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.
Legacy range options have been removed (Level/community#86). If you previously did:
db.iterator({ start: 'a', end: 'z' })An error would now be thrown and you must instead do:
db.iterator({ gte: 'a', lte: 'z' })A db created or opened with this release cannot be opened by earlier versions, because RocksDB has been upgraded to 6.17.3 which effectively changes the default format_version from 2 to 4.
This release also drops support of Node.js 8 (Level/community#98).
This is a rewrite to N-API 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 rocksdb 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 rocksdb is now 8.6.0. Conversely, for node >= 12, rocksdb@4 is the minimum version.
Previously, they were downloaded from GitHub by an npm postinstall script. Prebuilds for 32 bits Windows are no longer included.
RocksDB writes debug information to this log (a file on disk). Should you need this RocksDB feature, it can be enabled with a new infoLogLevel option. Please see README.md for details.
Previously, range options like lt were passed through as-is by abstract-leveldown, unlike keys. For rocksdb 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 rocksdb, 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 rocksdb 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 operations like db.put() complete, 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, rocksdb 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')
})Dropped support for node 4. No other breaking changes.
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.