{~"],
+  ["space", "\t\n\r\r  "],
+  ["upper", "AZ"],
+  ["xdigit", "09AFaf"],
+]);
+
+/** What a pattern git cannot read matches: nothing, as git aborts the whole match. */
+const NOTHING = "(?!)";
+
+/** One code point as a member of an expression class. */
+function member(point: number): string {
+  return `\\u{${point.toString(16)}}`;
+}
+
+/**
+ * The class a bracket expression opening at `open` denotes, and the index past its `]`,
+ * read as git's wildmatch reads it: a leading `!` or `^` negates it, a `]` first is a
+ * member, `\` escapes one character, `x-y` is a range (one whose `y` is below `x` holds
+ * `x` alone) and `[:name:]` a POSIX class. An unclosed bracket or an unknown class name
+ * matches nothing. Neither form matches the solidus.
+ */
+function bracket(pattern: string, open: number): { source: string; end: number } {
+  let at = open + 1;
+  const negated = pattern[at] === "!" || pattern[at] === "^";
+  if (negated) {
+    at += 1;
+  }
+  const next = (): number | undefined => {
+    const point = pattern.codePointAt(at);
+    at += point !== undefined && point > 0xffff ? 2 : 1;
+    return point;
+  };
+  let members = "";
+  let low: number | undefined;
+  for (let first = true; ; first = false) {
+    let point = next();
+    if (point === undefined) {
+      return { source: NOTHING, end: pattern.length };
+    }
+    if (point === 0x5d && !first) {
+      return { source: negated ? `[^/${members}]` : `(?!/)[${members}]`, end: at };
+    }
+    if (point === 0x2d && low !== undefined && at < pattern.length && pattern[at] !== "]") {
+      let high = next();
+      if (high === 0x5c) {
+        high = next();
+      }
+      if (high === undefined) {
+        return { source: NOTHING, end: pattern.length };
+      }
+      members += high >= low ? `${member(low)}-${member(high)}` : "";
+      low = undefined;
+      continue;
+    }
+    if (point === 0x5b && pattern[at] === ":") {
+      const close = pattern.indexOf("]", at + 1);
+      if (close < 0) {
+        return { source: NOTHING, end: pattern.length };
+      }
+      if (close > at + 1 && pattern[close - 1] === ":") {
+        const ranges = CLASSES.get(pattern.slice(at + 1, close - 1));
+        if (ranges === undefined) {
+          return { source: NOTHING, end: pattern.length };
+        }
+        for (let one = 0; one < ranges.length; one += 2) {
+          members += `${member(ranges.charCodeAt(one))}-${member(ranges.charCodeAt(one + 1))}`;
+        }
+        low = undefined;
+        at = close + 1;
+        continue;
+      }
+    }
+    if (point === 0x5c) {
+      point = next();
+      if (point === undefined) {
+        return { source: NOTHING, end: pattern.length };
+      }
+    }
+    members += member(point);
+    low = point;
+  }
+}
+
+/** The expression source of one pattern with no brace list. */
+function source(pattern: string, form: GlobForm): string {
+  let text = "";
+  let at = 0;
+  while (at < pattern.length) {
+    if (pattern.startsWith("**/", at)) {
+      text += "(?:[^/]*/)*";
+      at += 3;
+      continue;
+    }
+    if (pattern.startsWith("**", at)) {
+      text += ".*";
+      at += 2;
+      continue;
+    }
+    const character = pattern[at] ?? "";
+    if (form === "gitignore" && character === "[") {
+      const read = bracket(pattern, at);
+      text += read.source;
+      at = read.end;
+      continue;
+    }
+    if (form === "gitignore" && character === "\\" && at + 1 < pattern.length) {
+      text += quoted(pattern[at + 1] ?? "");
+      at += 2;
+      continue;
+    }
+    at += 1;
+    text += character === "*" ? "[^/]*" : character === "?" ? "[^/]" : quoted(character);
+  }
+  return text;
+}
+
+/** A pattern with every brace list expanded into its alternatives; a lone `{` stands for itself. */
+export function braceExpanded(pattern: string): readonly string[] {
+  const open = pattern.indexOf("{");
+  const close = open < 0 ? -1 : pattern.indexOf("}", open);
+  if (close < 0) {
+    return [pattern];
+  }
+  const before = pattern.slice(0, open);
+  const after = pattern.slice(close + 1);
+  return pattern
+    .slice(open + 1, close)
+    .split(",")
+    .flatMap((member) => braceExpanded(before + member + after));
+}
+
+/** The expression one whole-path pattern of the configured form denotes. */
+export function globExpression(pattern: string): RegExp {
+  const alternatives = braceExpanded(pattern).map((one) => source(one, "configured"));
+  return new RegExp(`^(?:${alternatives.join("|")})$`, "u");
+}
+
+/** The expression one `.gitignore` pattern denotes over a path below its file's directory. */
+export function gitignoreExpression(pattern: string): RegExp {
+  return new RegExp(`^${source(pattern, "gitignore")}$`, "u");
+}
+
+const SEGMENTS = new Map<string, RegExp>();
+
+/**
+ * Whether one directory name matches one segment of a workspace's member pattern, its brace
+ * lists already expanded. Neither `*` nor `?` matches a leading dot, as npm's workspaces read
+ * them: https://docs.npmjs.com/cli/using-npm/workspaces
+ */
+export function segmentMatches(segment: string, name: string): boolean {
+  if (name.startsWith(".") && !segment.startsWith(".")) {
+    return false;
+  }
+  let expression = SEGMENTS.get(segment);
+  if (expression === undefined) {
+    expression = globExpression(segment);
+    SEGMENTS.set(segment, expression);
+  }
+  return expression.test(name);
+}
+
+/**
+ * Whether one path's segments match a member pattern's, `**` standing for any run of segments
+ * that starts with no dot.
+ */
+export function pathMatches(pattern: readonly string[], path: readonly string[]): boolean {
+  const [head, ...rest] = pattern;
+  if (head === undefined) {
+    return path.length === 0;
+  }
+  if (head === "**") {
+    return (
+      pathMatches(rest, path) ||
+      (path.length > 0 && !(path[0] ?? "").startsWith(".") && pathMatches(pattern, path.slice(1)))
+    );
+  }
+  const [name, ...below] = path;
+  return name !== undefined && segmentMatches(head, name) && pathMatches(rest, below);
+}