path: refactor for performance and consistency by nwoltman · Pull Request #1778 · nodejs/node

nwoltman · 2015-05-23T04:50:07Z

Convergence PR prompted by #35
Original PR: nodejs/node-v0.x-archive#9289

Original PR message (slightly modified):

Improve performance by:

Not leaking the arguments object in win32.join
Getting the last character of a string by index, instead of with .substr() or .slice()

Improve code consistency by:

Using [] instead of .charAt() where possible
Using a function declaration instead of a var declaration
Using .slice() with clearer arguments

Improve both by:

Making the reusable trimArray() function
Standardizing getting certain path statistics with the new win32StatPath() function

Benchmarks: (higher is better)

function (path.)	current (ops/sec)	this PR (ops/sec)
win32.resolve	315,019	399,543
win32.normalize	873,559	1,051,170
win32.isAbsolute	1,915,332	1,939,849
win32.join	452,879	815,033
win32.relative	153,821	219,928
win32.format	13,118,588	16,364,384
posix.relative	244,638	304,456

Benchmark code (gist)

mscdex · 2015-05-23T04:54:18Z

lib/path.js

First letter should be capitalized.

silverwind · 2015-05-23T05:54:42Z

Are there any API changes? (seeing that the tests were modified)

nwoltman · 2015-05-23T06:34:11Z

@silverwind I'm not sure if this counts as an API change, but as discussed here, if a user passes in an object where the dir property is null or undefined, these changes make it so that the function returns the correct value instead of incorrectly throwing.

jbergstroem · 2015-05-23T06:47:48Z

I'm keen on getting the benchmark in (benchmarks/fs/) as well – but we haven't started using benchmark.js just quite yet. If you're keen, rewriting it slightly would be nice.

nwoltman · 2015-05-23T17:06:59Z

@jbergstroem It looks like each path method would require it's own benchmark file. That's going to be a fair number of files so perhaps they should go in a new benchmark/path directory?
Then there's the question of whether or not the path.win32 and path.posix methods should be in their own files, or if they should be compared in the same file (similar to the Buffer/SlowBuffer comparison).

ChALkeR · 2015-05-23T17:23:03Z

Looks like #1752 conflicts with this. I will remove the path.win32 part from #1752 once this gets merged.

Btw,

In path.win32.join this PR changes the stack of the TypeError('Arguments to path.join must be strings'), removing the top at f (path.js:190:13) and at Object.filter (native) so now it is similar to the stack in path.posix.join.

Is also valid for this PR.

jasnell · 2015-05-27T18:01:35Z

LGTM

ChALkeR · 2015-05-27T18:39:31Z

Why using slice instead of substr/substring for strings?

nwoltman · 2015-05-27T22:38:16Z

@ChALkeR I haven't changed any code to use slice instead of substr/substring. If you're specifically referring to this line, yes I could have changed slice to substr, but instead I changed the input value to make the code cleaner and easier to read.

nwoltman · 2015-06-06T00:50:09Z

@jbergstroem I added some benchmarks. Do those look about right?

jbergstroem · 2015-06-13T03:56:02Z

@woollybogger apologies for the late reply. LGTM. CI here: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/827/

nwoltman · 2015-06-16T20:19:32Z

@jbergstroem A couple unrelated (and possibly flaky) tests failed. Does that matter?

jbergstroem · 2015-06-16T21:32:21Z

@woollybogger no, that's something we're working on improving. I can't access the old results -- can you rebase on top of latest master so we can get a new run going?

You might want to see if anyone else wants to review before I/someone else look at landing it. @ChALkeR?

nwoltman · 2015-06-16T21:48:30Z

Rebased

jbergstroem · 2015-06-16T22:00:42Z

CI here: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/

jbergstroem · 2015-06-16T22:45:14Z

(never mind the windows timeouts)

ChALkeR · 2015-06-17T06:09:12Z

The main part (lib/path.js) LGTM.

Minor thing: you replace single-char .substr() and .charAt() with [] in all places of path module, but this one:

posix.isAbsolute = function(path) {
  assertPath(path);
  return path.charAt(0) === '/';
};

Could you fix that for consistency?

nwoltman · 2015-06-17T08:11:17Z

@ChALkeR Fixed.

thefourtheye · 2015-06-17T08:19:52Z

lib/path.js

I am not sure if returning the original array and a new array based on conditions is a good idea.

It's just a little helper function. There's no point in making a copy of an array when the original is all you need.

nwoltman · 2015-06-26T04:51:19Z

bump 🚅

silverwind · 2015-07-03T08:22:07Z

lib/path.js

Could use shorthand properties here:

return { device, isUnc, isAbsolute: isUnc || !!result[2], // UNC paths are always absolute tail: result[3] };

silverwind · 2015-07-03T08:43:15Z

LGTM, few nits and a question above.

Improve performance by: + Not leaking the `arguments` object! + Getting the last character of a string by index, instead of with `.substr()` or `.slice()` Improve code consistency by: + Using `[]` instead of `.charAt()` where possible + Using a function declaration instead of a var declaration + Using `.slice()` with clearer arguments + Checking if `dir` is truthy in `win32.format` (added tests for this) Improve both by: + Making the reusable `trimArray()` function + Standardizing getting certain path statistics with the new `win32StatPath()` function

Path functions being benchmarked are: * format * isAbsolute * join * normalize * relative * resolve

nwoltman · 2015-07-03T17:02:54Z

Updated

silverwind · 2015-07-03T20:11:51Z

LGTM

Improve performance by: + Not leaking the `arguments` object! + Getting the last character of a string by index, instead of with `.substr()` or `.slice()` Improve code consistency by: + Using `[]` instead of `.charAt()` where possible + Using a function declaration instead of a var declaration + Using `.slice()` with clearer arguments + Checking if `dir` is truthy in `win32.format` (added tests for this) Improve both by: + Making the reusable `trimArray()` function + Standardizing getting certain path statistics with the new `win32StatPath()` function PR-URL: #1778 Reviewed-By: Сковорода Никита Андреевич <chalkerx@gmail.com> Reviewed-By: Roman Reiss <me@silverwind.io>

Path functions being benchmarked are: * format * isAbsolute * join * normalize * relative * resolve PR-URL: #1778 Reviewed-By: Сковорода Никита Андреевич <chalkerx@gmail.com> Reviewed-By: Roman Reiss <me@silverwind.io>

silverwind · 2015-07-04T11:31:21Z

Landed in bca53dc and 0d15161

Improve performance by: + Not leaking the `arguments` object! + Getting the last character of a string by index, instead of with `.substr()` or `.slice()` Improve code consistency by: + Using `[]` instead of `.charAt()` where possible + Using a function declaration instead of a var declaration + Using `.slice()` with clearer arguments + Checking if `dir` is truthy in `win32.format` (added tests for this) Improve both by: + Making the reusable `trimArray()` function + Standardizing getting certain path statistics with the new `win32StatPath()` function PR-URL: nodejs#1778 Reviewed-By: Сковорода Никита Андреевич <chalkerx@gmail.com> Reviewed-By: Roman Reiss <me@silverwind.io>

Path functions being benchmarked are: * format * isAbsolute * join * normalize * relative * resolve PR-URL: nodejs#1778 Reviewed-By: Сковорода Никита Андреевич <chalkerx@gmail.com> Reviewed-By: Roman Reiss <me@silverwind.io>

mscdex added the path Issues and PRs related to the path subsystem. label May 23, 2015

mscdex reviewed May 23, 2015
View reviewed changes

lib/path.js Outdated

Copy link

Contributor

mscdex May 23, 2015

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

First letter should be capitalized.

jasnell mentioned this pull request May 27, 2015

[Converge] path: refactor for performance and consistency nodejs/node-convergence-archive#35

Closed

bnoordhuis force-pushed the master branch from 311c3e4 to b926718 Compare June 2, 2015 17:25

thefourtheye reviewed Jun 17, 2015
View reviewed changes

rvagg force-pushed the master branch from 9d21135 to 628a3ab Compare June 25, 2015 09:18

brendanashworth added the benchmark Issues and PRs related to the benchmark subsystem. label Jun 26, 2015

silverwind reviewed Jul 3, 2015
View reviewed changes

benchmark: Add some path benchmarks for #1778

363335c

Path functions being benchmarked are: * format * isAbsolute * join * normalize * relative * resolve

silverwind closed this Jul 4, 2015

Fishrock123 mentioned this pull request Jul 6, 2015

Release proposal: v2.3.4 #2115

Merged

nwoltman deleted the path branch July 8, 2015 03:36

snyk-bot mentioned this pull request Apr 2, 2021

[Snyk] Security upgrade snyk from 1.437.3 to 1.518.0 turkdevops/node#241

Closed

snyk-bot mentioned this pull request Aug 24, 2021

[Snyk] Security upgrade snyk from 1.437.3 to 1.518.0 turkdevops/node#317

Closed

Uh oh!

Conversation

nwoltman commented May 23, 2015

Benchmarks: (higher is better)

Benchmark code (gist)

Uh oh!

mscdex May 23, 2015

Choose a reason for hiding this comment

Uh oh!

silverwind commented May 23, 2015

Uh oh!

nwoltman commented May 23, 2015

Uh oh!

jbergstroem commented May 23, 2015

Uh oh!

nwoltman commented May 23, 2015

Uh oh!

ChALkeR commented May 23, 2015

Uh oh!

jasnell commented May 27, 2015

Uh oh!

ChALkeR commented May 27, 2015

Uh oh!

nwoltman commented May 27, 2015

Uh oh!

nwoltman commented Jun 6, 2015

Uh oh!

jbergstroem commented Jun 13, 2015

Uh oh!

nwoltman commented Jun 16, 2015

Uh oh!

jbergstroem commented Jun 16, 2015

Uh oh!

nwoltman commented Jun 16, 2015

Uh oh!

jbergstroem commented Jun 16, 2015

Uh oh!

jbergstroem commented Jun 16, 2015

Uh oh!

ChALkeR commented Jun 17, 2015

Uh oh!

nwoltman commented Jun 17, 2015

Uh oh!

thefourtheye Jun 17, 2015

Choose a reason for hiding this comment

Uh oh!

nwoltman Jun 17, 2015

Choose a reason for hiding this comment

Uh oh!

nwoltman commented Jun 26, 2015

Uh oh!

silverwind Jul 3, 2015

Choose a reason for hiding this comment

Uh oh!

silverwind commented Jul 3, 2015

Uh oh!

nwoltman commented Jul 3, 2015

Uh oh!

silverwind commented Jul 3, 2015

Uh oh!

silverwind commented Jul 4, 2015

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants