preserves/implementations/javascript/packages/core/src/decoder.ts

385 lines
12 KiB
TypeScript

import { Annotated } from "./annotated";
import { DecodeError, ShortPacket } from "./codec";
import { Tag } from "./constants";
import { Set, Dictionary } from "./dictionary";
import { DoubleFloat, SingleFloat } from "./float";
import { Record } from "./record";
import { Bytes, BytesLike, underlying } from "./bytes";
import { Value } from "./values";
import { is } from "./is";
import { embed, GenericEmbedded, Embedded, EmbeddedTypeDecode } from "./embedded";
export interface DecoderOptions {
includeAnnotations?: boolean;
}
export interface DecoderEmbeddedOptions<T> extends DecoderOptions {
embeddedDecode?: EmbeddedTypeDecode<T>;
}
export interface TypedDecoder<T> {
atEnd(): boolean;
mark(): any;
restoreMark(m: any): void;
skip(): void;
next(): Value<T>;
withEmbeddedDecode<S, R>(
embeddedDecode: EmbeddedTypeDecode<S>,
body: (d: TypedDecoder<S>) => R): R;
nextBoolean(): boolean | undefined;
nextFloat(): SingleFloat | undefined;
nextDouble(): DoubleFloat | undefined;
nextEmbedded(): Embedded<T> | undefined;
nextSignedInteger(): number | undefined;
nextString(): string | undefined;
nextByteString(): Bytes | undefined;
nextSymbol(): symbol | undefined;
openRecord(): boolean;
openSequence(): boolean;
openSet(): boolean;
openDictionary(): boolean;
closeCompound(): boolean;
}
export function asLiteral<T, E extends Exclude<Value<T>, Annotated<T>>>(
actual: Value<T>,
expected: E): E | undefined
{
return is(actual, expected) ? expected : void 0;
}
export class DecoderState {
packet: Uint8Array;
index = 0;
options: DecoderOptions;
constructor(packet: BytesLike, options: DecoderOptions) {
this.packet = underlying(packet);
this.options = options;
}
get includeAnnotations(): boolean {
return this.options.includeAnnotations ?? false;
}
write(data: BytesLike) {
if (this.index === this.packet.length) {
this.packet = underlying(data);
} else {
this.packet = Bytes.concat([this.packet.slice(this.index), data])._view;
}
this.index = 0;
}
atEnd(): boolean {
return this.index >= this.packet.length;
}
mark(): number {
return this.index;
}
restoreMark(m: number): void {
this.index = m;
}
shortGuard<R>(body: () => R, short: () => R): R {
if (this.atEnd()) return short();
// ^ important somewhat-common case optimization - avoid the exception
const start = this.mark();
try {
return body();
} catch (e) {
if (ShortPacket.isShortPacket(e)) {
this.restoreMark(start);
return short();
}
throw e;
}
}
nextbyte(): number {
if (this.atEnd()) throw new ShortPacket("Short packet");
return this.packet[this.index++];
}
nextbytes(n: number): DataView {
const start = this.index;
this.index += n;
if (this.index > this.packet.length) throw new ShortPacket("Short packet");
// ^ NOTE: greater-than, not greater-than-or-equal-to - this makes atEnd() inappropriate
return new DataView(this.packet.buffer, this.packet.byteOffset + start, n);
}
varint(): number {
// TODO: Bignums :-/
const v = this.nextbyte();
if (v < 128) return v;
return (this.varint() << 7) + (v - 128);
}
peekend(): boolean {
return (this.nextbyte() === Tag.End) || (this.index--, false);
}
nextint(n: number): number {
// TODO: Bignums :-/
if (n === 0) return 0;
let acc = this.nextbyte();
if (acc & 0x80) acc -= 256;
for (let i = 1; i < n; i++) acc = (acc * 256) + this.nextbyte();
return acc;
}
nextSmallOrMediumInteger(tag: number): number | undefined {
if (tag >= Tag.SmallInteger_lo && tag <= Tag.SmallInteger_lo + 15) {
const v = tag - Tag.SmallInteger_lo;
return v > 12 ? v - 16 : v;
}
if (tag >= Tag.MediumInteger_lo && tag <= Tag.MediumInteger_lo + 15) {
const n = tag - Tag.MediumInteger_lo;
return this.nextint(n + 1);
}
return void 0;
}
wrap<T>(v: Value<T>): Value<T> {
return this.includeAnnotations ? new Annotated(v) : v;
}
unshiftAnnotation<T>(a: Value<T>, v: Annotated<T>): Annotated<T> {
if (this.includeAnnotations) {
v.annotations.unshift(a);
}
return v;
}
}
export const neverEmbeddedTypeDecode: EmbeddedTypeDecode<never> = {
decode(_s: DecoderState): never {
throw new Error("Embeddeds not permitted at this point in Preserves document");
},
fromValue(_v: Value<GenericEmbedded>): never {
throw new Error("Embeddeds not permitted at this point in Preserves document");
},
};
export class Decoder<T = never> implements TypedDecoder<T> {
state: DecoderState;
embeddedDecode: EmbeddedTypeDecode<T>;
constructor(state: DecoderState, embeddedDecode?: EmbeddedTypeDecode<T>);
constructor(packet?: BytesLike, options?: DecoderEmbeddedOptions<T>);
constructor(
packet_or_state: (DecoderState | BytesLike) = new Uint8Array(0),
options_or_embeddedDecode?: (DecoderEmbeddedOptions<T> | EmbeddedTypeDecode<T>))
{
if (packet_or_state instanceof DecoderState) {
this.state = packet_or_state;
this.embeddedDecode = (options_or_embeddedDecode as EmbeddedTypeDecode<T>) ?? neverEmbeddedTypeDecode;
} else {
const options = (options_or_embeddedDecode as DecoderEmbeddedOptions<T>) ?? {};
this.state = new DecoderState(packet_or_state, options);
this.embeddedDecode = options.embeddedDecode ?? neverEmbeddedTypeDecode;
}
}
write(data: BytesLike) {
this.state.write(data);
}
nextvalues(): Value<T>[] {
const result = [];
while (!this.state.peekend()) result.push(this.next());
return result;
}
static dictionaryFromArray<T>(vs: Value<T>[]): Dictionary<T> {
const d = new Dictionary<T>();
if (vs.length % 2) throw new DecodeError("Missing dictionary value");
for (let i = 0; i < vs.length; i += 2) {
d.set(vs[i], vs[i+1]);
}
return d;
}
next(): Value<T> {
const tag = this.state.nextbyte();
switch (tag) {
case Tag.False: return this.state.wrap<T>(false);
case Tag.True: return this.state.wrap<T>(true);
case Tag.Float: return this.state.wrap<T>(new SingleFloat(this.state.nextbytes(4).getFloat32(0, false)));
case Tag.Double: return this.state.wrap<T>(new DoubleFloat(this.state.nextbytes(8).getFloat64(0, false)));
case Tag.End: throw new DecodeError("Unexpected Compound end marker");
case Tag.Annotation: {
const a = this.next();
const v = this.next() as Annotated<T>;
return this.state.unshiftAnnotation(a, v);
}
case Tag.Embedded: return this.state.wrap<T>(embed(this.embeddedDecode.decode(this.state)));
case Tag.SignedInteger: return this.state.wrap<T>(this.state.nextint(this.state.varint()));
case Tag.String: return this.state.wrap<T>(Bytes.from(this.state.nextbytes(this.state.varint())).fromUtf8());
case Tag.ByteString: return this.state.wrap<T>(Bytes.from(this.state.nextbytes(this.state.varint())));
case Tag.Symbol: return this.state.wrap<T>(Symbol.for(Bytes.from(this.state.nextbytes(this.state.varint())).fromUtf8()));
case Tag.Record: {
const vs = this.nextvalues();
if (vs.length === 0) throw new DecodeError("Too few elements in encoded record");
return this.state.wrap<T>(Record(vs[0], vs.slice(1)));
}
case Tag.Sequence: return this.state.wrap<T>(this.nextvalues());
case Tag.Set: return this.state.wrap<T>(new Set(this.nextvalues()));
case Tag.Dictionary: return this.state.wrap<T>(Decoder.dictionaryFromArray(this.nextvalues()));
default: {
const v = this.state.nextSmallOrMediumInteger(tag);
if (v === void 0) {
throw new DecodeError("Unsupported Preserves tag: " + tag);
}
return this.state.wrap<T>(v);
}
}
}
try_next(): Value<T> | undefined {
return this.state.shortGuard(() => this.next(), () => void 0);
}
atEnd(): boolean {
return this.state.atEnd();
}
mark(): any {
return this.state.mark();
}
restoreMark(m: any): void {
this.state.restoreMark(m);
}
skip(): void {
// TODO: be more efficient
this.next();
}
withEmbeddedDecode<S, R>(
embeddedDecode: EmbeddedTypeDecode<S>,
body: (d: TypedDecoder<S>) => R): R
{
return body(new Decoder(this.state, embeddedDecode));
}
skipAnnotations(): void {
if (!this.state.atEnd() && this.state.packet[this.state.index] === Tag.Annotation) {
this.state.index++;
this.skip();
}
}
nextBoolean(): boolean | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.False: return false;
case Tag.True: return true;
default: return void 0;
}
}
nextFloat(): SingleFloat | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.Float: return new SingleFloat(this.state.nextbytes(4).getFloat32(0, false));
default: return void 0;
}
}
nextDouble(): DoubleFloat | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.Double: return new DoubleFloat(this.state.nextbytes(8).getFloat64(0, false));
default: return void 0;
}
}
nextEmbedded(): Embedded<T> | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.Embedded: return embed(this.embeddedDecode.decode(this.state));
default: return void 0;
}
}
nextSignedInteger(): number | undefined {
this.skipAnnotations();
const b = this.state.nextbyte();
switch (b) {
case Tag.SignedInteger: return this.state.nextint(this.state.varint());
default: return this.state.nextSmallOrMediumInteger(b);
}
}
nextString(): string | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.String: return Bytes.from(this.state.nextbytes(this.state.varint())).fromUtf8();
default: return void 0;
}
}
nextByteString(): Bytes | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.ByteString: return Bytes.from(this.state.nextbytes(this.state.varint()));
default: return void 0;
}
}
nextSymbol(): symbol | undefined {
this.skipAnnotations();
switch (this.state.nextbyte()) {
case Tag.Symbol:
return Symbol.for(Bytes.from(this.state.nextbytes(this.state.varint())).fromUtf8());
default:
return void 0;
}
}
openRecord(): boolean {
this.skipAnnotations();
return (this.state.nextbyte() === Tag.Record) || (this.state.index--, false);
}
openSequence(): boolean {
this.skipAnnotations();
return (this.state.nextbyte() === Tag.Sequence) || (this.state.index--, false);
}
openSet(): boolean {
this.skipAnnotations();
return (this.state.nextbyte() === Tag.Set) || (this.state.index--, false);
}
openDictionary(): boolean {
this.skipAnnotations();
return (this.state.nextbyte() === Tag.Dictionary) || (this.state.index--, false);
}
closeCompound(): boolean {
return this.state.peekend();
}
}
export function decode<T>(bs: BytesLike, options: DecoderEmbeddedOptions<T> = {}): Value<T> {
return new Decoder(bs, options).next();
}
export function decodeWithAnnotations<T>(bs: BytesLike,
options: DecoderEmbeddedOptions<T> = {}): Annotated<T> {
return decode(bs, { ... options, includeAnnotations: true }) as Annotated<T>;
}