Improve performance of addLayers by preallocating array

[pmarrapese]

Presently, addLayers pushes a new element onto MarkerClusterGroup._needsClustering for each qualifying element in layersArray. This is inefficient because the array is being resized upon every iteration, which is a very expensive operation. This problem really starts to become evident when adding a large number of markers.

This pull request resolves this problem by preallocating the array instead of repeatedly resizing it. Using a dataset of 120,866 markers, this simple fix reduced my average addLayers call time from 8.9s to 0.5s.

[goapunk]

thanks for maintaining this project! Is there any chance someone could take a look at this?

[pmarrapese]

Made a minor but important update to this: this._needsClustering.push was resulting in a "Maximum call stack size exceeded" exception on some mobile browsers (Safari on iPhone) when working with very large sets of markers. this._needsClustering.concat prevents this from occurring.

[nconnector]

I was half-way into building a cusom static clustered grid to solve this.

6.5s --> 1.2s with this solution

@pmarrapese this is fantastic, please pull this :)

[ykzeng]

Hi @pmarrapese , this seems like an important perf improvement and def would like to merge it. any minimal examples that could reproduce the efficiency improvements on jsfiddle?

[pmarrapese]

@ykzeng - Sorry for the wait here! Finally had some time to do a demo. Open the console for logging.
Tested on Chrome 114.0.5735.199 / Windows 10 (64-bit).

My changes; using leaflet.markercluster-src.js built from pmarrapese@f43d89c

• URL: https://jsfiddle.net/pmarrapese/vh8zc0t7/
• Adding 205,705 layers takes ~0.13s

Without my changes; leaflet.markercluster-src.js built from bd2794c

• URL: https://jsfiddle.net/pmarrapese/tyo3hbnf/
• Adding 205,705 layers takes ~21.15s

[MariaAbadi]

@IvanSanchez @ykzeng please either start reviewing the PRs (this one is 4 years old!) or add more maintainers to the project with merge rights.