In the Stream Reasoning (SR) research, working prototypes often came along with foundational investigations. For RDF Stream Processing (RSP) in particular, RSP engines empirically proved the approach feasibility and paved the road to application design and comparative analyses. Observing these real systems highlighted their heterogeneity and fosters new foundational achievements: RSP-QL, i.e. a reference a model that explains, uni es and can be used for correctness checking and optimization of RSP approaches. In this paper, we present Yasper 1.0 a brand new RSP engine that implements RSP-QL semantics and, we hope, will foster new empirical research on RSP.
In this section, we summarize the notions of RSP-QL [
An RDF Stream is a sequence of pairs (Oi, ti), where ti is a non-decreasing timestamp and Oi is either an RDF Graph or an RDF Triple.
The time-based sliding window operator W is a Stream-to-Relation (S2R)
operator [
A Time-Varying Graph is a function that takes a time instant as input and produces as output RDF Graph which is called Instantaneous RDF Graph. The application of W on a RDF Stream S produces a Time-Varying Graph TVGW;S that for any given time instant t at which W is de ned outputs an Instantaneous RDF Graph, which is the result of coalescing all the RDF Graphs or triples that the current window contains3.
An streaming dataset (SDS) is an extension of SPARQL dataset4 that is composed by: an optional default graph A0, n (n 0) named Time-Varying Graphs, and m (m 0) named sliding windows over k (k m) data streams.
An RSP-QL query is continuously evaluated against an SDS by an RSP engine. The set ET of all the instants at which the evaluation occurs is determined by the reporting policy of the RSP engine that executes the query. RSP-QL de nes a reporting policy for an RSP as a combination of one or more of the following strategies: CC Content Change { the engine reports if the content of the current window changes{, WC Window Close { the engine reports if the current window closes {, NC Non-empty Content { the engine reports if the the current window is not empty {, and P Periodic { the engine reports periodically.
The evaluation of a RSP-QL query outputs an instantaneous multiset of
solution mappings for each evaluation time instant in "ET". Relation-to-Stream
(R2S) [
In this section, we introduce Yasper's architecture and how it implements RSPQL concepts. Figure 1 shows the following Yasper's modules: Streams, Windowing, SDS, Querying and Reasoning. Modules that expose Yasper as a REST service are available5 but not discussed.
The Stream module, Fig 1 (a), contains the classes to represent a Stream, which is identi ed by an URI and its content. A StreamItem is identi ed by 3 The current window identi ed by W with the oldest closing time instant at t 4 https://www.w3.org/TR/rdf-sparql-query/#specifyingDataset 5 https://github.com/streamreasoning/rspservices the triple < ti; te; O > where ti is the time when s entered the system (Ingestion Time); te is the time when s occurred (Event Time) and O is a generic representation of the data in s, i.e. for RSP either a RDF Graph or a Triple6.
The Windowing module, Fig 1 (b), is based on Esper7, i.e. an open-source Data Stream Management Systems that relies on the Event Processing Language (EPL) and special objects called listeners that continuously receive the EPL queries outputs. Each EPL statement together with a dedicated listener represents a (named) time-based sliding window operator (henceforth referred as just window operator) on a RDF stream. The windowing is performed by means of temporal annotations of the StreamItems. Yasper works by default using Event Time, but it can be con gured to work with Ingestion Time8.
An Esper-based window operator maintains one time-varying graph. A window operator has two alternative ways to deliver content to Time-Varying Graph: (i) as a Snapshot { i.e. the window operator pushes to the Time-Varying Graph the whole window content { or (ii) as Deltas { i.e. window operator pushes to the Time-Varying Graph only the di erences between the current window and the previous one in terms of additions and deletions. By default Yasper works in Snapshot mode, but it can be con gured to work with Deltas. In both the ways, the Time-Varying Graphs reactively generates a Instantaneous RDF Graph any time the engine reports. Yasper reports the results on window close (WC) without empty content (NC).
Since the set ET is determined reading the StreamItems, at each evaluation time instant we can identify an Instantaneous SDS to evaluate a RSP-QL query against. A streaming dataset SDS, Fig 1 (c), can be reactively consolidated into a set of (named) Instantaneous Graphs9 at the time t at which a Time-Varying 6 We implement O using Apache Jena 3. 7 www.espertech.com 8 Event Time does not guarantee total ordering of StreamItems 9 Slowly evolving RDF graph are represented as a (named) Time-Varying Graph too.
REGISTER STREAM <example> AS CONSTRUCT ISTREAM f? s ?p ? og FROM NAMED WINDOW : win1 [RANGE 5 s , SLIDE 2 s ] ON STREAM : stream1 FROM NAMED WINDOW : win2 [RANGE 5 s , SLIDE 5 s ] ON STREAM : stream2 WHERE f WINDOW ?w1 f? s ?p ? og
WINDOW ?w2 f? s ?p ? og FILTER ( ?w1 != ?w2) g
Listing 1.1: An example of RSP-QL Query.
Graph is updated. We implemented the SDS and this mechanism by extending Apache Jena 3 in-memory dataset.
The Querying module, Fig 1 (d), contains the elements for query instantiation
and continuous execution. At this stage of development, Yasper accepts SELECT
and CONSTRUCT queries written in RSP-QL syntax (e.g. Listing 1.1).
Moreover, as the reader can observe in Listing 1.1, Yasper supports multi-streams
query, despite [
Finally, the reasoning module, Fig 1 (e), allows to answer RSP-QL queries under OWL entailment by means of Jena Rule reasoner. In the future we want to integrate RDFox, Ontop and DL reasoners e.g. Pellet or Hermit. 4
In this paper, we presented Yasper, an RSP engine for RSP-QL queries. Yasper adopts a generic stream representation, it can work with event time or ingestion time and it implements multi-stream queries evaluation. At the moment of writing, Yasper supports SELECT and CONSTRUCT queries. ASK support is a work in progress, while DESCRIBE requires further investigations for RSP. We plan to support all the reporting Policies in Yasper, either by con guration or as an extension of the proposed syntax.
Finally, we plan to perform an empirical evaluation [