Very exciting! This looks amazing. It almost looks like half a SQL interface ;-)
On Fri, Jan 16, 2015 at 11:04 AM, Aljoscha Krettek <aljos...@apache.org> wrote: > Hi, > I did some work recently on adding support for SQL-like queries on top > of DataSets. (This is known as "named datasets" in the jira issue: > > https://issues.apache.org/jira/browse/FLINK-947?jql=project%20%3D%20FLINK%20AND%20assignee%20%3D%20currentUser()%20AND%20resolution%20%3D%20Unresolved > ). > > I have support for filter, join, grouping and aggregation. I think the > basis is quite strong now but we can add support for more data types > and supported operations in the select expressions. > > Please have a look at my branch if you're interested: > https://github.com/aljoscha/flink/tree/linq You can look at the new > Expression ITCases to see what features are currently available and > how the interface is used. There are also two complete programs: > PageRankExpression and TPCHQuery3Expression. > > And now at last, a sneak peek at how the new interface is used: > > in.group('key).select('key, ('a + 10).avg + " the average", 'a.count) > > The notation 'foo are Scala symbols, I use them in the DSL to > reference named fields. > > Cheers, > Aljoscha >
