Skip to content

Commit 8722aa9

Browse files
creationixcreationix
authored
Merge pull request #31 from hayes/big-strings
improve memory usage when parsing large tokens
2 parents 044b268 + 5643e52 commit 8722aa9

File tree

2 files changed

+104
-21
lines changed

2 files changed

+104
-21
lines changed

‎jsonparse.js

Lines changed: 80 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -40,13 +40,24 @@ var KEY = C.KEY = 0x72;
4040
// Parser Modes
4141
var OBJECT = C.OBJECT = 0x81;
4242
var ARRAY = C.ARRAY = 0x82;
43+
// Character constants
44+
var BACK_SLASH = "\\".charCodeAt(0);
45+
var FORWARD_SLASH = "\/".charCodeAt(0);
46+
var BACKSPACE = "\b".charCodeAt(0);
47+
var FORM_FEED = "\f".charCodeAt(0);
48+
var NEWLINE = "\n".charCodeAt(0);
49+
var CARRIAGE_RETURN = "\r".charCodeAt(0);
50+
var TAB = "\t".charCodeAt(0);
4351

52+
var STRING_BUFFER_SIZE = 64 * 1024;
4453

4554
function Parser() {
4655
this.tState = START;
4756
this.value = undefined;
4857

4958
this.string = undefined; // string data
59+
this.stringBuffer = Buffer.alloc ? Buffer.alloc(STRING_BUFFER_SIZE) : new Buffer(STRING_BUFFER_SIZE);
60+
this.stringBufferOffset = 0;
5061
this.unicode = undefined; // unicode escapes
5162

5263
this.key = undefined;
@@ -77,6 +88,41 @@ proto.charError = function (buffer, i) {
7788
this.tState = STOP;
7889
this.onError(new Error("Unexpected " + JSON.stringify(String.fromCharCode(buffer[i])) + " at position " + i + " in state " + Parser.toknam(this.tState)));
7990
};
91+
proto.appendStringChar = function (char) {
92+
if (this.stringBufferOffset >= STRING_BUFFER_SIZE) {
93+
this.string += this.stringBuffer.toString('utf8');
94+
this.stringBufferOffset = 0;
95+
}
96+
97+
this.stringBuffer[this.stringBufferOffset++] = char;
98+
};
99+
proto.appendStringBuf = function (buf, start, end) {
100+
var size = buf.length;
101+
if (typeof start === 'number') {
102+
if (typeof end === 'number') {
103+
if (end < 0) {
104+
// adding a negative end decreeses the size
105+
size = buf.length - start + end;
106+
} else {
107+
size = end - start;
108+
}
109+
} else {
110+
size = buf.length - start;
111+
}
112+
}
113+
114+
if (size < 0) {
115+
size = 0;
116+
}
117+
118+
if (this.stringBufferOffset + size > STRING_BUFFER_SIZE) {
119+
this.string += this.stringBuffer.toString('utf8', 0, this.stringBufferOffset);
120+
this.stringBufferOffset = 0;
121+
}
122+
123+
buf.copy(this.stringBuffer, this.stringBufferOffset, start, end);
124+
this.stringBufferOffset += size;
125+
};
80126
proto.write = function (buffer) {
81127
if (typeof buffer === "string") buffer = new Buffer(buffer);
82128
var n;
@@ -93,7 +139,10 @@ proto.write = function (buffer) {
93139
}else if(n === 0x74){ this.tState = TRUE1; // t
94140
}else if(n === 0x66){ this.tState = FALSE1; // f
95141
}else if(n === 0x6e){ this.tState = NULL1; // n
96-
}else if(n === 0x22){ this.string = ""; this.tState = STRING1; // "
142+
}else if(n === 0x22){ // "
143+
this.string = "";
144+
this.stringBufferOffset = 0;
145+
this.tState = STRING1;
97146
}else if(n === 0x2d){ this.string = "-"; this.tState = NUMBER1; // -
98147
}else{
99148
if (n >= 0x30 && n < 0x40) { // 1-9
@@ -112,7 +161,8 @@ proto.write = function (buffer) {
112161
for (var j = 0; j < this.bytes_remaining; j++) {
113162
this.temp_buffs[this.bytes_in_sequence][this.bytes_in_sequence - this.bytes_remaining + j] = buffer[j];
114163
}
115-
this.string += this.temp_buffs[this.bytes_in_sequence].toString();
164+
165+
this.appendStringBuf(this.temp_buffs[this.bytes_in_sequence]);
116166
this.bytes_in_sequence = this.bytes_remaining = 0;
117167
i = i + j - 1;
118168
} else if (this.bytes_remaining === 0 && n >= 128) { // else if no remainder bytes carried over, parse multi byte (>=128) chars one at a time
@@ -129,38 +179,47 @@ proto.write = function (buffer) {
129179
this.bytes_remaining = (i + this.bytes_in_sequence) - buffer.length;
130180
i = buffer.length - 1;
131181
} else {
132-
this.string += buffer.slice(i, (i + this.bytes_in_sequence)).toString();
182+
this.appendStringBuf(buffer, i, i + this.bytes_in_sequence);
133183
i = i + this.bytes_in_sequence - 1;
134184
}
135-
} else if (n === 0x22) { this.tState = START; this.onToken(STRING, this.string); this.offset += Buffer.byteLength(this.string, 'utf8') + 1; this.string = undefined; }
136-
else if (n === 0x5c) { this.tState = STRING2; }
137-
else if (n >= 0x20) { this.string += String.fromCharCode(n); }
185+
} else if (n === 0x22) {
186+
this.tState = START;
187+
this.string += this.stringBuffer.toString('utf8', 0, this.stringBufferOffset);
188+
this.stringBufferOffset = 0;
189+
this.onToken(STRING, this.string);
190+
this.offset += Buffer.byteLength(this.string, 'utf8') + 1;
191+
this.string = undefined;
192+
}
193+
else if (n === 0x5c) {
194+
this.tState = STRING2;
195+
}
196+
else if (n >= 0x20) { this.appendStringChar(n); }
138197
else {
139198
return this.charError(buffer, i);
140199
}
141200
}else if (this.tState === STRING2){ // After backslash
142201
n = buffer[i];
143-
if(n === 0x22){ this.string += "\""; this.tState = STRING1;
144-
}else if(n === 0x5c){ this.string += "\\"; this.tState = STRING1;
145-
}else if(n === 0x2f){ this.string += "\/"; this.tState = STRING1;
146-
}else if(n === 0x62){ this.string += "\b"; this.tState = STRING1;
147-
}else if(n === 0x66){ this.string += "\f"; this.tState = STRING1;
148-
}else if(n === 0x6e){ this.string += "\n"; this.tState = STRING1;
149-
}else if(n === 0x72){ this.string += "\r"; this.tState = STRING1;
150-
}else if(n === 0x74){ this.string += "\t"; this.tState = STRING1;
202+
if(n === 0x22){ this.appendStringChar(n); this.tState = STRING1;
203+
}else if(n === 0x5c){ this.appendStringChar(BACK_SLASH); this.tState = STRING1;
204+
}else if(n === 0x2f){ this.appendStringChar(FORWARD_SLASH); this.tState = STRING1;
205+
}else if(n === 0x62){ this.appendStringChar(BACKSPACE); this.tState = STRING1;
206+
}else if(n === 0x66){ this.appendStringChar(FORM_FEED); this.tState = STRING1;
207+
}else if(n === 0x6e){ this.appendStringChar(NEWLINE); this.tState = STRING1;
208+
}else if(n === 0x72){ this.appendStringChar(CARRIAGE_RETURN); this.tState = STRING1;
209+
}else if(n === 0x74){ this.appendStringChar(TAB); this.tState = STRING1;
151210
}else if(n === 0x75){ this.unicode = ""; this.tState = STRING3;
152-
}else{
153-
return this.charError(buffer, i);
211+
}else{
212+
return this.charError(buffer, i);
154213
}
155214
}else if (this.tState === STRING3 || this.tState === STRING4 || this.tState === STRING5 || this.tState === STRING6){ // unicode hex codes
156215
n = buffer[i];
157216
// 0-9 A-F a-f
158217
if ((n >= 0x30 && n < 0x40) || (n > 0x40 && n <= 0x46) || (n > 0x60 && n <= 0x66)) {
159218
this.unicode += String.fromCharCode(n);
160219
if (this.tState++ === STRING6) {
161-
this.string += String.fromCharCode(parseInt(this.unicode, 16));
220+
this.appendStringBuf(Buffer(String.fromCharCode(parseInt(this.unicode, 16))));
162221
this.unicode = undefined;
163-
this.tState = STRING1;
222+
this.tState = STRING1;
164223
}
165224
} else {
166225
return this.charError(buffer, i);
@@ -266,14 +325,14 @@ proto.emit = function (value) {
266325
};
267326
proto.onValue = function (value) {
268327
// Override me
269-
};
328+
};
270329
proto.onToken = function (token, value) {
271330
if(this.state === VALUE){
272331
if(token === STRING || token === NUMBER || token === TRUE || token === FALSE || token === NULL){
273332
if (this.value) {
274333
this.value[this.key] = value;
275334
}
276-
this.emit(value);
335+
this.emit(value);
277336
}else if(token === LEFT_BRACE){
278337
this.push();
279338
if (this.value) {
@@ -322,7 +381,7 @@ proto.onToken = function (token, value) {
322381
if (token === COLON) { this.state = VALUE; }
323382
else { return this.parseError(token, value); }
324383
}else if(this.state === COMMA){
325-
if (token === COMMA) {
384+
if (token === COMMA) {
326385
if (this.mode === ARRAY) { this.key++; this.state = VALUE; }
327386
else if (this.mode === OBJECT) { this.state = KEY; }
328387

‎test/big-token.js

Lines changed: 24 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,24 @@
1+
var stream = require('stream');
2+
var JsonParse = require('../jsonparse');
3+
var test = require('tape');
4+
5+
test('can handle large tokens without running out of memory', function (t) {
6+
var parser = new JsonParse();
7+
var chunkSize = 1024;
8+
var chunks = 1024 * 200; // 200mb
9+
var quote = Buffer.from ? Buffer.from('"') : new Buffer('"');
10+
t.plan(1);
11+
12+
parser.onToken = function (type, value) {
13+
t.equal(value.length, chunkSize * chunks, 'token should be size of input json');
14+
t.end();
15+
};
16+
17+
parser.write(quote);
18+
for (var i = 0; i < chunks; ++i) {
19+
var buf = Buffer.alloc ? Buffer.alloc(chunkSize) : new Buffer(chunkSize);
20+
buf.fill('a');
21+
parser.write(buf);
22+
}
23+
parser.write(quote);
24+
});

0 commit comments

Comments
 (0)