1 | const { isReadableStream } = require('./utils');
|
---|
2 | const TextDecoder = require('./text-decoder');
|
---|
3 |
|
---|
4 | const STACK_OBJECT = 1;
|
---|
5 | const STACK_ARRAY = 2;
|
---|
6 | const decoder = new TextDecoder();
|
---|
7 |
|
---|
8 | function isObject(value) {
|
---|
9 | return value !== null && typeof value === 'object';
|
---|
10 | }
|
---|
11 |
|
---|
12 | function adjustPosition(error, parser) {
|
---|
13 | if (error.name === 'SyntaxError' && parser.jsonParseOffset) {
|
---|
14 | error.message = error.message.replace(/at position (\d+)/, (_, pos) =>
|
---|
15 | 'at position ' + (Number(pos) + parser.jsonParseOffset)
|
---|
16 | );
|
---|
17 | }
|
---|
18 |
|
---|
19 | return error;
|
---|
20 | }
|
---|
21 |
|
---|
22 | function append(array, elements) {
|
---|
23 | // Note: Avoid to use array.push(...elements) since it may lead to
|
---|
24 | // "RangeError: Maximum call stack size exceeded" for a long arrays
|
---|
25 | const initialLength = array.length;
|
---|
26 | array.length += elements.length;
|
---|
27 |
|
---|
28 | for (let i = 0; i < elements.length; i++) {
|
---|
29 | array[initialLength + i] = elements[i];
|
---|
30 | }
|
---|
31 | }
|
---|
32 |
|
---|
33 | module.exports = function(chunkEmitter) {
|
---|
34 | let parser = new ChunkParser();
|
---|
35 |
|
---|
36 | if (isObject(chunkEmitter) && isReadableStream(chunkEmitter)) {
|
---|
37 | return new Promise((resolve, reject) => {
|
---|
38 | chunkEmitter
|
---|
39 | .on('data', chunk => {
|
---|
40 | try {
|
---|
41 | parser.push(chunk);
|
---|
42 | } catch (e) {
|
---|
43 | reject(adjustPosition(e, parser));
|
---|
44 | parser = null;
|
---|
45 | }
|
---|
46 | })
|
---|
47 | .on('error', (e) => {
|
---|
48 | parser = null;
|
---|
49 | reject(e);
|
---|
50 | })
|
---|
51 | .on('end', () => {
|
---|
52 | try {
|
---|
53 | resolve(parser.finish());
|
---|
54 | } catch (e) {
|
---|
55 | reject(adjustPosition(e, parser));
|
---|
56 | } finally {
|
---|
57 | parser = null;
|
---|
58 | }
|
---|
59 | });
|
---|
60 | });
|
---|
61 | }
|
---|
62 |
|
---|
63 | if (typeof chunkEmitter === 'function') {
|
---|
64 | const iterator = chunkEmitter();
|
---|
65 |
|
---|
66 | if (isObject(iterator) && (Symbol.iterator in iterator || Symbol.asyncIterator in iterator)) {
|
---|
67 | return new Promise(async (resolve, reject) => {
|
---|
68 | try {
|
---|
69 | for await (const chunk of iterator) {
|
---|
70 | parser.push(chunk);
|
---|
71 | }
|
---|
72 |
|
---|
73 | resolve(parser.finish());
|
---|
74 | } catch (e) {
|
---|
75 | reject(adjustPosition(e, parser));
|
---|
76 | } finally {
|
---|
77 | parser = null;
|
---|
78 | }
|
---|
79 | });
|
---|
80 | }
|
---|
81 | }
|
---|
82 |
|
---|
83 | throw new Error(
|
---|
84 | 'Chunk emitter should be readable stream, generator, ' +
|
---|
85 | 'async generator or function returning an iterable object'
|
---|
86 | );
|
---|
87 | };
|
---|
88 |
|
---|
89 | class ChunkParser {
|
---|
90 | constructor() {
|
---|
91 | this.value = undefined;
|
---|
92 | this.valueStack = null;
|
---|
93 |
|
---|
94 | this.stack = new Array(100);
|
---|
95 | this.lastFlushDepth = 0;
|
---|
96 | this.flushDepth = 0;
|
---|
97 | this.stateString = false;
|
---|
98 | this.stateStringEscape = false;
|
---|
99 | this.pendingByteSeq = null;
|
---|
100 | this.pendingChunk = null;
|
---|
101 | this.chunkOffset = 0;
|
---|
102 | this.jsonParseOffset = 0;
|
---|
103 | }
|
---|
104 |
|
---|
105 | parseAndAppend(fragment, wrap) {
|
---|
106 | // Append new entries or elements
|
---|
107 | if (this.stack[this.lastFlushDepth - 1] === STACK_OBJECT) {
|
---|
108 | if (wrap) {
|
---|
109 | this.jsonParseOffset--;
|
---|
110 | fragment = '{' + fragment + '}';
|
---|
111 | }
|
---|
112 |
|
---|
113 | Object.assign(this.valueStack.value, JSON.parse(fragment));
|
---|
114 | } else {
|
---|
115 | if (wrap) {
|
---|
116 | this.jsonParseOffset--;
|
---|
117 | fragment = '[' + fragment + ']';
|
---|
118 | }
|
---|
119 |
|
---|
120 | append(this.valueStack.value, JSON.parse(fragment));
|
---|
121 | }
|
---|
122 | }
|
---|
123 |
|
---|
124 | prepareAddition(fragment) {
|
---|
125 | const { value } = this.valueStack;
|
---|
126 | const expectComma = Array.isArray(value)
|
---|
127 | ? value.length !== 0
|
---|
128 | : Object.keys(value).length !== 0;
|
---|
129 |
|
---|
130 | if (expectComma) {
|
---|
131 | // Skip a comma at the beginning of fragment, otherwise it would
|
---|
132 | // fail to parse
|
---|
133 | if (fragment[0] === ',') {
|
---|
134 | this.jsonParseOffset++;
|
---|
135 | return fragment.slice(1);
|
---|
136 | }
|
---|
137 |
|
---|
138 | // When value (an object or array) is not empty and a fragment
|
---|
139 | // doesn't start with a comma, a single valid fragment starting
|
---|
140 | // is a closing bracket. If it's not, a prefix is adding to fail
|
---|
141 | // parsing. Otherwise, the sequence of chunks can be successfully
|
---|
142 | // parsed, although it should not, e.g. ["[{}", "{}]"]
|
---|
143 | if (fragment[0] !== '}' && fragment[0] !== ']') {
|
---|
144 | this.jsonParseOffset -= 3;
|
---|
145 | return '[[]' + fragment;
|
---|
146 | }
|
---|
147 | }
|
---|
148 |
|
---|
149 | return fragment;
|
---|
150 | }
|
---|
151 |
|
---|
152 | flush(chunk, start, end) {
|
---|
153 | let fragment = chunk.slice(start, end);
|
---|
154 |
|
---|
155 | // Save position correction an error in JSON.parse() if any
|
---|
156 | this.jsonParseOffset = this.chunkOffset + start;
|
---|
157 |
|
---|
158 | // Prepend pending chunk if any
|
---|
159 | if (this.pendingChunk !== null) {
|
---|
160 | fragment = this.pendingChunk + fragment;
|
---|
161 | this.jsonParseOffset -= this.pendingChunk.length;
|
---|
162 | this.pendingChunk = null;
|
---|
163 | }
|
---|
164 |
|
---|
165 | if (this.flushDepth === this.lastFlushDepth) {
|
---|
166 | // Depth didn't changed, so it's a root value or entry/element set
|
---|
167 | if (this.flushDepth > 0) {
|
---|
168 | this.parseAndAppend(this.prepareAddition(fragment), true);
|
---|
169 | } else {
|
---|
170 | // That's an entire value on a top level
|
---|
171 | this.value = JSON.parse(fragment);
|
---|
172 | this.valueStack = {
|
---|
173 | value: this.value,
|
---|
174 | prev: null
|
---|
175 | };
|
---|
176 | }
|
---|
177 | } else if (this.flushDepth > this.lastFlushDepth) {
|
---|
178 | // Add missed closing brackets/parentheses
|
---|
179 | for (let i = this.flushDepth - 1; i >= this.lastFlushDepth; i--) {
|
---|
180 | fragment += this.stack[i] === STACK_OBJECT ? '}' : ']';
|
---|
181 | }
|
---|
182 |
|
---|
183 | if (this.lastFlushDepth === 0) {
|
---|
184 | // That's a root value
|
---|
185 | this.value = JSON.parse(fragment);
|
---|
186 | this.valueStack = {
|
---|
187 | value: this.value,
|
---|
188 | prev: null
|
---|
189 | };
|
---|
190 | } else {
|
---|
191 | this.parseAndAppend(this.prepareAddition(fragment), true);
|
---|
192 | }
|
---|
193 |
|
---|
194 | // Move down to the depths to the last object/array, which is current now
|
---|
195 | for (let i = this.lastFlushDepth || 1; i < this.flushDepth; i++) {
|
---|
196 | let value = this.valueStack.value;
|
---|
197 |
|
---|
198 | if (this.stack[i - 1] === STACK_OBJECT) {
|
---|
199 | // find last entry
|
---|
200 | let key;
|
---|
201 | // eslint-disable-next-line curly
|
---|
202 | for (key in value);
|
---|
203 | value = value[key];
|
---|
204 | } else {
|
---|
205 | // last element
|
---|
206 | value = value[value.length - 1];
|
---|
207 | }
|
---|
208 |
|
---|
209 | this.valueStack = {
|
---|
210 | value,
|
---|
211 | prev: this.valueStack
|
---|
212 | };
|
---|
213 | }
|
---|
214 | } else /* this.flushDepth < this.lastFlushDepth */ {
|
---|
215 | fragment = this.prepareAddition(fragment);
|
---|
216 |
|
---|
217 | // Add missed opening brackets/parentheses
|
---|
218 | for (let i = this.lastFlushDepth - 1; i >= this.flushDepth; i--) {
|
---|
219 | this.jsonParseOffset--;
|
---|
220 | fragment = (this.stack[i] === STACK_OBJECT ? '{' : '[') + fragment;
|
---|
221 | }
|
---|
222 |
|
---|
223 | this.parseAndAppend(fragment, false);
|
---|
224 |
|
---|
225 | for (let i = this.lastFlushDepth - 1; i >= this.flushDepth; i--) {
|
---|
226 | this.valueStack = this.valueStack.prev;
|
---|
227 | }
|
---|
228 | }
|
---|
229 |
|
---|
230 | this.lastFlushDepth = this.flushDepth;
|
---|
231 | }
|
---|
232 |
|
---|
233 | push(chunk) {
|
---|
234 | if (typeof chunk !== 'string') {
|
---|
235 | // Suppose chunk is Buffer or Uint8Array
|
---|
236 |
|
---|
237 | // Prepend uncompleted byte sequence if any
|
---|
238 | if (this.pendingByteSeq !== null) {
|
---|
239 | const origRawChunk = chunk;
|
---|
240 | chunk = new Uint8Array(this.pendingByteSeq.length + origRawChunk.length);
|
---|
241 | chunk.set(this.pendingByteSeq);
|
---|
242 | chunk.set(origRawChunk, this.pendingByteSeq.length);
|
---|
243 | this.pendingByteSeq = null;
|
---|
244 | }
|
---|
245 |
|
---|
246 | // In case Buffer/Uint8Array, an input is encoded in UTF8
|
---|
247 | // Seek for parts of uncompleted UTF8 symbol on the ending
|
---|
248 | // This makes sense only if we expect more chunks and last char is not multi-bytes
|
---|
249 | if (chunk[chunk.length - 1] > 127) {
|
---|
250 | for (let seqLength = 0; seqLength < chunk.length; seqLength++) {
|
---|
251 | const byte = chunk[chunk.length - 1 - seqLength];
|
---|
252 |
|
---|
253 | // 10xxxxxx - 2nd, 3rd or 4th byte
|
---|
254 | // 110xxxxx – first byte of 2-byte sequence
|
---|
255 | // 1110xxxx - first byte of 3-byte sequence
|
---|
256 | // 11110xxx - first byte of 4-byte sequence
|
---|
257 | if (byte >> 6 === 3) {
|
---|
258 | seqLength++;
|
---|
259 |
|
---|
260 | // If the sequence is really incomplete, then preserve it
|
---|
261 | // for the future chunk and cut off it from the current chunk
|
---|
262 | if ((seqLength !== 4 && byte >> 3 === 0b11110) ||
|
---|
263 | (seqLength !== 3 && byte >> 4 === 0b1110) ||
|
---|
264 | (seqLength !== 2 && byte >> 5 === 0b110)) {
|
---|
265 | this.pendingByteSeq = chunk.slice(chunk.length - seqLength);
|
---|
266 | chunk = chunk.slice(0, -seqLength);
|
---|
267 | }
|
---|
268 |
|
---|
269 | break;
|
---|
270 | }
|
---|
271 | }
|
---|
272 | }
|
---|
273 |
|
---|
274 | // Convert chunk to a string, since single decode per chunk
|
---|
275 | // is much effective than decode multiple small substrings
|
---|
276 | chunk = decoder.decode(chunk);
|
---|
277 | }
|
---|
278 |
|
---|
279 | const chunkLength = chunk.length;
|
---|
280 | let lastFlushPoint = 0;
|
---|
281 | let flushPoint = 0;
|
---|
282 |
|
---|
283 | // Main scan loop
|
---|
284 | scan: for (let i = 0; i < chunkLength; i++) {
|
---|
285 | if (this.stateString) {
|
---|
286 | for (; i < chunkLength; i++) {
|
---|
287 | if (this.stateStringEscape) {
|
---|
288 | this.stateStringEscape = false;
|
---|
289 | } else {
|
---|
290 | switch (chunk.charCodeAt(i)) {
|
---|
291 | case 0x22: /* " */
|
---|
292 | this.stateString = false;
|
---|
293 | continue scan;
|
---|
294 |
|
---|
295 | case 0x5C: /* \ */
|
---|
296 | this.stateStringEscape = true;
|
---|
297 | }
|
---|
298 | }
|
---|
299 | }
|
---|
300 |
|
---|
301 | break;
|
---|
302 | }
|
---|
303 |
|
---|
304 | switch (chunk.charCodeAt(i)) {
|
---|
305 | case 0x22: /* " */
|
---|
306 | this.stateString = true;
|
---|
307 | this.stateStringEscape = false;
|
---|
308 | break;
|
---|
309 |
|
---|
310 | case 0x2C: /* , */
|
---|
311 | flushPoint = i;
|
---|
312 | break;
|
---|
313 |
|
---|
314 | case 0x7B: /* { */
|
---|
315 | // Open an object
|
---|
316 | flushPoint = i + 1;
|
---|
317 | this.stack[this.flushDepth++] = STACK_OBJECT;
|
---|
318 | break;
|
---|
319 |
|
---|
320 | case 0x5B: /* [ */
|
---|
321 | // Open an array
|
---|
322 | flushPoint = i + 1;
|
---|
323 | this.stack[this.flushDepth++] = STACK_ARRAY;
|
---|
324 | break;
|
---|
325 |
|
---|
326 | case 0x5D: /* ] */
|
---|
327 | case 0x7D: /* } */
|
---|
328 | // Close an object or array
|
---|
329 | flushPoint = i + 1;
|
---|
330 | this.flushDepth--;
|
---|
331 |
|
---|
332 | if (this.flushDepth < this.lastFlushDepth) {
|
---|
333 | this.flush(chunk, lastFlushPoint, flushPoint);
|
---|
334 | lastFlushPoint = flushPoint;
|
---|
335 | }
|
---|
336 |
|
---|
337 | break;
|
---|
338 |
|
---|
339 | case 0x09: /* \t */
|
---|
340 | case 0x0A: /* \n */
|
---|
341 | case 0x0D: /* \r */
|
---|
342 | case 0x20: /* space */
|
---|
343 | // Move points forward when they points on current position and it's a whitespace
|
---|
344 | if (lastFlushPoint === i) {
|
---|
345 | lastFlushPoint++;
|
---|
346 | }
|
---|
347 |
|
---|
348 | if (flushPoint === i) {
|
---|
349 | flushPoint++;
|
---|
350 | }
|
---|
351 |
|
---|
352 | break;
|
---|
353 | }
|
---|
354 | }
|
---|
355 |
|
---|
356 | if (flushPoint > lastFlushPoint) {
|
---|
357 | this.flush(chunk, lastFlushPoint, flushPoint);
|
---|
358 | }
|
---|
359 |
|
---|
360 | // Produce pendingChunk if something left
|
---|
361 | if (flushPoint < chunkLength) {
|
---|
362 | if (this.pendingChunk !== null) {
|
---|
363 | // When there is already a pending chunk then no flush happened,
|
---|
364 | // appending entire chunk to pending one
|
---|
365 | this.pendingChunk += chunk;
|
---|
366 | } else {
|
---|
367 | // Create a pending chunk, it will start with non-whitespace since
|
---|
368 | // flushPoint was moved forward away from whitespaces on scan
|
---|
369 | this.pendingChunk = chunk.slice(flushPoint, chunkLength);
|
---|
370 | }
|
---|
371 | }
|
---|
372 |
|
---|
373 | this.chunkOffset += chunkLength;
|
---|
374 | }
|
---|
375 |
|
---|
376 | finish() {
|
---|
377 | if (this.pendingChunk !== null) {
|
---|
378 | this.flush('', 0, 0);
|
---|
379 | this.pendingChunk = null;
|
---|
380 | }
|
---|
381 |
|
---|
382 | return this.value;
|
---|
383 | }
|
---|
384 | };
|
---|