This document describes breaking changes and how to upgrade. For a complete list of changes including minor and patch releases, please refer to the changelog.
This is a rewrite to N-API - which is a huge milestone, achieved without an impact on write performance - and an upgrade to abstract-leveldown v6, which solves long-standing issues around serialization and type support.
N-API is the latest interface for native addons in Node.js. Main benefit is that leveldown became independent of the V8 version, which means it will be compatible with future versions of Node.js. As long as the native code doesn't change, it doesn't need to be recompiled as new versions of Node.js are released. This helps greatly with both maintenance and when delivering code to consumers.
Because N-API has an experimental status in node 6 and early 8.x releases, support of node 6 has been dropped and the minimum node version for leveldown is now 8.6.0. Conversely, for node >= 12, leveldown@5 is the minimum version.
Previously, they were downloaded from GitHub by an npm postinstall script. In addition, leveldown now includes prebuilt binaries for Linux ARMv7, Linux ARM64, Android ARMv7 and Android ARM64. The latter target node core (rather than the formerly needed nodejs-mobile fork).
Previously, range options like lt were passed through as-is by abstract-leveldown, unlike keys. For leveldown it means that range option types other than a string or Buffer will be stringified.
Because null, undefined, zero-length strings and zero-length buffers are significant types in encodings like bytewise and charwise, they became valid as range options in abstract-leveldown. This means db.iterator({ gt: undefined }) is not the same as db.iterator({}).
In the case of leveldown, when used by itself, the aforementioned change means that db.iterator({ gt: undefined }) is now effectively the same as db.iterator({ gt: 'undefined' }), making it explicit that leveldown only supports strings and buffers.
As a result of this, the target argument in iterator.seek(target) is now serialized. Meaning any type other than a string or Buffer will be stringified. Like before, if the result is a zero-length string or Buffer, an error will be thrown.
In addition to rejecting null and undefined as keys, abstract-leveldown now also rejects these types as values, due to preexisting significance in streams and iterators.
Though this was already the case, abstract-leveldown has replaced the behavior with an explicit Array.isArray() check and a new error message.
The sync option has moved to chainedBatch.write(options). Previously, sync was half-documented and half-implemented.
It is now safe to call db.close() before db.put() completes, to call db.iterator() on a non-open db and to call db.close() having created many iterators regardless of their state (idle, nexting, ending or ended). To achieve this, leveldown waits for pending operations before closing:
db.put('key', 'value', function (err) {
console.log('this happens first')
})
db.close(function (err) {
console.log('this happens second')
})A future release will do the same for other operations like get and batch.
Dropped support for node 4. No other breaking changes.
If you're using node v10 you'll need at least leveldown@2.0.1 to successfully compile. In addition, if you want prebuilt binaries you'll need at least leveldown@3.0.1.
This major release contains an upgrade to abstract-leveldown with a breaking change for the array version of .batch(). This change ensures all elements in the batch array are objects.
If you previously passed arrays to .batch() that contained undefined or null, they would be silently ignored. Now this will produce an error.