-
Notifications
You must be signed in to change notification settings - Fork 29.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
buffer: inspect extra properties #25150
Conversation
FWIW Here's what I came up with for the JS implementation: function isAllDigits(s) {
if (s.length === 0)
return false;
for (var i = 0; i < s.length; ++i) {
const code = s.charCodeAt(i);
if (code < 48 || code > 57)
return false;
}
return true;
}
function getOwnNonIndexProperties(obj, filter) {
const props = [];
var p = 0;
if (filter === ONLY_ENUMERABLE) {
const keys = Object.keys(obj);
for (var i = 0; i < keys.length; ++i) {
const key = keys[i];
if (!isAllDigits(key))
props[p++] = key;
}
} else {
// TODO or defer to C++ implementation
}
return props;
} It's at least twice as fast for a variety of objects using node master. |
@mscdex how would that look like? I can't think of any way to access the keys like that in JS. The functions returns the keys of an object that are not numbers (keys are sorted by spec. First the numbers, then strings, then symbols). |
@BridgeAR I updated my comment with the implementation I benchmarked with. |
@mscdex that will be very slow for objects with lots of properties. The reason is that the array with all keys has to be allocated first and that would be very expensive. |
@BridgeAR How many properties are we talking about here? |
@mscdex in my direct comparison I can't get the JS function to be faster under any circumstances. The C++ function has a constant time while the JS function is over linear and even for small buffers, the C++ function wins when running it multiple times. |
@BridgeAR FWIW here is the complete benchmark code I used (comment out the case not being tested): Benchmark codeconst n = 1e6;
const vals = [
{ foo: 'bar', baz: 5, quux: true },
{ '1': 'bar', '2': 5, '3': 5.1 },
{ '-1': 'bar', '2': 5, '3a': 40000 },
{ '': () => {}, '1234567890-0987654321': false, quux: false }
];
const {
propertyFilter: {
ALL_PROPERTIES,
ONLY_ENUMERABLE
}
} = process.binding('util');
const filter = ONLY_ENUMERABLE;
var name;
// Case #1
//~ const {
//~ getOwnNonIndexProperties,
//~ } = process.binding('util');
//~ name = 'C++';
// Case #2
function isAllDigits(s) {
if (s.length === 0)
return false;
for (var i = 0; i < s.length; ++i) {
const code = s.charCodeAt(i);
if (code < 48 || code > 57)
return false;
}
return true;
}
function getOwnNonIndexProperties(obj, filter) {
const props = [];
var p = 0;
if (filter === ONLY_ENUMERABLE) {
const keys = Object.keys(obj);
for (var i = 0; i < keys.length; ++i) {
const key = keys[i];
if (!isAllDigits(key))
props[p++] = key;
}
} else {
}
return props;
}
name = 'JS';
// Run the benchmark
console.time(name);
for (var i = 0; i < n; ++i) {
for (var j = 0; j < vals.length; ++j)
getOwnNonIndexProperties(vals[j], filter);
}
console.timeEnd(name); Running this I get ~1100ms for C++ and ~510ms for JS on node master. EDIT: For mostly numeric keys the C++ implementation does seem to be a little faster, but not by a whole lot (~50ms), when using objects with the same number of properties as in the cases provided above. |
@mscdex this function is used for array types (in this case Buffer / Uint8Array), not for regular objects. |
CI https://ci.nodejs.org/job/node-test-pull-request/19763/ @nodejs/buffer @nodejs/util PTAL |
Resume Build CI: https://ci.nodejs.org/job/node-test-pull-request/19767/ ✔️ |
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer.
60a5f3d
to
1be4412
Compare
Rebased due to conflicts. CI https://ci.nodejs.org/job/node-test-pull-request/19787/ ✔️ |
@nodejs/buffer @nodejs/util PTAL. This needs another review. |
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer. PR-URL: nodejs#25150 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com>
Landed in d385e2c 🎉 |
Should this be backported to |
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer. PR-URL: nodejs#25150 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com>
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer. PR-URL: #25150 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com>
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer. PR-URL: #25150 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com>
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer. PR-URL: nodejs#25150 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com>
This makes sure extra properties on buffers are not ignored anymore when inspecting the buffer. PR-URL: nodejs#25150 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com>
This makes sure extra properties on buffers are not ignored anymore
when inspecting the buffer.
I am not sure about this either being a patch or semver-major. We mainly consider changes to
util.inspect
as patch, so I guess this could be considered a similar thing.The implementation uses some internal knowledge and it would require more overhead otherwise but this seemed the most straight forward way to do this.
Checklist
make -j4 test
(UNIX), orvcbuild test
(Windows) passes