spy16 / fabric

Fabric is a simple triplestore written in Golang

Fabric

Fabric is a triple-store written in Go. Fabric provides simple functions and store options to deal with "Subject->Predicate->Object" relations or so called triples.

Usage

Get fabric by using go get -u github.com/spy16/fabric (Fabric as a library has no external dependencies)

// See next snippet for using persistent SQL backend
fab := fabric.New(&fabric.InMemoryStore{})

fab.Insert(context.Background(), fabric.Triple{
    Source: "Bob",
    Predicate: "Knows",
    Target: "John",
})

fab.Query(context.Background(), fabric.Query{
    Source: fabric.Clause{
        Type: "equal",
        Value: "Bob",
    },
})

To use a SQL database for storing the triples, use the following snippet:

db, err := sql.Open("sqlite3", "fabric.db")
if err != nil {
    panic(err)
}

store := &fabric.SQLStore{DB: db}
store.Setup(context.Background()) // to create required tables

fab := fabric.New(store)

Fabric SQLStore uses Go's standard database/sql package. So any SQL database supported through this interface (includes most major SQL databases) should work.

Additional store support can be added by implementing the Store interface.

type Store interface {
    Insert(ctx context.Context, tri Triple) error
    Query(ctx context.Context, q Query) ([]Triple, error)
    Delete(ctx context.Context, q Query) (int, error)
}

Optional Counter and ReWeighter can be implemented by the store implementations to support extended query options.

REST API

The server package exposes REST APIs (/triples endpoint) which can be used to query, insert/delete or reweight triples using any HTTP client.

fabric's Issues

query interface

before looking more closely here, may i ask a question?
we have a triples file (file.nt) with multi-million items.
if we installed fabric, and loaded fabric with that file, how would we query fabric+file.nt?
for example, could we query using sparqler (http://sparql.org/sparql.html) where fabric would reveal a graph endpoint?
we are looking for a triple store that does not involve the overhead associated with apache jena/fuseki.
we would like to have the equivalent of a REST server that can be queried by a SPARQL processor (like sparqler).
we would install fabric on AWS, so the ideal configuration would be that the data in the triple store is on S3.
clearly, AWS aurora is an option, but we would like the lightest-weight solution for a release-1 implementation.
thanks in advance for advice and guidance.