On the surface I'm not sure this explanation passes the smell test. Almost irrespective of how you get the data from your network card into v8 / nodejs, you're going to be crossing c++/v8 boundaries. It's possible that a native binding goes through this expense multiple times, but that's not likely now most of these bindings work (pg at least for sure).
As the author noted elsewhere in thread, the performance gains are mostly coming from prepared statements and pipelining which seems a bit of an apples to oranges benchmark as most of these native drivers will support similar features, just not enable them transparently like this library does.
Out of curiosity, do you have links to these other projects where they have similar benchmarking attempts/results?
> On the surface I'm not sure this explanation passes the smell test. Almost irrespective of how you get the data from your network card into v8 / nodejs, you're going to be crossing c++/v8 boundaries.
yes, you are, but the differences are the object creation that occurs. a single buffer coming from c/c++ (a socket, let's say) can be parsed and turned into a large number of objects in javascript much more quickly. yes, you're passing through that barrier once, but creating all of those objects from c++ and passing through it 20-30 times is a lot more expensive.
> Out of curiosity, do you have links to these other projects where they have similar benchmarking attempts/results?
and unfortunately, I cannot find the original discussions from when node-redis went from native to pure javascript, but it was about a 30-40% speed increase originally if memory serves (I was the one who did that original conversion after a lot of deep dives into v8 and performance crossing the barrier).
as an aside, I'm also the maintainer of plv8, and am happy to discuss the same types of performance issues of dealing with jsonb vs json (which in Postgres is text): creating objects vs a simple JSON.parse() in c++ is a significant difference.
As the author noted elsewhere in thread, the performance gains are mostly coming from prepared statements and pipelining which seems a bit of an apples to oranges benchmark as most of these native drivers will support similar features, just not enable them transparently like this library does.
Out of curiosity, do you have links to these other projects where they have similar benchmarking attempts/results?