[issue22458] Add fractions benchmark

[STINNER Victor]

STINNER Victor added the comment:

The issue is not closed, but moved to 
https://github.com/python/performance/pull/10

Please continue the discussion there. I'm now slowly trying to move benchmark 
issues to GitHub.

--
resolution:  -> third party
status: open -> closed

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Krah]

Stefan Krah added the comment:

Which Stefan? :)


Anyway, the first half of this issue was centered around the
proposition that fractions are a "better decimal", and the
latest pull request still has the "decimal backend" in it. :)


Fast fractions have been around for a long time (Lisp/sbcl),
but apparently they have never been widely used.


A fraction-only benchmark is okay, but comparing fractions/decimals
in one particular "official" benchmark is a bit strange.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[STINNER Victor]

STINNER Victor added the comment:

Stefan Krah added the comment:
> Fractions will shine in most benchmarks where decimal is exact anyway.

According to Stefan's latest message, the purpose of the benchmark is
not the compare decimal to fractions, but measure the performance of
fractions for one Python implementation. For example, it can be used
to check that fractions is faster in Python 3.6 than Python 3.4
fractions.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Krah]

Stefan Krah added the comment:

The reason is that the benchmark is a softball for fractions. Using the fastest 
fraction implementation (gmpy.mpq) and the best printing method for both types, 
gmpy.mpq seems to be about 2.5 to 3 times slower than decimal.

It is however easy to generate benchmarks where mpq is 3000 times slower, 
depending on how many prime factors accumulate in the denominator.


Fractions will shine in most benchmarks where decimal is exact anyway.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

Stefan Behnel added the comment:

> So this benchmark cannot be used to show the superiority of exact fractions.

I don't see how a benchmark would be a way to show that. It's certainly not the 
goal of this benchmark to show that one is computationally better than the 
other. But if a benchmark for one part of Python allows a direct, reasonable 
speed comparison with another, why would you object to make use of that?


> is the purpose of this benchmark to show that fractions are slow and generate 
> interest in changing the situation?

There have been a couple of attempts to make them faster already. The speed 
improvements in Py3.5 and 3.6 directly result from those changes. Thus, making 
these improvements visible and comparable across Python versions and different 
implementations is one of the goals of this benchmark.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Krah]

Stefan Krah added the comment:

I've left comments on GitHub.


[scoder] 
> As I said, where ever exact calculations are needed..

Even if the formatting comment is addressed, the main problem with
this benchmark is that *both* fraction and decimal calculations
are in fact exact here.

You can check that by printing out the values for "t" in the loop
like I suggested on GitHub.


This is of course one of the main features of Decimal: If you're
willing to take the penalty of a large precision, calculations
can often be exact; similar to taking the penalty of potentially
huge numerators and denominators in fractions.


So this benchmark cannot be used to show the superiority of exact
fractions.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[STINNER Victor]

STINNER Victor added the comment:

@Stefan Krah: Please review the pull request.

https://github.com/python/performance/pull/10

I suggested to only run the benchmark with the fractions module. I don't see 
the point of running the benchmark with the decimal module.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Krah]

Stefan Krah added the comment:

I'm also not opposed to adding it (-0.000) as long as we rename it to 
bm_fractions.py and change the docstring to "based on the telco benchmark".

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[STINNER Victor]

STINNER Victor added the comment:

> I think string conversion should be part of this benchmark, or it should be 
> renamed to fraction-arith or something.

The "telco" benchmark seems to be standard and well defined. I agree
that we should use a different name.

 > That said, is the purpose of this benchmark to show that fractions
are slow and generate interest in changing the situation?

I'm not really convinced that the benchmark is useful in the Python
benchmark suite, but I'm not strongly opposed to adding it neither :-)

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Krah]

Stefan Krah added the comment:

I think string conversion should be part of this benchmark, or it should be 
renamed to fraction-arith or something.

The formatting function can be a significant bottleneck, and if you use 
Fractions for financial calculations the result will still need to be printed 
in decimal floating point form.


That said, is the purpose of this benchmark to show that fractions are slow and 
generate interest in changing the situation?

--
nosy: +skrah

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

Stefan Behnel added the comment:

Done:
https://github.com/python/performance/pull/10

Note that I didn't replace the existing telco benchmark as it is more specific 
to Decimal. The new benchmark makes it possible to compare the decimal and 
fractions modules for similar operations, though.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Mark Dickinson]

Changes by Mark Dickinson :


--
nosy:  -mark.dickinson

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[STINNER Victor]

STINNER Victor added the comment:

Can you please send a pull request to the new 
https://github.com/python/performance project?

--
nosy: +haypo

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Krah]

Changes by Stefan Krah stefan-use...@bytereef.org:


--
nosy:  -skrah

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

New submission from Stefan Behnel:

Fractions are great for all sorts of exact computations (including 
money/currency calculations), but are quite slow due to the need for 
normalisation at instantiation time.

I adapted the existing telco benchmark to use Fraction instead of Decimal to 
make this problem more visible. One change I made was to take the data reading 
out of the measured loop. I/O is not part of what the benchmark should measure.

Please consider adding it to the benchmark suite.

--
components: Benchmarks
files: telco_fractions.py
messages: 227252
nosy: scoder
priority: normal
severity: normal
status: open
title: Add fractions benchmark
type: enhancement
Added file: http://bugs.python.org/file36684/telco_fractions.py

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Mark Dickinson]

Changes by Mark Dickinson dicki...@gmail.com:


--
nosy: +mark.dickinson

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

Stefan Behnel added the comment:

I just thought that it might also be nice to have a direct comparison with 
Decimal, so here's an updated benchmark that has an option --use-decimal to 
run the same code with Decimal instead of Fraction.

Decimal is about 66x faster with Py3.4 on my side (due to the C implementation).

--
Added file: http://bugs.python.org/file36685/bm_telco_fractions.py

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Antoine Pitrou]

Antoine Pitrou added the comment:

Where are Fractions used in the real world?

--
nosy: +pitrou, skrah

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

Stefan Behnel added the comment:

As I said, where ever exact calculations are needed. I use them for currency 
calculations, for example, as they inherently avoid rounding errors during the 
calculations regardless of the relative size of values. They are basically like 
Decimal but with arbitrary precision.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Antoine Pitrou]

Antoine Pitrou added the comment:

Le 22/09/2014 14:51, Stefan Behnel a écrit :
 
 I use them for currency calculations, for example,
 as they inherently avoid rounding errors during the
 calculations regardless of the relative size of values.

Do other people use them for that purpose, or are you
the only one?

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

Stefan Behnel added the comment:

I admit that I keep meeting developers who are not aware of their merits, 
simply because many other programming languages don't have them available. 
Specifically, many developers with a Java background firmly believe that 
BigDecimal is the only way to do money calculations. That's clearly wrong, and 
the heap of (Big-)Decimal bugs in their software that I've been fixing over the 
years shows just how wrong it is. Using Fractions would have avoided most of 
them. Decimal is surprisingly difficult to use right, Fractions are actually 
difficult to get wrong. They are a major feature of the standard library.

Your question is another good indication that the general awareness for them 
still needs to be raised. I'm not entirely sure how this relates to this 
ticket, though. I doubt that you are seriously suggesting that I am the only 
person using a module in the stdlib, so what exactly are you trying to say?

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Antoine Pitrou]

Antoine Pitrou added the comment:

My point is that if fractions are little used right now, there's little point 
in adding a benchmark about them in the official benchmark suite. The 
benchmark suite does not aim at measuring every possible aspect of Python 
performance, but at showcasing representative workloads.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Mark Dickinson]

Mark Dickinson added the comment:

I've always viewed the `Fraction` type as more of a teaching tool than 
something intended for real-world calculations.  If that's not how others see 
it, then it might be worth investing some effort in reimplementing 
`Fractions.gcd` in C: the `Fraction` type is almost unusably slow for serious 
work, and the gcd computations are a clear bottleneck. I've experimented in the 
past with a C-based implementation of Lehmer's algorithm for gcd computation, 
getting around a ten-fold speedup.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Stefan Behnel]

Stefan Behnel added the comment:

Speed improvements for fractions should have their own ticket(s). I created 
issue 22464 for this.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue22458] Add fractions benchmark

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

 Where are Fractions used in the real world?

For example Andrew Svetlov uses them in his online game [1]. May be it will 
explain this in detail.

[1] http://asvetlov.blogspot.com/2012/08/numerics.html (on Russian).

--
nosy: +asvetlov, serhiy.storchaka

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue22458
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com