You can do this efficiently with Apache Beam but you would need to write code which converts a users expression into a set of PTransforms or create a few pipeline variants for commonly computed outcomes. There are already many transforms which can compute things like min, max, average. Take a look at the javadoc[1]. It seems like you would want to structure your pipeline like: ReadFromFile -> FilterRecordsBasedUponTimestamp -> Min.perKey/Max.perKey/Average.perKey/... -> OutputToFile
It doesn't seem like windowing/triggers would provide you much value based upon what you describe. Also, it sounds like you would be interested in the SQL development that is ongoing which would allow users to write these kinds of queries without needing to write a complicated pipeline. The feature branch[2] is looking to be merged into master soon and become part of the next release. 1: https://beam.apache.org/documentation/sdks/javadoc/2.0.0/index.html?org/apache/beam/sdk/transforms/Min.html 2: https://github.com/apache/beam/tree/DSL_SQL On Wed, Jul 19, 2017 at 4:31 AM, <[email protected]> wrote: > > > Hello, > > I want to create a lib which generates features for potentially very large > datasets. > > Each file 'F' of my dataset is composed of at least : > - an id ( string or int ) > - a timestamp ( or a long value ) > - a value ( int or string ) > > I want my tool to : > - compute aggregate function for many couple 'instants + duration' > ===> FOR EXAMPLE : > ===== compute for the instant 't = 2001-01-01' aggregate functions for > data between 't-1month and t' and 't-12months and t-9months' and this, FOR > EACH ID ! > ( aggregate function such as min/max/count/distinct/last/mode or user > defined ) > > My constraints : > - I don't want to compute aggregate for each tuple of 'F' > ---> I want to provide a list of couples 'instants + duration' ( > potentially large ) > - My 'window' defined by the duration may be really large ( but may > contain only a few values... ) > - I may have many id... > - I may have many timestamps... > > ======================================================== > ======================================================== > ======================================================== > > Let me describe this with some kind of example to see if Apache Beam may > help me to do that : > > Let's imagine that I have all my data in a DB or a file with the following > columns : > id | timestamp(ms) | value > A | 1000000 | 100 > A | 1000500 | 66 > B | 1000000 | 100 > B | 1000010 | 50 > B | 1000020 | 200 > B | 2500000 | 500 > > ( The timestamp is a long value, so as to be able to express date in ms > from 0000-01-01 to today ) > > I want to compute operations such as min, max, average, last on the value > column, for a these couples : > -> instant = 1000500 / [-1000ms, 0 ] ( i.e. : agg. data betweem [ t-1000ms > and t ] > -> instant = 1333333 / [-5000ms, -2500 ] ( i.e. : agg. data betweem [ > t-5000ms and t-2500ms ] > > > And this will produce this kind of output : > > id | timestamp(ms) | min_value | max_value | avg_value | last_value > ------------------------------------------------------------------- > A | 1000500 | min... | max.... | avg.... | last.... > B | 1000500 | min... | max.... | avg.... | last.... > A | 1333333 | min... | max.... | avg.... | last.... > B | 1333333 | min... | max.... | avg.... | last.... > > > > Do you think we can do this efficiently with Apache beam, and do you have > an idea on "how" ? > > > Thanks a lot .... >
