buffer: implement iterable interface
#204
Conversation
Dtrace probes were removed from libuv recently, but their usage by node was not completely removed, causing build breaks on SmartOS. Even though the build is working on other platforms, these probes are not fired by libuv anymore, so there's no point in using them on these platforms too. Reviewed-by: Trevor Norris <trev.norris@gmail.com>
PR-URL: nodejs/node-v0.x-archive#8847 Reviewed-by: Trevor Norris <trev.norris@gmail.com>
Marking these two tests as flaky, since they have been failing intermittenly in recent builds: test-debug-signal-cluster test-cluster-basic
PR-URL: nodejs/node-v0.x-archive#8546 Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
PR-URL: nodejs/node-v0.x-archive#8845 Reviewed-by: Trevor Norris <trev.norris@gmail.com>
Clarify the fd option: it is preferred to the path parameter, omits the "open" event if given, and is available on WriteStreams as well. PR-URL: nodejs/node-v0.x-archive#7707 Fixes: nodejs/node-v0.x-archive#7707 Fixes: nodejs/node-v0.x-archive#7708 Fixes: nodejs/node-v0.x-archive#4367 Reviewed-By: Chris Dickinson <christopher.s.dickinson@gmail.com>
Reviewed-By: Fedor Indutny <fedor@indutny.com> PR-URL: nodejs/node-v0.x-archive#6442
Fix Interface.setBreakpoint() to correctly handle an attempt to set a breakpoint in the current script when there is no current script. This usually happens when the debugged process is not paused. Fixes: nodejs/node-v0.x-archive#6453 PR-URL: nodejs/node-v0.x-archive#6460 Reviewed-By: Chris Dickinson <christopher.s.dickinson@gmail.com>
Fix a Windows-only build error that was introduced in commit 1183ba4 ("zlib: support concatenated gzip files"). Rename the NO_ERROR and FAILED enumerations, they conflict with macros of the same name in <winerror.h>. PR-URL: nodejs/node-v0.x-archive#8893 Reviewed-By: Fedor Indutny <fedor@indutny.com> Reviewed-By: Rod Vagg <rod@vagg.org> Reviewed-by: Timothy J Fontaine <tjfontaine@gmail.com>
In cases where many small writes are made to a stream lacking _writev, the array data structure backing the WriteReq buffer would greatly increase GC pressure. Specifically, in the fs.WriteStream case, the clearBuffer routine would only clear a single WriteReq from the buffer before exiting, but would cause the entire backing array to be GC'd. Switching to [].shift lessened pressure, but still the bulk of the time was spent in memcpy. This replaces that structure with a linked list-backed queue so that adding and removing from the queue is O(1). In the _writev case, collecting the buffer requires an O(N) loop over the buffer, but that was already being performed to collect callbacks, so slowdown should be neglible. PR-URL: nodejs/node-v0.x-archive#8826 Reviewed-by: Timothy J Fontaine <tjfontaine@gmail.com> Reviewed-by: Trevor Norris <trev.norris@gmail.com>
Add documentation for the callback parameter of http.ClientRequest's and http.ServerResponse's end methods. Signed-off-by: Julien Gilli <julien.gilli@joyent.com>
| // 2 | ||
| // 3 | ||
|
|
||
| Additionaly, `buffer.values()`, `buffer.keys()` and `buffer.entries()` |
There was a problem hiding this comment.
s/Additionaly/Additionally/
vkurchatkin
force-pushed
the
iterable-buffer
branch
from
43ff8f6
to
685ec25
Compare
| if (kind === ITERATOR_KIND_ENTRIES) | ||
| return new BufferIteratorResult([index, buffer[index]], false); | ||
|
|
||
| return new BufferIteratorResult(index, false); |
There was a problem hiding this comment.
I'm curious -- can iterators reuse results? If so that might be ideal, otherwise for large buffers we might see a huge number of throwaway objects generated when iterated with for... of.
There was a problem hiding this comment.
Wouldn't using flyweights here just be reinventing smis, badly? If you really wanted to do that, though, you could probably do it in the BufferIteratorResult constructor by having it return from a cache of previously-seen values. There's only 256 potential values when iterating via for..of, right?
There was a problem hiding this comment.
I hope this would be somehow optimized eventually. I mean, the whole for..of thing.
There was a problem hiding this comment.
In my unscientific benchmarking, using a cache makes it about 15% faster.
Benchmark:
var buf = Buffer(1 << 26), idx = 0;
buf.fill(0); // Touch memory, force delayed allocation.
for (var b of buf) buf[idx++] = b;Cache:
diff --git a/lib/buffer.js b/lib/buffer.js
index 9149b65..eb75761 100644
--- a/lib/buffer.js
+++ b/lib/buffer.js
@@ -969,6 +969,11 @@ function BufferIteratorResult(value, done) {
this.done = done;
}
+var cache = new Array(256);
+
+for (var i = 0; i < cache.length; i += 1)
+ cache[i] = new BufferIteratorResult(i, false);
+
BufferIterator.prototype.next = function() {
var buffer = this._buffer;
var kind = this._kind;
@@ -980,7 +985,7 @@ BufferIterator.prototype.next = function() {
this._index++;
if (kind === ITERATOR_KIND_VALUES)
- return new BufferIteratorResult(buffer[index], false);
+ return cache[buffer[index]];
if (kind === ITERATOR_KIND_ENTRIES)
return new BufferIteratorResult([index, buffer[index]], false);Peak RSS and the number of page faults with that benchmark are consistently and repeatedly 2-3% lower so it makes sense that it's faster.
On a related topic, it would be good to add one or two benchmarks in benchmark/buffers/
There was a problem hiding this comment.
@othiym23 ah, I was thinking that each iterator could instantiate a single BufferIteratorResult and swap its value on .next() -- to avoid the problem of having to instantiate a whole new object just to add a .done parameter. Could you expand on the "reinventing smis, badly" bit? I'm afraid I don't follow.
@vkurchatkin In that these objects will probably only exist in the new-generation pool and be quickly GC'd, it's optimized, though it would still increase GC pressure (I think.) If we can get away it, I imagine only creating two objects per iteration (vs. N + 1) would be less expensive.
There was a problem hiding this comment.
@chrisdickinson I was actually thinking that it's possible to avoid object creation at all in for..of. Unfortunately, iterators can be consumed directly and it's possible to mutate result objects:
var b = new Buffer(255);
for (var i = 0; i < 255; i++)
b[i] = i;
var it = b[Symbol.iterator](), r;
while ((r = it.next()) && !r.done)
r.value = Math.floor(Math.random() * 255);There was a problem hiding this comment.
Is there anything in the spec that forbids making the the cached objects immutable with Object.freeze()?
There was a problem hiding this comment.
Well, I think result object are supposed to be simple mutable data objects (https://people.mozilla.org/~jorendorff/es6-draft.html#sec-createiterresultobject). It is not specified anywhere that custom iterators should work the same way (https://people.mozilla.org/~jorendorff/es6-draft.html#sec-iteratorresult-interface), but I think they should.
There was a problem hiding this comment.
Object.freeze is definitely allowed.
vkurchatkin
force-pushed
the
iterable-buffer
branch
from
685ec25
to
299bf12
Compare
|
I think a simpler version of this patch (untested) would be Buffer.prototype[Symbol.iterator] = Array.prototype[Symbol.iterator];
Buffer.prototype.values = Array.prototype.values;
Buffer.prototype.keys = Array.prototype.keys;
Buffer.prototype.entries = Array.prototype.entries; |
|
@domenic That makes the array methods polymorphic and probably slows them down. |
|
@domenic thought that too, typed arrays do that, but we obviously want some optimizations. @bnoordhuis added a benchmark and result caching. Results on my machine: Before: After: |
|
@bnoordhuis none of |
|
@timoxley I think |
| var bench = common.createBenchmark(main, { | ||
| size: [16, 512, 1024, 4096, 16386], | ||
| type: ['fast', 'slow'], | ||
| itertation: ['for', 'forOf'] |
There was a problem hiding this comment.
Typo: s/itertation/iteration/ (or maybe just 'method')
|
@vkurchatkin "buffer: iteratior optimization" might want to amend that commit message before the typo is made permanent. Also running your tests I get: Have I forgotten to do something? |
| } | ||
|
|
||
| function benchForOf(buffer) { | ||
| bench.start(); |
vkurchatkin
force-pushed
the
iterable-buffer
branch
2 times, most recently
from
7ef9fe7
to
60cdc32
Compare
| arr = []; | ||
|
|
||
| for (b of buffer) | ||
| arr.push(b); |
vkurchatkin
force-pushed
the
iterable-buffer
branch
from
60cdc32
to
00a04e3
Compare
|
@timoxley my bad! too many fuck ups for a simple commit |
This makes possible to use for..of loop with buffers. Also related keys, values and entries methods are added for feature parity with Uint8Array.
This is basically the code from v8 array iterator implementation minus spec stuff.