Re: [Monotone-devel] Re: HOWTO: benchmarking monotone (was Re: "memory exhausted" error for 'mtn list status' command)

[Nathaniel Smith]

On Wed, Oct 11, 2006 at 11:52:26PM +0200, Lapo Luchini wrote:
> Nathaniel Smith wrote:
> > As an example, let's look at the InitialPull benchmark.
> 
> I know no Python, I have no idea what a "requirement" is, I wonder why

"requirements" are part of the benchmark harness, nothing to do with
python -- before the harness runs some benchmark, it looks at the
benchmark object to see if it has any "requirements" (other runnable
objects), and if it does, it runs those first.  That's all :-).

> "this" as to be declared as a first parameter (linke in a language born
> when OOP meanth nothing 0_o???), but thanks to this clear "tutorial" I

People say all sorts of crazy things about the initial "self" argument
on python methods, and it does take a bit to get used to typing it at
the right times, but the idea is that there are no "invisible
variables" -- 'this' in C++ for example basically _is_ a function
argument, it's passed in from outside and it has function scope, but,
it's magic and implicit.  In python, all the local variables come from
the function arguments.

> managed anyway to do first real "code" contribution:
> the CVSImport benchmark ;-)

Awesome!

> I'll time-test it on a variety of platforms and examples, for now I can
> only say that, on Cygwin, mtn-0.30 is consistently 6% faster at
> cvs_import than mtn-0.26 was 0=)

Awesome^2!

(Nice to hear that the harness works on cygwin, too, there are
definitely a bunch of unixisms in there :-).)

-- Nathaniel

-- 
Eternity is very long, especially towards the end.
  -- Woody Allen