An in-process key-value store
tiny is an in-process document/object store for node.js.
It is largely inspired by nStore, however, its goal was to implement real querying which goes easy on the memory.
Tiny is very simple, there are no schemas, just store your objects. It supports mongo-style querying, or alternatively a "mapreduce-like" interface similar to CouchDB's views.
$ npm install tiny
Tiny takes advantage of the fact that, normally, when you query for records in a database, you're only comparing small properties (<128b) in the query itself. For example, when you query for articles on a weblog, you'll usually only be comparing the timestamp of the article, the title, the author, the category, the tags, etc. - pretty much everything except the content of the article itself.
Tiny stores each document/object's property individually in the DB file and caches all the small properties into memory when the DB loads, leaving anything above 128b behind. When a query is performed, Tiny only lets you compare the properties stored in memory, which is what you were going to do anyway. Once the query is complete, Tiny will perform lookups on the FD to grab the large properties and put them in their respective objects before results are returned to you.
This my attempt at combining what I think the best aspects of nStore and node-dirty are. node-dirty is incredibly fast and simple (everything is in-memory), and nStore is very memory effecient, (but this only lasts until you perform a query). node-tiny allows for queries that perform lookups on the db file, and it selectively caches properties as well, so it's fast and easier on memory.
var Tiny = require'./tiny';Tiny'articles.tiny'var time = Datenowlow = time - 60*60*1000high = time - 30*60*1000;// mongo-style querydbfind$or:timestamp: $lte: lowtimestamp: $gte: highdesc'timestamp'limit3console.log'Results:' results;;// is equivalent to...dbfetchdesc: 'timestamp'limit: 3if doctimestamp <= low|| doctimestamp >= highconsole.log'Found:' key;return true;console.log'Results:' results;;;
The mongo-style querying should be fairly self-explanatory. The second query is
supposed to be similar to a mapreduce interface, but it's the rough equivalent
Note: there is a
shallow parameter for
it will only lookup properties that are under 128b in size. This is to go
easy on the memory.
.all are shallow by default, but they do have
deep parameter, (which I don't recommend using).
// save a documentdbset'myDocument'title: 'a document'content: 'hello world'console.log'set!';;// .each will iterate through// every object in the database// it is shallow by defaultdbeachconsole.logdoctitle;;// returns every object in the DB// in an array, this is shallow// by defaultdballconsole.logdocslength;;// remove a docdbremove'myDocument'console.log'deleted';;// retrieve an object from the databasedbget'someOtherThing'// data._key is a property which// holds the key of every objectconsole.log'found:' data_key;;// updates the object// without overwriting its other propertiesdbupdate'article_1'title: 'new title'console.log'done';;// close the file descriptordbcloseconsole.log'db closed';;// clean up the messdbcompactconsole.log'done';;// dump the entire database to a JSON file// in the same directory as the DB file// (with an optional pretty-print parameter)dbdumptrueconsole.log'dump complete';;
Because of the way Tiny works, there are ways to alter your data to make it more memory efficient. For example, if you have several properties on your objects that aren't necessary to for queries, its best to nest them in an object.
user:name: 'joe'prop1: 'data'prop2: 'data'prop3: 'data'user:name: 'joe'data:prop1: 'data'prop2: 'data'prop3: 'data'
That way, the data will not be cached if it exceeds 128b collectively.
Eventually there may be an
ignore method or an
index method, which will be
explicitly inclusive or exclusive to which properties are cached and which
properties are able to be referenced within a query.
Creates and returns a database with the given name.
var db;Tiny'./articles.tiny'if err throw err;db = db_;;
Dumps the a database to a JSON file with the name as name.json. Pretty specifies whether to indent each line with two spaces or not. Alternatively, dump(func) can be called.
Closes the Tiny database file handle. A new Tiny object must be made to reopen the file.
Closes the Tiny database file, deletes the file and all the data in the database, and then creates a new database with the same name and file.
dbkillconsole.log'db has been destroyed and a new db has been loaded';;
Saves a object
doc to database under the key
docKey. Ideally, docKey should
be 128b or smaller.
dbset'myDocument'title: 'a document'content: 'hello world'console.log'set!';;
Iterates through every object in the database.
docfrom the database
trueif every object should be returned,
falseor unset if only cacheable objects should be returned (ones smaller than 128b)
If you contribute code to this project, you are implicitly allowing your code
to be distributed under the MIT license. You are also implicitly verifying that
all code is your original work.
Copyright (c) 2011-2014, Christopher Jeffrey. (MIT License)
See LICENSE for more info.