Re: Stdio.Buffer.set_buffer_mode revisited

Per Hedbor () @ Pike (-) importm?te f?r mailinglistan wrote:
>A class like this might be useful to have somewhere, incidentally:

>| protected Thread.Thread backend_thread;
>| /* initialized in various ways in our code. */
>|
>| class Blocking( function(function(mixed:void),mixed...:void) async_func )

>The main advantage of this is that there is no longer any need for a
>single thread in the low-level code, and in fact, it works even if the
>client code does not use threads either. Everything is non-blocking,
>the blocking API either waits for the result or runs the backend, as
>needed.

I'll try and wrap my head around this and see if it helps me reduce/eliminate
the threads in the pgsql driver.

>Personally I tend to see use of threads as a failure. :)

Well, usually I'd agree, however, if you want to process incoming network
data in a streaming fashion, then using callbacks usually means that at
some point you *are* copying data.  For small amounts that won't be a problem,
but in case of the database, we sometimes can handle queries that return
several megabytes worth of data, so I want to avoid all unnecessary copies.
Using threading allows me to keep state information in the thread without
needing to copy things around to (re)package it for the callback.
-- 
Stephen.