@hew-lang/sandbox-vm 0.6.1 → 0.6.3

package/README.md

@@ -5,8 +5,9 @@
 borrows, cleanup, checked failures, actor protocols and suspension edges.
 
 Actors, tasks and pipes use resumable frames on a deterministic scheduler.
-Timers advance virtual time. Native filesystem and network operations are
-refused before execution; other missing operations are reported separately as
+Timers advance virtual time. Hosts can supply real origin-private filesystem
+(OPFS) directory handles; network and ambient filesystem operations are refused
+before execution. Other missing operations are reported separately as
 `not_implemented`. The loader is the capability admission authority.
 
 ## Browser API
@@ -17,16 +18,43 @@ import { runBytecode } from "@hew-lang/sandbox-vm";
 
 const { diagnostics, bytecode } = JSON.parse(compileToSandboxBytecode(source));
 if (bytecode) {
-  const trace = runBytecode(bytecode, { replay: { step_budget: 100000 } });
+  const trace = await runBytecode(bytecode, {
+    replay: { step_budget: 100000 },
+  });
 }
 ```
 
 Initialize the Wasm compiler using its generated default export before calling
 it. `runProgram(source, stdin)` is also available for existing page bridges;
-install `globalThis.compileToSandboxBytecode` first. Its result includes
+install `globalThis.compileToSandboxBytecode` first. Both entrypoints return a Promise. The page result includes
 `stdout`, `exit_code`, `diagnostics`, `status`, `sandbox_rejections`,
 `compiler_version` and `hew_version`.
 
+To grant filesystem authority, pass only the directories the program needs:
+
+```js
+const trace = await runBytecode(bytecode, {
+  directoryRoots: { read: inputDirectory, write: outputDirectory },
+  signal: controller.signal,
+});
+```
+
+Both roots are optional and default to no authority. The handles must belong to
+this origin's OPFS; external picker-backed handles are refused because their
+symlink confinement is not standardized. Existing Hew `fs.open_read_dir(".")`
+and `fs.open_write_dir(".")` acquire the supplied roots. Relative traversal stays
+beneath the current directory owner, including after narrowing with `subdir`.
+Children remain usable after closing a parent. Each `file.write` publishes its
+bytes before it returns, even while the file owner remains open.
+
+`AbortSignal` requests cancellation through the source's authored cleanup edges.
+Filesystem acquisition publishes its owner before observing cancellation, as on
+native hosts. An in-progress file operation drains or aborts its browser writer
+before cancellation releases its inputs. Pure execution retains the deterministic
+virtual scheduler and runs until a host wait or completion. Pass the same options
+as the fourth argument to `runProgram(source, stdin, files, options)`; signals are
+for direct calls rather than worker message payloads.
+
 A `sandbox_rejected` result includes rejection entries with `category`, `code`,
 `capability`, `message` and `span`. Categories are `native_only`,
 `not_implemented` and `invalid_package`. Compiler errors, language panics,
@@ -48,8 +76,24 @@ The [package contract](bytecode/package-v1.md) describes the SIR projection.
 Source fixtures are compiled by the public compiler and executed to record
 deterministic traces. Tests also compare executable sources with native Hew.
 
+With `hew-wasm/pkg` and `hew-sandbox-vm/dist` built, serve the repository on
+localhost and open `hew-sandbox-vm/test/browser/directory-handles.html` to run the
+real browser filesystem controls. They compile the ordinary Hew source fixture,
+use a temporary OPFS directory, verify disk contents, confinement and cancellation,
+then remove their entries. The page accepts a `compiler` module URL query parameter
+to reuse another built compiler package.
+
+`RunOptions.argv` supplies process arguments, including the executable name at
+index zero. It defaults to `["main.hew"]` and is copied for each run. For example,
+`runBytecode(bytecode, { argv: ["greet", "--", "Zoë", ""] })` supplies those exact
+strings to `std.os.args()`. The VM never reads Node or browser process arguments.
+
 Typed serialization executes the shared Rust codecs in `@hew-lang/wasm`.
 Pass its initialized `executeCodec` export as `runBytecode`'s `codecBridge`
 option. The Node test host loads that package automatically; in a source
 checkout, run `make wasm` to build `hew-wasm/pkg` first. An explicit
 `HEW_WASM_MODULE` module URL can select another build of the same package.
+
+UTC calendar operations also require the initialized `executeCalendar`
+export as `calendarBridge`. Both hosts use the shared Chrono calendar engine;
+the bridge preserves integer seconds and nanoseconds exactly.

package/cli/test-runner.mjs

@@ -22,6 +22,7 @@ try {
   const args = parseArgs(process.argv.slice(2));
   const packageValue = JSON.parse(fs.readFileSync(args.path, "utf8"));
   let codecBridge;
+  let calendarBridge;
   if (packageValue.codecs?.length) {
     const local = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../../hew-wasm/pkg/hew_wasm.js");
     const location = process.env.HEW_WASM_MODULE ?? (fs.existsSync(local) ? pathToFileURL(local).href : "@hew-lang/wasm");
@@ -31,11 +32,13 @@ try {
       await wasm.default({ module_or_path: fs.readFileSync(new URL(resolved.replace(/\.js$/, "_bg.wasm"))) });
     }
     codecBridge = wasm.executeCodec ?? wasm.default?.executeCodec;
+    calendarBridge = wasm.executeCalendar ?? wasm.default?.executeCalendar;
     if (typeof codecBridge !== "function") throw new Error("hew-wasm executeCodec export is missing; rebuild the Wasm package");
   }
   const propertyPath = reportPath ? path.join(path.dirname(reportPath), "property.cbor") : null;
-  const trace = runBytecode(packageValue, {
+  const trace = await runBytecode(packageValue, {
     codecBridge,
+    calendarBridge,
     propertyInput: propertyPath && fs.existsSync(propertyPath) ? [...fs.readFileSync(propertyPath)] : [],
     fixtureId: "hew-test",
     traceId: "trace:hew-test",

package/dist/interpreter/interpreter.d.ts

@@ -1,2 +1,2 @@
 import type { RunOptions, SandboxTrace } from "./types.js";
-export declare function runBytecode(input: unknown, options?: RunOptions): SandboxTrace;
+export declare function runBytecode(input: unknown, options?: RunOptions): Promise<SandboxTrace>;

package/dist/interpreter/interpreter.js

@@ -1,7 +1,7 @@
 import { validateBytecodePackage } from "./schema-validator.js";
 import { runPackageV1 } from "./v1/exec.js";
 import { TraceBuilder } from "./trace.js";
-export function runBytecode(input, options = {}) {
+export async function runBytecode(input, options = {}) {
     let bytecode;
     try {
         bytecode = validateBytecodePackage(input);

package/dist/interpreter/parity-runner.js

@@ -45,7 +45,7 @@ function parseArgs(argv) {
 try {
     const args = parseArgs(process.argv.slice(2));
     const bytecode = JSON.parse(fs.readFileSync(args.bytecodePath, "utf8"));
-    const trace = runBytecode(bytecode, {
+    const trace = await runBytecode(bytecode, {
         replay: {
             seed: args.seed,
             step_budget: DEFAULT_STEP_BUDGET,

package/dist/interpreter/run-program.d.ts

@@ -1,5 +1,5 @@
 import type { PackageV1 } from "./v1/package.js";
-import type { JsonValue, RuntimeStatus, SandboxRejection } from "./types.js";
+import type { RunOptions, JsonValue, RuntimeStatus, SandboxRejection } from "./types.js";
 export interface Diagnostic {
     severity: string;
     phase: string;
@@ -24,7 +24,8 @@ type SandboxCompiler = (source: string, profile?: string, files?: Map<string, st
 declare global {
     var __hewSandboxCompileToSandboxBytecode: SandboxCompiler | undefined;
     var __hewSandboxExecuteCodec: ((request: string) => string) | undefined;
+    var __hewSandboxExecuteCalendar: ((request: string) => string) | undefined;
     var compileToSandboxBytecode: SandboxCompiler | undefined;
 }
-export declare function runProgram(source: string, stdin: string, files?: Map<string, string | Uint8Array>): RunProgramResult;
+export declare function runProgram(source: string, stdin: string, files?: Map<string, string | Uint8Array>, options?: RunOptions): Promise<RunProgramResult>;
 export {};

package/dist/interpreter/run-program.js

@@ -1,6 +1,6 @@
 import { runBytecode } from "./interpreter.js";
 const SANDBOX_PROFILE = "sandbox-vm-export";
-export function runProgram(source, stdin, files) {
+export async function runProgram(source, stdin, files, options = {}) {
     const compileOutput = compileSource(source, files);
     if (hasErrorDiagnostics(compileOutput.diagnostics) ||
         compileOutput.bytecode === null) {
@@ -15,9 +15,11 @@ export function runProgram(source, stdin, files) {
         };
     }
     const bytecode = compileOutput.bytecode;
-    const trace = runBytecode(bytecode, {
+    const trace = await runBytecode(bytecode, {
+        ...options,
         fixtureId: "page-run",
-        codecBridge: globalThis.__hewSandboxExecuteCodec,
+        codecBridge: options.codecBridge ?? globalThis.__hewSandboxExecuteCodec,
+        calendarBridge: options.calendarBridge ?? globalThis.__hewSandboxExecuteCalendar,
         traceId: "trace:page-run",
         replay: { inputs: [{ kind: "stdin", data: stdin }] },
     });

package/dist/interpreter/schema-validator.d.ts

@@ -1,8 +1,13 @@
-import { type PackageV1 } from "./v1/package.js";
+import type { PackageV1 } from "./v1/package.js";
+declare const checkedPackage: unique symbol;
+export type CheckedPackage = PackageV1 & {
+    readonly [checkedPackage]: true;
+};
 export declare class BytecodeValidationError extends Error {
     readonly errors: string[];
     constructor(errors: string[]);
 }
-/** Check the package envelope before capability admission traverses its tables.
- * Operation support belongs to v1/validate, with one shared executor registry. */
-export declare function validateBytecodePackage(input: unknown): PackageV1;
+/** Decode an isolated package before capability admission or any host effect.
+ * Cloning prevents a caller or host callback from changing the checked input. */
+export declare function validateBytecodePackage(input: unknown): CheckedPackage;
+export {};

package/dist/interpreter/schema-validator.js

@@ -1,4 +1,5 @@
-import { isPackageV1 } from "./v1/package.js";
+import { validPackage } from "./v1/package-validator.generated.js";
+import { validatePackageGraph } from "./v1/package-graph.js";
 export class BytecodeValidationError extends Error {
     errors;
     constructor(errors) {
@@ -7,55 +8,20 @@ export class BytecodeValidationError extends Error {
         this.name = "BytecodeValidationError";
     }
 }
-/** Check the package envelope before capability admission traverses its tables.
- * Operation support belongs to v1/validate, with one shared executor registry. */
+/** Decode an isolated package before capability admission or any host effect.
+ * Cloning prevents a caller or host callback from changing the checked input. */
 export function validateBytecodePackage(input) {
-    if (!isPackageV1(input))
-        throw new BytecodeValidationError(["expected hew.sandbox.bytecode.v1"]);
-    const errors = [];
-    for (const key of ["profile", "hew_version", "compiler_version"]) {
-        if (typeof input[key] !== "string")
-            errors.push(`${key} must be a string`);
-    }
-    for (const key of [
-        "functions",
-        "strings",
-        "bytes",
-        "regex_patterns",
-        "aggregates",
-        "variants",
-        "runtime_families",
-        "externs",
-        "suspend_kinds",
-        "value_capabilities",
-        "closures",
-        "vtables",
-    ]) {
-        if (!Array.isArray(input[key]))
-            errors.push(`${key} must be an array`);
-    }
-    if (errors.length)
-        throw new BytecodeValidationError(errors);
-    for (const fn of input.functions) {
-        if (!fn ||
-            !Number.isInteger(fn.id) ||
-            !Array.isArray(fn.params) ||
-            !Array.isArray(fn.places) ||
-            !Array.isArray(fn.blocks)) {
-            errors.push("malformed function");
-            continue;
-        }
-        for (const block of fn.blocks) {
-            if (!block ||
-                !Number.isInteger(block.id) ||
-                !Array.isArray(block.params) ||
-                !Array.isArray(block.ops) ||
-                !block.term ||
-                typeof block.term.op !== "string")
-                errors.push("malformed block");
+    try {
+        const decoded = structuredClone(input);
+        if (!validPackage(decoded)) {
+            const reasons = (validPackage.errors ?? []).map((issue) => `${issue.instancePath || "/"} ${issue.message ?? "invalid value"}`);
+            throw new Error(reasons.join("; ") || "malformed sandbox bytecode package");
         }
+        const pkg = decoded;
+        validatePackageGraph(pkg);
+        return pkg;
+    }
+    catch (error) {
+        throw new BytecodeValidationError([String(error)]);
     }
-    if (errors.length)
-        throw new BytecodeValidationError(errors);
-    return input;
 }

package/dist/interpreter/strings.d.ts

@@ -0,0 +1,21 @@
+export declare class IndexedString {
+    readonly kind = "string";
+    private storage;
+    private constructor();
+    static fromText(text: string): IndexedString;
+    private live;
+    get value(): string;
+    get scalarLength(): number;
+    get byteLength(): number;
+    encoded(): Uint8Array;
+    clone(): IndexedString;
+    release(): void;
+    scalarOffset(index: number): number | undefined;
+    at(index: number): string | undefined;
+    slice(start: number, end: number): IndexedString | undefined;
+    private scalarIndex;
+    find(needle: IndexedString, start?: number): number | undefined;
+    /** Consume this receiver, retaining the suffix's borrowed snapshot. */
+    append(suffix: IndexedString): IndexedString;
+}
+export declare function stringValue(value: string): IndexedString;

package/dist/interpreter/strings.js

@@ -0,0 +1,191 @@
+/** The VM's owned UTF-8 string storage. Scalar positions use the same
+ * bounded-stride index as native strings; no prefix is rescanned on access. */
+const INDEX_STRIDE = 32;
+const ENCODER = new TextEncoder();
+const DECODER = new TextDecoder("utf-8", { fatal: true, ignoreBOM: true });
+function scalarWidth(first) {
+    return first < 0x80 ? 1 : first < 0xe0 ? 2 : first < 0xf0 ? 3 : 4;
+}
+function indexStorage(storage, byte, scalar) {
+    const offsets = storage.offsets;
+    while (byte < storage.byteLength) {
+        if (scalar % INDEX_STRIDE === 0)
+            offsets[scalar / INDEX_STRIDE] = byte;
+        byte += scalarWidth(storage.bytes[byte]);
+        scalar++;
+    }
+}
+function newStorage(bytes, capacity = bytes.length) {
+    const buffer = new Uint8Array(capacity);
+    buffer.set(bytes);
+    let scalars = 0;
+    for (let at = 0; at < bytes.length; at += scalarWidth(bytes[at]))
+        scalars++;
+    const storage = {
+        bytes: buffer,
+        byteLength: bytes.length,
+        scalarLength: scalars,
+        offsets: scalars === bytes.length ? null : [],
+        owners: 1,
+    };
+    if (storage.offsets)
+        indexStorage(storage, 0, 0);
+    return storage;
+}
+export class IndexedString {
+    kind = "string";
+    storage;
+    constructor(storage) {
+        this.storage = storage;
+    }
+    static fromText(text) {
+        return new IndexedString(newStorage(ENCODER.encode(text)));
+    }
+    live() {
+        if (!this.storage)
+            throw new Error("cannot read a released string");
+        return this.storage;
+    }
+    get value() {
+        return DECODER.decode(this.encoded());
+    }
+    get scalarLength() {
+        return this.live().scalarLength;
+    }
+    get byteLength() {
+        return this.live().byteLength;
+    }
+    encoded() {
+        const storage = this.live();
+        return storage.bytes.subarray(0, storage.byteLength);
+    }
+    clone() {
+        const storage = this.live();
+        storage.owners++;
+        return new IndexedString(storage);
+    }
+    release() {
+        if (this.storage) {
+            this.storage.owners--;
+            this.storage = null;
+        }
+    }
+    scalarOffset(index) {
+        const storage = this.live();
+        if (!Number.isSafeInteger(index) ||
+            index < 0 ||
+            index > storage.scalarLength)
+            return undefined;
+        if (index === storage.scalarLength)
+            return storage.byteLength;
+        if (!storage.offsets)
+            return index;
+        let byte = storage.offsets[Math.floor(index / INDEX_STRIDE)];
+        for (let tail = index % INDEX_STRIDE; tail > 0; tail--)
+            byte += scalarWidth(storage.bytes[byte]);
+        return byte;
+    }
+    at(index) {
+        if (index === this.scalarLength)
+            return undefined;
+        const byte = this.scalarOffset(index);
+        if (byte === undefined)
+            return undefined;
+        const storage = this.live();
+        return DECODER.decode(storage.bytes.subarray(byte, byte + scalarWidth(storage.bytes[byte])));
+    }
+    slice(start, end) {
+        const from = this.scalarOffset(start);
+        const to = this.scalarOffset(end);
+        if (from === undefined || to === undefined || from > to)
+            return undefined;
+        return new IndexedString(newStorage(this.live().bytes.subarray(from, to)));
+    }
+    scalarIndex(byte) {
+        const storage = this.live();
+        if (!storage.offsets)
+            return byte;
+        if (byte === storage.byteLength)
+            return storage.scalarLength;
+        let low = 0;
+        let high = storage.offsets.length;
+        while (low + 1 < high) {
+            const mid = Math.floor((low + high) / 2);
+            if (storage.offsets[mid] <= byte)
+                low = mid;
+            else
+                high = mid;
+        }
+        let scalar = low * INDEX_STRIDE;
+        for (let at = storage.offsets[low]; at < byte; at += scalarWidth(storage.bytes[at]))
+            scalar++;
+        return scalar;
+    }
+    find(needle, start = 0) {
+        const from = this.scalarOffset(start);
+        if (from === undefined)
+            return undefined;
+        const bytes = this.encoded();
+        const pattern = needle.encoded();
+        if (pattern.length === 0)
+            return start;
+        // KMP searches only the suffix requested by the caller. Decoding the
+        // complete value for String.indexOf would rescan its prefix on every call.
+        const prefixes = new Array(pattern.length).fill(0);
+        for (let i = 1, matched = 0; i < pattern.length; i++) {
+            while (matched && pattern[i] !== pattern[matched])
+                matched = prefixes[matched - 1];
+            if (pattern[i] === pattern[matched])
+                matched++;
+            prefixes[i] = matched;
+        }
+        for (let i = from, matched = 0; i < bytes.length; i++) {
+            while (matched && bytes[i] !== pattern[matched])
+                matched = prefixes[matched - 1];
+            if (bytes[i] === pattern[matched])
+                matched++;
+            if (matched === pattern.length)
+                return this.scalarIndex(i + 1 - matched);
+        }
+        return undefined;
+    }
+    /** Consume this receiver, retaining the suffix's borrowed snapshot. */
+    append(suffix) {
+        const bytes = suffix.encoded();
+        const added = suffix.scalarLength;
+        if (!bytes.length)
+            return this;
+        let storage = this.live();
+        const oldLength = storage.byteLength;
+        const oldScalars = storage.scalarLength;
+        const newLength = oldLength + bytes.length;
+        const capacity = newLength > storage.bytes.length
+            ? Math.max(newLength, storage.bytes.length * 2, 16)
+            : storage.bytes.length;
+        if (storage.owners !== 1) {
+            storage.owners--;
+            storage = newStorage(this.encoded(), capacity);
+            this.storage = storage;
+        }
+        else if (capacity !== storage.bytes.length) {
+            const buffer = new Uint8Array(capacity);
+            buffer.set(storage.bytes.subarray(0, oldLength));
+            storage.bytes = buffer;
+        }
+        storage.bytes.set(bytes, oldLength);
+        storage.byteLength = newLength;
+        storage.scalarLength += added;
+        if (storage.scalarLength !== newLength) {
+            if (storage.offsets)
+                indexStorage(storage, oldLength, oldScalars);
+            else {
+                storage.offsets = [];
+                indexStorage(storage, 0, 0);
+            }
+        }
+        return this;
+    }
+}
+export function stringValue(value) {
+    return IndexedString.fromText(value);
+}

package/dist/interpreter/types.d.ts

@@ -117,7 +117,13 @@ export interface SandboxTrace {
     };
 }
 export interface RunOptions {
+    /** Explicit browser filesystem authority; absent roots grant no access. */
+    directoryRoots?: import("./v1/browser-directory.js").DirectoryRoots;
+    signal?: AbortSignal;
+    /** Supplied process arguments including argv[0]. Defaults to ["main.hew"]. */
+    argv?: readonly string[];
     codecBridge?: (request: string) => string;
+    calendarBridge?: (request: string) => string;
     propertyInput?: number[];
     fixtureId?: string;
     traceId?: string;

package/dist/interpreter/unicode-data.d.ts

@@ -0,0 +1,4 @@
+export declare const CATEGORY_RANGES: readonly number[];
+export declare const UPPER_MAPPING: readonly number[];
+export declare const LOWER_MAPPING: readonly number[];
+export declare const TITLE_MAPPING: readonly number[];