♻️ improves merge proxies

[ekwoka]

This PR refactors mergeProxies to increase performance and reduce memory costs.

• 🐛 Allows JSON stringifying $data (solving an issue presented in discord)
• ♻️ Uses Reflect (to reduce code and memory)
• ⚡ Shortcircuits on unscopables (to increase performance on every access of data in expressions)
• ♻️ Extract proxyTraps (to reduce memory)
• ✅ Uses legacy syntax (to pass tests)

[calebporzio]

Thank you for this @ekwoka - do you have any benchmarks to go with it to more concretely demonstrate the benefits? Thanks!

[ekwoka]

@calebporzio Sure thing. Here's some tests that basically just compares both mergeProxies methods accessing at different depths in some different run times.

So naturally old and new is self explanatory, then the first number is the number of repititions, the second is the depth of the test.

The depth is used to generate a proxy out of that many objects, and similarly, in the accessing, it accesses a property on each of those objects within a with statement.

So its 100k loops on a depth of 26, and 1M loops on a depth of 10.

You can see the new one is basically always about 30% faster than the old in these test configuration So every time a property is accessed in an expression, this could make that up to 30% faster (deeper component trees will benefit more, smaller trees it's much more modest (around 3-5%).

For memory, this script just puts that number of merged proxies into an array so we can actually measure the memory size (otherwise it's basically impossible to really tell) and the time in ms. So less memory can also have a performance impact (less garbage to collect).

About 75% less memory per merged proxy (runtime memory was about 3mb so not meaningful to these measurements)

[calebporzio]

Great, thank you for the extra context!