chore(): cache cleanup

[ShaMan123]

Proposed Changes

Related to #26, preliminary work
Closes #25

Indexing paths means the index needs to hold the path in memory, making caching redundant.
So I wish to unblock myself before moving forward.

I have encapsulated all caching logic into a single method getPathCurve.
This is a main entry point that will not be used by the indexing method.
This is where path parsing is handled as well.

I don't really understand why there are 3 levels of caching (curve, abs, arr) and I find the caching engine a bit weird (does it rm an entry after 100 invocations?) but that is off topic and will be made redundant once indexing is used.

Additionally I took the opportunity to inspect pathClone usages. They were called excessively so I have removed those. The place that mutates path data is pathToCurve.
Then I looked into the implementation and decided to simplify it as well (for perf, readability and code quality).

I exposed parsePathCurve so a consumer can opt out of internal path caching.

Finally, I've extracted isBBoxIntersect early return up to its caller since the spatial index will be handling bbox intersection logic.

Checklist

To ensure you provided everything we need to look at your PR:

• Brief textual description of the changes present
• Visual demo attached
• Steps to try out present, i.e. using the @bpmn-io/sr tool
• Related issue linked via Closes {LINK_TO_ISSUE} or Related to {LINK_TO_ISSUE}

[CLAassistant]

All committers have signed the CLA.

[ShaMan123]

stable and ready for review

[ShaMan123]

I don't understand this...

Suggested change

	(!isArray(path) || !isArray(path && path[0]))) ?
	typeof path === 'string' ?

[ShaMan123]

@nikku @barmac can I commit this?

[barmac]

The current CI failure is not related to your PR. Let me fix that...

[barmac]

The current CI failure is not related to your PR. Let me fix that...

Problem will be fixed via #28
You can rebase your work on main after the PR is merged.

[ShaMan123]

add prettierrc to make my life easier but still there is whitespace mayham

[nikku]

add prettierrc to make my life easier but still there is whitespace mayham

Please do not do that. Keep changes minimal, easy to follow.

[ShaMan123]

I believe I kept changes minimal as possible.
I can revert prettierrc but I added it because I didn't want to commit whitespace changes.
If prettier is installed in vscode it takes over...
Nvm

[ShaMan123]

I expected lint to fix whitespace but it doesn't

[ShaMan123]

expose parsePathCurve

[nikku]

@ShaMan123 CI is failing. Please ensure npm run all passes locally (on your machine).

[nikku]

You could add a test case to verify caching (intersecting between two cached paths should be faster than intersecting between two uncached paths).

[ShaMan123]

You could add a test case to verify caching (intersecting between two cached paths should be faster than intersecting between two uncached paths).

Since the cache is not exposed I have no means of resetting it, right?

I am not sure if the test is reliable enough, but it is added

[ShaMan123]

Fixed failing TS and added a test to verify findIntersections does not mutate a path passed to it

[ShaMan123]

added a cache timing test but I do not recommend keeping it. It is flaky and IMO gains nothing.

[ShaMan123]

Revisited the caching perf test to make it more robust and found out that for 100/1000 path segments the difference in runtime in 3ms. For 2 segments caching was degrading performance.
All in all seems that for most cases the code's caching is degrading or not boosting enough to make it worthwhile.
With this PR a consumer can parse and cache paths so I would remove caching in the next major release.