Python is DOOMED! Again!

Mario Figueiredo marfig at gmail.com
Sat Jan 24 23:30:11 CET 2015


In article <54c2299d$0$13005$c3e8da3$5496439d at news.astraweb.com>, 
steve+comp.lang.python at pearwood.info says...
> 
> I don't think that a raise of 0.10000000000000001 (10%),
> 0.035000000000000003 (3.5%) or 0.070000000000000007 (7%) is quite what
> people intended.
> 
> (Don't use binary floating point numbers for anything related to
> money. Just don't.)

Few rules are written in stone and that rule isn't either. It all 
depends on the context. People often read these bits of well intentioned 
advise and go on to immediately make some kind of gospel out of it. Why 
is that?

And why is that? Why do you think I can't ever use floats for money? Is 
there some kind of unspoken rule that money must always be dealt with 
with enough precision to give a microscope an headache?

- What if I am storing money as an integer or a 2 decimal place Decimal 
to manage my electrical bill and I use a float to express a percentage?
- What if the float operation is not a repetitive operation that would 
indeed invariably lead to round errors, but instead a once in a lifetime 
operation?

I'm not saying I don't agree we should avoid it. I'm saying we need also 
to properly contextualize it before we decide to do so. If I'm writing a 
banking system, or a POS, you will be damn sure it will be hard to spot 
a float in my code.

But if I'm writing my household electrical bill yearly report, or I'm 
writting a damn code snippet on a python group to illustrate how type 
hinting mangles your code, who gives a flying arse if I'm using a float 
to express money? Sheesh!



More information about the Python-list mailing list