Example use cases for various options

[littledan]

I'm glad to see that the README provides various use cases in the JS ecosystem--this does a good job showing that there's a strong chance that this feature will improve application performance. For the various options, it's less clear to me where they are needed. How much of the time would the defaults just work? It'd be good have example use cases for various option combinations broken out.

[BridgeAR]

That is a very valid point!
I know of use cases in Node.js for propably all permutations. I'll add those. If someoen else could support and look for some in the wider ecosystem, that would be great!

[BridgeAR]

I just add a few examples here until I open a PR

https://github.com/nodejs/node/blob/1540fc6a37e42fe002d3061a516431741e8e6d3d/lib/internal/util/comparisons.js#L221-L225
This uses a C++ method to get the enumerable nonIndexString properties of an array and in one mode also the symbol properties. It depends on the algorithm mode. It could just get the count instead. That would hopefully be even a tad faster for the default case, while it would absolutely allow the algorithm outside of Node.js to use a fast path. That is important for a browser implementation.

https://github.com/nodejs/node/blob/1540fc6a37e42fe002d3061a516431741e8e6d3d/lib/internal/util/comparisons.js#L385
This gets the symbol properties. The algorithm only needs enumerable symbol properties, so if there are any, they are filtered afterwards. As a fast path, this could use enumerable symbols.

https://github.com/nodejs/node/blob/1540fc6a37e42fe002d3061a516431741e8e6d3d/lib/internal/util/inspect.js#L755-L777
This gets the symbol properties and that is not necessary in case showHidden is false and there are no enumerable ones (which is the default when being called).

https://github.com/nodejs/node/blob/1540fc6a37e42fe002d3061a516431741e8e6d3d/lib/internal/util/comparisons.js#L915-L924
This detects if an array is sparse or not by checking for the entries being own properties or not. It would be possible to detect a sparse array immediately by checking for the index properties on the array. If they match the length, it is a dense array.

https://github.com/nodejs/node/blob/1540fc6a37e42fe002d3061a516431741e8e6d3d/lib/internal/util/inspect.js#L1752-L1756
This detects if a key on a sparse array is actually an index key or not. That could be skipped if the length of Object.keys is identical to the count of all index properties.

These are examples for:

• type: ['index'], enumerable: true
• type: ['symbol'], enumerable: true
• type: ['nonIndexString', 'symbol'], enumerable: true

[bakkot]

FWIW I checked a couple local codebases and found ~ a dozen uses of Object.keys().length and no uses of counting properties otherwise (that I could easily find).

[BridgeAR]

React

https://github.com/facebook/react/blob/336614679600af371b06371c0fbdd31fd9838231/packages/react-devtools-shared/src/utils.js#L109-L119
The symbols and descriptor calls may be skipped for

• type: ['symbol'], enumerable: true

https://github.com/facebook/react/blob/336614679600af371b06371c0fbdd31fd9838231/packages/react-server/src/ReactFlightServer.js#L2787-L2788

https://github.com/facebook/react/blob/336614679600af371b06371c0fbdd31fd9838231/packages/react-refresh/src/ReactFreshRuntime.js#L669-L670
Counts the entries

• type: ['index', 'nonIndexString'], enumerable: 'all'

https://github.com/facebook/react/blob/336614679600af371b06371c0fbdd31fd9838231/packages/react-dom-bindings/src/client/ReactFiberConfigDOM.js#L4425

Next.js

https://github.com/vercel/next.js/blob/82ab39f8017e2ba1aef668269e5b68cdea94f335/packages/next/src/compiled/%40babel/runtime/helpers/objectSpread.js#L7-L9

https://github.com/vercel/next.js/blob/82ab39f8017e2ba1aef668269e5b68cdea94f335/packages/next/src/compiled/react-server-dom-webpack/cjs/react-server-dom-webpack-client.node.development.js#L685-L686

vscode

https://github.com/microsoft/vscode/blob/a32a31b6ea1e1bc92504879d7ce96ad53e43ef2c/src/vs/workbench/contrib/debug/common/replModel.ts#L133

https://github.com/microsoft/vscode/blob/a32a31b6ea1e1bc92504879d7ce96ad53e43ef2c/src/vs/workbench/contrib/notebook/common/model/notebookTextModel.ts#L1066-L1069

TypeScript

https://github.com/microsoft/TypeScript/blob/83dc0bb2ed91fe0815ab28dc3ff95fae7425e413/src/compiler/factory/emitHelpers.ts#L905-L908

Angular

https://github.com/angular/angular/blob/6fa8d441979fdbabb88dddd246f54587e17126e8/packages/zone.js/lib/browser/define-property.ts#L48-L60

[BridgeAR]

@bakkot the use cases are often not just directly using a length check in those cases. It is possible to improve algorithms by adding a fast path for the default cases in many situations. E.g., checking for enumerable symbols.