Re: [clojure-mongodb] Monger internals, sync model ?

[Michael Klishin]

2013/4/28 Yakovlev Roman <feli...@gmail.com>

What happened in monger ? for example 

(monger/find "documents" {:first_name "Ringo"}) is this request sync ? so it waits for response and returns data some time later ?

 Execution is hanging waiting for the data ?

Monger and MongoDB Java driver often return a form of a future (DBCursor or a lazy Clojure sequence).
DBCursor then loads results as needed: if you have 1000 documents in the response but only ever use 20,

it will only load one batch (the size of the batch can be set per connection).

In addition, MongoDB Java driver uses a pool of threads that execute actual requests. AFAIR

its size is in medium hundreds by default. The size can be bumped up but it's only effective

to maybe 2-3K on a 4-8 core machine.

In general, the idea that functions that return values are "blocking" only applies to

environments where you don't have threads (or similar, e.g. Erlang processes), such as Nodejs. On the JVM

plenty of libraries return values that are computed in other threads. So you get both

concurrent (and, depending on the hardware, parallel) execution and a program flow

control you can reason about.

In any case, I hardly believe the client will ever be the bottleneck in any realistic application.

--
MK

http://github.com/michaelklishin
http://twitter.com/michaelklishin