Performance issue with URL.searchParams.append for large numbers of params

Version

21.6.0

Platform

Darwin [snip] 23.2.0 Darwin Kernel Version 23.2.0: Wed Nov 15 21:53:18 PST 2023; root:xnu-10002.61.3~2/RELEASE_ARM64_T6000 arm64

Subsystem

No response

What steps will reproduce the bug?

Using URL.searchParams.append:

time node -e "const url = new url('/sitelet?url=https%3A%2F%2Fexample.com%2F'); console.log(new Date()); Array.from('a'.repeat(50000)).map(a => url.searchParams.append('test', a)); console.log(new Date()); console.log(url.toString().length); console.log(new Date());"

2024-01-19T01:31:47.867Z
2024-01-19T01:34:12.696Z
350020
2024-01-19T01:34:12.697Z
node -e   140.94s user 14.34s system 106% cpu 2:25.17 total

vs. using URLSearchParams.append:

time node -e "const params = new URLSearchParams(); console.log(new Date()); Array.from('a'.repeat(50000)).map(a => params.append('test', a)); console.log(new Date()); console.log(('https://example.com/?' + params.toString()).length); console.log(new Date());"

2024-01-19T01:32:25.111Z
2024-01-19T01:32:25.120Z
350020
2024-01-19T01:32:25.127Z
node -e   0.04s user 0.01s system 79% cpu 0.066 total

How often does it reproduce? Is there a required condition?

No response

What is the expected behavior? Why is that the expected behavior?

No response

What do you see instead?

The URL.searchParams.append snippet with 50,000 params being append takes 2-3 minutes to run on an M1, which is a ridiculous amount of time. The equivalent snippet with URLSearchParams.append takes 50-100 milliseconds to run.

Additional information

I talked through this performance issue with Yagiz on Twitter and got to what appears to be the cause of the issue, every time URL.searchParams.append is called, the whole URL essentially gets stringified to update properties on URL itself: https://twitter.com/yagiznizipli/status/1748151781534650545

I believe it to be this call that is the cause:

node/lib/internal/url.js

Lines 478 to 480 in eb4432c

    
           if (this.#context) { 
        
             this.#context.search = this.toString(); 
        
           }

With #context being set here:

node/lib/internal/url.js

Lines 1023 to 1024 in eb4432c

    
           this.#searchParams = new URLSearchParams(this.search); 
        
           setURLSearchParamsContext(this.#searchParams, this);

And the #context.search setter being here:

node/lib/internal/url.js

Lines 1012 to 1017 in eb4432c

    
           set search(value) { 
        
             const href = bindingUrl.update(this.#context.href, updateActions.kSearch, StringPrototypeToWellFormed(`${value}`)); 
        
             if (href) { 
        
               this.#updateContext(href); 
        
             } 
        
           }

I'm unsure whether the performance issue is with the toString call in URLSearchParams, or whether it is within the logic of the #context.search setter. I've not worked with Node.js source before, so not sure where to start on profiling that, or submitting a fix (hence opening this issue to track it for anyone to grab).

I suspect that an "easy" fix either way is to update URLSearchParams such that it doesn't write back to URL, and instead update the relevant bits of URL (href, search, etc.) to essentially call the logic that is currently in the #context.search setter on-demand as getters when needed, if URLSearchParams has been created for the URL.

(unrelated to Node.js, but bonus points to anyone that feels like surfacing a similar fix with browser engines, as they seem to suffer from the same bottleneck)

As I understand it currently, URL stores its entire value in #context.href as a single string (and associated pointers for search etc.). When instantiated, URL does not have an associated URLSearchParams. If, and only if, a user accesses searchParams on URL, then a URLSearchParams is instantiated for the URL. However, URL still relies on its own #context.href value always, and so the data in URLSearchParams is essentially a duplicate, and any update to it needs to call the search setter on URL so that #context.href can be updated properly.

Thinking through this, I think the approach to fixing this might be:

Remove the URL #context from URLSearchParams, so it is a properly independent data store
Update the search + href getters in URL to interface with searchParams if it exists
Update the search + href setters in URL to interface with searchParams if it exists, and not pass any search to the internal #context.href, so it is just stored in searchParams (or, rather, update #updateContext to slice the returned href and not store the search index when searchParams exists)
Update searchParams so that when it instantiates URLSeachParams, it also removes any search from the internal #context.href, so it is just stored in searchParams (probably via the same #updateContext modification)

This does mean that the internal #context.href value of URL will not contain the search string once searchParams has been accessed, but I think that is probably the best thing to do so there is no risk of future confusion with the state of that being potentially out of sync with searchParams.

@anonrig

@anonrig @nodejs/performance

url: remove #context from URLSearchParams

url: don't update URL immediately on update to URLSearchParams

Navigation Menu

Repository navigation

Performance issue with URL.searchParams.append for large numbers of params #51518

Version

Platform

Subsystem

What steps will reproduce the bug?

How often does it reproduce? Is there a required condition?

What is the expected behavior? Why is that the expected behavior?

What do you see instead?

Additional information

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	if (this.#context) {
	this.#context.search = this.toString();
	}

	this.#searchParams = new URLSearchParams(this.search);
	setURLSearchParamsContext(this.#searchParams, this);

	set search(value) {
	const href = bindingUrl.update(this.#context.href, updateActions.kSearch, StringPrototypeToWellFormed(`${value}`));
	if (href) {
	this.#updateContext(href);
	}
	}

Performance issue with URL.searchParams.append for large numbers of params #51518

Description

Version

Platform

Subsystem

What steps will reproduce the bug?

How often does it reproduce? Is there a required condition?

What is the expected behavior? Why is that the expected behavior?

What do you see instead?

Additional information

Activity

MattIPv4 commented on Jan 19, 2024

MattIPv4 commented on Jan 19, 2024

benjamingr commented on Jan 21, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions