LinkRest at JeeConf 2017
•Download as PPTX, PDF•
1 like•396 views
LinkRest (http://linkrest.io) is a small HTTP-based protocol and a Java framework that rethinks how REST APIs should be built and consumed. It turns each API endpoint into a graph query engine, giving the client full control over the shape and contents of the requested data sets, while the server controls query strategy and security. The talk will start with the easy parts – how to reuse an existing Apache Cayenne ORM model for the REST service and write simple Java one-liners to get it up and running. Then we will delve into alternative backends, graph access security and parallel data processing.
Report
Share
LinkRest at JeeConf 2017
- 1. RESTful Data Services with LinkRest by Andrus Adamchik @andrus_a 1
- 2. About Me • Open source developer: ORM, DI, REST, no-container • Member and ex-VP of the Apache Software Foundation • Run a company called ObjectStyle 2
- 3. I Love REST • HTTP is a clean and universal protocol • Near-infinite scalability 3
- 4. Hi! I am a REST API Client 4
- 7. Aggregate-Centric REST API • An aggregate handled by a single REST resource • Decoupling / encapsulation • Assumes "normal" use cases don't cross aggregate boundaries 7
- 8. Problem #1: Too Much Data curl -i 'https://api.github.com/repos/nhl/link-rest/issues?state=all' 8 • Returns 126K JSON • We need 3K ("title", "number", "id")
- 9. Problem #2: Too Little Data 9 curl -i 'https://api.github.com/repos/nhl/link-rest/issues?state=all' { // O(N) problem: // for each issue fetch events via a separate REST call "events_url" : "https://api.github.com/.../issues/207/events" ... }
- 10. LinkRest • From aggregates to graphs • Client decides on request shape / size • One size fits all • Reuse existing (ORM) models 10
- 11. Demo: Hockey Schedule App 11
- 14. 14 ?cayenneExp=inProgress=true SELECT t0.* FROM game t0 WHERE t0.in_progress = 1 SELECT t0.* FROM game t0 JOIN team t1 ON (t0.home_team_id = t1.id) WHERE t1.name LIKE 'New York%' ?cayenneExp=["homeTeam.name like $t","New York%"] Filtering
- 15. 15 ?sort=dateTime SELECT t0.* FROM game t0 ORDER BY t0.date_time SELECT t0.* FROM game t0 ORDER BY t0.date_time DESC ?sort=dateTime&dir=DESC SELECT t0.* FROM game t0 JOIN team t1 ON (t0.home_team_id = t1.id) ORDER BY t0.in_progress, t1.name DESC ?sort=[{"property":"inProgress"},{"property":"homeTeam.name","direction":"DESC"}] Sorting
- 16. 16 ?start=2 {"data":[ /*objects 2..9*/ ], "total":10} Pagination (for a list of 10 objects) {"data":[ /*objects 0..4*/ ], "total":10} ?limit=5 {"data":[ /*objects 2..6*/ ], "total":10} ?start=2&limit=5
- 17. 17 ?include={"path":"homeGames","cayenneExp":"arena = 'Madison Square Garden'","sort":"dateTime"}&include=homeGames.id { "homeGames": [ { "id": 2016020075 }, { "id": 2016020790 } ] } Advanced Include
- 18. Constraints API 18 • Prevents DOS attacks • Fine-grained attribute security ConstraintsBuilder<Team> tc = Constraint.idAndAttributes(Team.class); ConstraintsBuilder<Game> gc = Constraint.idAndAttributes(Game.class) .path("homeTeam", tc) .path("visitingTeam", tc); return LinkRest.select(Game.class, configuration) .uri(uri) .constraint(gc) .get();
- 20. Create/Update Flavors • create(..) • update(..) • createOrUpdate(..) • idempotentCreateOrUpdate(..) • idempotentFullSync(..) • delete(..) 20
- 21. Demo: Live Game Events App 21
- 22. Takeaway • POJOs can be exposed via LinkRest using annotations • Custom backend is plugged in as a "stage listener" • "include" / "exclude" / "limit" supported automatically • Other protocol features are up to the backend to implement 22
- 23. Demo: Hockey Schedule with Live Scores 23
- 24. Takeaway • LinkRest can do parallel fetching and aggregation from multiple data sources. 24
- 25. Great Minds Think Alike 25