1 | // a transform stream is a readable/writable stream where you do
|
2 | // something with the data. Sometimes it's called a "filter",
|
3 | // but that's not a great name for it, since that implies a thing where
|
4 | // some bits pass through, and others are simply ignored. (That would
|
5 | // be a valid example of a transform, of course.)
|
6 | //
|
7 | // While the output is causally related to the input, it's not a
|
8 | // necessarily symmetric or synchronous transformation. For example,
|
9 | // a zlib stream might take multiple plain-text writes(), and then
|
10 | // emit a single compressed chunk some time in the future.
|
11 | //
|
12 | // Here's how this works:
|
13 | //
|
14 | // The Transform stream has all the aspects of the readable and writable
|
15 | // stream classes. When you write(chunk), that calls _write(chunk,cb)
|
16 | // internally, and returns false if there's a lot of pending writes
|
17 | // buffered up. When you call read(), that calls _read(n) until
|
18 | // there's enough pending readable data buffered up.
|
19 | //
|
20 | // In a transform stream, the written data is placed in a buffer. When
|
21 | // _read(n) is called, it transforms the queued up data, calling the
|
22 | // buffered _write cb's as it consumes chunks. If consuming a single
|
23 | // written chunk would result in multiple output chunks, then the first
|
24 | // outputted bit calls the readcb, and subsequent chunks just go into
|
25 | // the read buffer, and will cause it to emit 'readable' if necessary.
|
26 | //
|
27 | // This way, back-pressure is actually determined by the reading side,
|
28 | // since _read has to be called to start processing a new chunk. However,
|
29 | // a pathological inflate type of transform can cause excessive buffering
|
30 | // here. For example, imagine a stream where every byte of input is
|
31 | // interpreted as an integer from 0-255, and then results in that many
|
32 | // bytes of output. Writing the 4 bytes {ff,ff,ff,ff} would result in
|
33 | // 1kb of data being output. In this case, you could write a very small
|
34 | // amount of input, and end up with a very large amount of output. In
|
35 | // such a pathological inflating mechanism, there'd be no way to tell
|
36 | // the system to stop doing the transform. A single 4MB write could
|
37 | // cause the system to run out of memory.
|
38 | //
|
39 | // However, even in such a pathological case, only a single written chunk
|
40 | // would be consumed, and then the rest would wait (un-transformed) until
|
41 | // the results of the previous transformed chunk were consumed.
|
42 |
|
43 |
|
44 | import {Duplex} from './duplex';
|
45 |
|
46 |
|
47 | import {inherits} from 'util';
|
48 | inherits(Transform, Duplex);
|
49 |
|
50 | function TransformState(stream) {
|
51 | this.afterTransform = function (er, data) {
|
52 | return afterTransform(stream, er, data);
|
53 | };
|
54 |
|
55 | this.needTransform = false;
|
56 | this.transforming = false;
|
57 | this.writecb = null;
|
58 | this.writechunk = null;
|
59 | this.writeencoding = null;
|
60 | }
|
61 |
|
62 | function afterTransform(stream, er, data) {
|
63 | var ts = stream._transformState;
|
64 | ts.transforming = false;
|
65 |
|
66 | var cb = ts.writecb;
|
67 |
|
68 | if (!cb) return stream.emit('error', new Error('no writecb in Transform class'));
|
69 |
|
70 | ts.writechunk = null;
|
71 | ts.writecb = null;
|
72 |
|
73 | if (data !== null && data !== undefined) stream.push(data);
|
74 |
|
75 | cb(er);
|
76 |
|
77 | var rs = stream._readableState;
|
78 | rs.reading = false;
|
79 | if (rs.needReadable || rs.length < rs.highWaterMark) {
|
80 | stream._read(rs.highWaterMark);
|
81 | }
|
82 | }
|
83 | export default Transform;
|
84 | export function Transform(options) {
|
85 | if (!(this instanceof Transform)) return new Transform(options);
|
86 |
|
87 | Duplex.call(this, options);
|
88 |
|
89 | this._transformState = new TransformState(this);
|
90 |
|
91 | // when the writable side finishes, then flush out anything remaining.
|
92 | var stream = this;
|
93 |
|
94 | // start out asking for a readable event once data is transformed.
|
95 | this._readableState.needReadable = true;
|
96 |
|
97 | // we have implemented the _read method, and done the other things
|
98 | // that Readable wants before the first _read call, so unset the
|
99 | // sync guard flag.
|
100 | this._readableState.sync = false;
|
101 |
|
102 | if (options) {
|
103 | if (typeof options.transform === 'function') this._transform = options.transform;
|
104 |
|
105 | if (typeof options.flush === 'function') this._flush = options.flush;
|
106 | }
|
107 |
|
108 | this.once('prefinish', function () {
|
109 | if (typeof this._flush === 'function') this._flush(function (er) {
|
110 | done(stream, er);
|
111 | });else done(stream);
|
112 | });
|
113 | }
|
114 |
|
115 | Transform.prototype.push = function (chunk, encoding) {
|
116 | this._transformState.needTransform = false;
|
117 | return Duplex.prototype.push.call(this, chunk, encoding);
|
118 | };
|
119 |
|
120 | // This is the part where you do stuff!
|
121 | // override this function in implementation classes.
|
122 | // 'chunk' is an input chunk.
|
123 | //
|
124 | // Call `push(newChunk)` to pass along transformed output
|
125 | // to the readable side. You may call 'push' zero or more times.
|
126 | //
|
127 | // Call `cb(err)` when you are done with this chunk. If you pass
|
128 | // an error, then that'll put the hurt on the whole operation. If you
|
129 | // never call cb(), then you'll never get another chunk.
|
130 | Transform.prototype._transform = function (chunk, encoding, cb) {
|
131 | throw new Error('Not implemented');
|
132 | };
|
133 |
|
134 | Transform.prototype._write = function (chunk, encoding, cb) {
|
135 | var ts = this._transformState;
|
136 | ts.writecb = cb;
|
137 | ts.writechunk = chunk;
|
138 | ts.writeencoding = encoding;
|
139 | if (!ts.transforming) {
|
140 | var rs = this._readableState;
|
141 | if (ts.needTransform || rs.needReadable || rs.length < rs.highWaterMark) this._read(rs.highWaterMark);
|
142 | }
|
143 | };
|
144 |
|
145 | // Doesn't matter what the args are here.
|
146 | // _transform does all the work.
|
147 | // That we got here means that the readable side wants more data.
|
148 | Transform.prototype._read = function (n) {
|
149 | var ts = this._transformState;
|
150 |
|
151 | if (ts.writechunk !== null && ts.writecb && !ts.transforming) {
|
152 | ts.transforming = true;
|
153 | this._transform(ts.writechunk, ts.writeencoding, ts.afterTransform);
|
154 | } else {
|
155 | // mark that we need a transform, so that any data that comes in
|
156 | // will get processed, now that we've asked for it.
|
157 | ts.needTransform = true;
|
158 | }
|
159 | };
|
160 |
|
161 | function done(stream, er) {
|
162 | if (er) return stream.emit('error', er);
|
163 |
|
164 | // if there's nothing in the write buffer, then that means
|
165 | // that nothing more will ever be provided
|
166 | var ws = stream._writableState;
|
167 | var ts = stream._transformState;
|
168 |
|
169 | if (ws.length) throw new Error('Calling transform done when ws.length != 0');
|
170 |
|
171 | if (ts.transforming) throw new Error('Calling transform done when still transforming');
|
172 |
|
173 | return stream.push(null);
|
174 | }
|