How to write Inline Functions in Python?

[Michael Stenner]

On Thu, Nov 14, 2002 at 05:20:40PM +1100, Delaney, Timothy wrote:
> > From: arivu at qmaxtest.com [mailto:arivu at qmaxtest.com]
> > Is Inline Functions available in python?
> 
> No.
> 
> I assume you're coming from C++. In that case, I suggest you take a good
> hard look at why you want to write inline functions/methods anyway. If
> you're doing it for performance reasons, your compiler is almost certainly
> going to be better at deciding what to inline, and has the option to do so
> unless you explicitly tell it not to. If you're doing for any other reason
> (such as putting code in header files) I strongly advise against it.
> 
> Inline was developed when computers were slower and compilers were more
> stupid. These days I see no need to inline whatsoever.

OK, I'm sincerely curious here and not trying to be argumentative.  I
present the following example.  I admit that it's a bit extreme, but
still:

==============================================================
#!/usr/bin/python2

N = 10000

def times_two(x):
    return 2 * x

def via_function():
    for i in range(N):
        y = times_two(i)

def inline():
    for i in range(N):
        y = 2 * i

def run():
    inline()
    via_function()

import profile
profile.run('run()')
==============================================================

with the following results:
   ncalls  tottime  percall  cumtime  percall
filename:lineno(function)
        1    0.000    0.000    0.220    0.220 <string>:1(?)
        0    0.000             0.000          profile:0(profiler)
        1    0.020    0.020    0.240    0.240 profile:0(run())
        1    0.010    0.010    0.010    0.010 py-inline.py:12(inline)
        1    0.000    0.000    0.220    0.220 py-inline.py:16(run)
    10000    0.050    0.000    0.050    0.000 py-inline.py:5(times_two)
        1    0.160    0.160    0.210    0.210 py-inline.py:8(via_function)

There's a resolution issue here, sometimes inline takes 0.020 rather
than 0.010, but conservatively, the "inline" version is 10 times
faster.  Now, how much of a problem this is in real live programs is
debatable.  I've often wished for an inline function for reasons that
are not too far from this.

					-Michael

-- 
  Michael Stenner                       Office Phone: 919-660-2513
  Duke University, Dept. of Physics       mstenner at phy.duke.edu
  Box 90305, Durham N.C. 27708-0305