In theory, the streaming APIs are meant to make it possible to operate on streaming database queries, where the entire result set is simply too large to fit int

In theory, the streaming APIs are meant to make it possible to operate on streaming database queries, where the entire result set is simply too large to fit into memory. A single, unified API over both in-memory datastructures and streaming queries is supposed to make it easy to do both with the same set of APIs without needing reason about the details, nor materialize the entire intermediate result in memory [1]. But I find the opposite to be true; it's hard to reason about where/when the entire result might need to be collected, and the streaming APIs cannot really match the query language underneath (SQL, e.g.). Instead, streaming APIs are frankly confusing, and less efficient than just doing the straightforward for loops, IMHO. Especially when things get complicated, with multiple joins and map/reduce. [1] Another way to achieve the incremental streaming result effect is to write everything in terms of generators. It is sooo much clearer to see a loop over a data structure and a yield to know how much the computation is actually incrementalized, IMHO.