Tracking issue: planned breaking changes in v2

[vweevers]

In order of significance:

• Breaking: remove callbacks in favor of promises #50
  • Replace use of process.nextTick() with queueMicrotask()
• Breaking: use undefined instead of error for non-existing entries #49
• Add hooks and deprecate batch, put & del events #45
  • Follow-up: Mark hooks as experimental #53
• Remove batch, put and del events (postponed)
• Breaking: remove deferred chained batch #51
• Refactor keys() and values() internals #13
• Breaking: require snapshots to be created synchronously #54
• Support hooks and events on db.clear(). I don't know how yet, but may require breaking changes. (postponed)
• Remove _checkKey() and _checkValue(), an abstract-leveldown leftover that has questionable value because after these methods are called, we run keys and values through encodings. In level-transcoder we should throw if the json encoding returns undefined, as a replacement for undefined checks. Other encodings don't need such checks. (postponed)
• Breaking: remove ready alias of open event #48
• Throw error in default private API methods that must be overridden. (postponed; can be semver-minor)
• Keep track of iterator end #56 (not breaking per se)
• Remove compatibility checks for levelup & friends #52

These won't necessarily land in a single major version. I wanted a place to list them and allow people to object.

[vweevers]

I did a quick test on memory-level, replacing _put(key, value, options, callback) with async _put(key, value, options). As a result, await db.put(key, value) is 2x faster!

$ level-bench plot put
benchmark put on memory-level@1.0.0 win32 x64
node@16.9.1 n=1M concurrency=4 valueSize=100B keys=random values=random

1  memory-level#v2    69565 ops/s ±8.83%  fastest
2  memory-level#main  33750 ops/s ±8.07%  +51.14%

[vweevers]

Removed a bunch of items from the list above, to keep the release manageable (see v2/UPGRADING.md for current content).

[vweevers]

Done; remaining items can go into a future release.