You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
codec for js values (including arrays) that preserves lexiographic
sort order when encoded. (the order is compatible with bytewise and thus indexeddb and couchdb,
but the encoded format is different)
for building indexes on top of leveldb, bytewise is great!
it lets you structure you keys and reason about how they
will be ordered in a very simple and reliable way.
But bytewise is too slow! it's slow enough to have quite visible
effects on a bulk load on a reasonable sized database with a couple
of indexes.
(i.e. 100k secure-scuttlebutt messages with indexes, measured by bench-ssb)
stability: experimental
Expect breaking changes to encoded format. We are still making
breaking changes if necessary to improve performance.
(although, codec api is fully stable and will not change)
simple benchmark
run a simple benchmark for one second, encoding & decoding ops
in one second.
It was easy to make charwise faster than bytewise when
it was only a partial implementation, but once correct escaping
and nested arrays where added it got slow.
But then @PaulBlanchehad the genious idea
of encoding items in an array with their depth inside the array.
This supports deeply nested arrays or shallowly nested arrays
with only one pass escaping the items. This made encoding much faster
again!
Examples
constlevel=require('level')constcharwise=require('charwise')constdb=level('./db8',{keyEncoding: charwise})awaitdb.batch([{type: 'put',key: ['users',2],value: 'example'},{type: 'put',key: ['users',10],value: 'example2'}])constuserStream=db.createStream({gte: ['users',charwise.LO],lte: ['users',charwise.HI]})// This will print ['users', 2], ['users', 10]// If you dont use charwise its sorted numerically and would// print ['users', 10] , ['users', 2]userStream.on('data',console.log)