Optimize and benchmark circular references

[trevorr]

I added a test for cloning circular references, and it performed very poorly:

$ node benchmark/circles
benchLodashCloneDeep*100: 878.215ms
benchFastCopy*100: 338.467ms
benchNanoCopy*100: 330.19ms
benchRamdaClone*100: 1.878s
benchRfdcCircles*100: 1:03.309 (m:ss.mmm)
benchRfdcCirclesProto*100: 1:03.243 (m:ss.mmm)
benchStructuredClone*100: 862.133ms

I figured the performance was likely due to using Array.indexOf() to handle looking up circular references. I used a Map instead, and the performance improved dramatically:

node benchmark/circles
benchLodashCloneDeep*100: 968.515ms
benchFastCopy*100: 351.399ms
benchNanoCopy*100: 334.315ms
benchRamdaClone*100: 1.945s
benchRfdcCircles*100: 452.196ms // 139x faster
benchRfdcCirclesProto*100: 455.135ms // 139x faster
benchStructuredClone*100: 899.574ms

On the original benchmark (no circular references), there is only a slight degradation on performance when using circles: true. Here's the original code:

node benchmark
benchDeepCopy*100: 360.976ms
benchLodashCloneDeep*100: 855.236ms
benchCloneDeep*100: 498.917ms
benchFastCopy*100: 429.037ms
benchFastestJsonCopy*100: 225.851ms
benchPlainObjectClone*100: 339.774ms
benchNanoCopy*100: 454.609ms
benchRamdaClone*100: 1.368s
benchJsonParseJsonStringify*100: 1.025s
benchRfdc*100: 213.324ms
benchRfdcProto*100: 223.08ms
benchRfdcCircles*100: 237.251ms
benchRfdcCirclesProto*100: 248.533ms
benchStructuredClone*100: 1.164s

And the code from this branch:

node benchmark
benchDeepCopy*100: 362.713ms
benchLodashCloneDeep*100: 845.653ms
benchCloneDeep*100: 495.222ms
benchFastCopy*100: 433.719ms
benchFastestJsonCopy*100: 233.539ms
benchPlainObjectClone*100: 303.27ms
benchNanoCopy*100: 454.411ms
benchRamdaClone*100: 1.369s
benchJsonParseJsonStringify*100: 1.030s
benchRfdc*100: 209.843ms
benchRfdcProto*100: 221.067ms
benchRfdcCircles*100: 309.416ms // 30% slower
benchRfdcCirclesProto*100: 318.529ms // 28% slower
benchStructuredClone*100: 1.143s

Overall, still very competitive, and no longer unusable when many cycles are encountered.

[mcollina]

lgtm

[mcollina]

CI is failing, couyld you fix?

[trevorr]

I don't think it has anything to do with my changes. I get the same error on master when I run npm run ci:

==> Uploading reports
/Users/trevor/git/rfdc/node_modules/codecov/lib/codecov.js:212
      } else if (result.split('\n').length !== 2) {
                        ^

TypeError: result.split is not a function
    at /Users/trevor/git/rfdc/node_modules/codecov/lib/codecov.js:212:25
    at /Users/trevor/git/rfdc/node_modules/teeny-request/build/src/index.js:217:17
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)

Node.js v20.11.1

[trevorr]

When I run it locally, result is an object rather than a string: { message: 'Repository not found' }.

[mcollina]

please rebase, I've removed codecov

[mcollina]

still lgtm

[mcollina]

for some odd reasons this is crashing on node v22. Can you take a look?

[trevorr]

This also fails on master. It looks like an internal node assert. Perhaps the test dependencies don't support Node 22? When I upgrade tap from 12 to 18+, Node no longer asserts, but the tests fail with messages like "isNot is not a function".