Its goal is to be a part of the developer's toolbox where Linked Data and graph-shaped data (semantic webs, social networks, etc) in general are concerned.
- Massive cleanup and restructuring is largely done, it should be even easier to add to Cayley. (thanks @kortschak)
- A couple new backends are in progress, namely Postgres and Cassandra -- PRs when they come around.
- Cayley is now in Homebrew, thanks to @whitlockjc
- Our first client API (for Clojure, thanks to @wjb) -- list is now started on the Client API wiki page
- Written in Go
- Easy to get running (3 or 4 commands, below)
- RESTful API
- or a REPL if you prefer
- Built-in query editor and visualizer
- Multiple query languages:
- Plays well with multiple backend stores:
- Modular design; easy to extend with new languages and backends
- Good test coverage
- Speed, where possible.
Rough performance testing shows that, on consumer hardware and an average disk, 134m triples in LevelDB is no problem and a multi-hop intersection query -- films starring X and Y -- takes ~150ms.
* Note that while it's not exactly Gremlin, it certainly takes inspiration from that API. For this flavor, see the documentation.
Grab the latest release binary and extract it wherever you like.
If you prefer to build from source, see the documentation on the wiki at How to start hacking on Cayley
cd to the directory and give it a quick test with:
./cayley repl --dbpath=testdata.nt
You should see a
cayley> REPL prompt. Go ahead and give it a try:
For somewhat more interesting data, a sample of 30k movies from Freebase comes in the checkout.
gzip -cd 30kmoviedata.nt.gz > 30kmovies.nt ./cayley repl --dbpath=30kmovies.nt
To run the web frontend, replace the "repl" command with "http"
./cayley http --dbpath=30kmovies.nt
And visit port 64210 on your machine, commonly http://localhost:64210
You'll notice we have a special object,
g, which is how you can interact with the graph.
The simplest query is merely to return a single vertex. Using the 30kmovies.nt dataset from above, let's walk through some simple queries:
// Query all vertices in the graph, limit to the first 5 vertices found. graph.Vertex().GetLimit(5) // Start with only one vertex, the literal name "Humphrey Bogart", and retreive all of them. graph.Vertex("Humphrey Bogart").All() // `g` and `V` are synonyms for `graph` and `Vertex` respectively, as they are quite common. g.V("Humphrey Bogart").All() // "Humphrey Bogart" is a name, but not an entity. Let's find the entities with this name in our dataset. // Follow links that are pointing In to our "Humphrey Bogart" node with the predicate "name". g.V("Humphrey Bogart").In("name").All() // Notice that "name" is a generic predicate in our dataset. // Starting with a movie gives a similar effect. g.V("Casablanca").In("name").All() // Relatedly, we can ask the reverse; all ids with the name "Casablanca" g.V().Has("name", "Casablanca").All()
You may start to notice a pattern here: with Gremlin, the query lines tend to:
Start somewhere in the graph | Follow a path | Run the query with "All" or "GetLimit"
g.V("Casablanca") | .In("name") | .All()
And these pipelines continue...
// Let's get the list of actors in the film g.V().Has("name","Casablanca") .Out("/film/film/starring").Out("/film/performance/actor") .Out("name").All() // But this is starting to get long. Let's use a morphism -- a pre-defined path stored in a variable -- as our linkage var filmToActor = g.Morphism().Out("/film/film/starring").Out("/film/performance/actor") g.V().Has("name", "Casablanca").Follow(filmToActor).Out("name").All()