To_f and the format of E notation

ruby-talk
1

Hi,

I am parsing an ASCII file with floating point numbers that have a slightly unusual format in order to save space (each floating point number is limited to 8 characters, one being the decimal point). For example, all the following strings are acceptable representations of the number 7:
7.0 .7E1 0.7+1 .70+1 7.E+0 70.-1

I thought that adding the missing E for numbers in E notation (but without the E) before calling to_f would be enough. So I came up with the following regex:

irb(main):001:0> "70.-1".sub(/([\d\.])[eE]?([+-])/, '\1e\2')
=> "70.e-1"

However, to_f gives a wrong number:

irb(main):002:0> "70.-1".sub(/([\d\.])[eE]?([+-])/, '\1e\2').to_f
=> 70.0

Inserting 0e instead of just e fixes the problem:

irb(main):003:0> "70.-1".sub(/([\d\.])[eE]?([+-])/, '\10e\2').to_f
=> 7.0

My questions:

* Why Ruby cannot convert the first string to a float correctly? It apparently expects digits after the decimal point, otherwise it stops the conversion. Is there some specification for the exponential notation format? I could not find any. On the other hand, other languages can handle it well, e.g., in Python:

float("70.e-1")

7.0
I tested in C/C++ and Java with the same (correct) result.
* Is there a better (faster, more elegant, more concise) way of converting such strings to floats?

Thanks,

Vladimir

2

irb(main):001:0> "70.-1".sub(/([\d\.])[eE]?([+-])/, '\1e\2')
=> "70.e-1"

Remember, we use "." for many things, the most important being method dispatch.

That string looks like sending a method named 'e' to the integer 70: 70.e() - 1

Full fledged floats have both an integer part and a fractional part, or they're integers with scientific notation:

70e-1
70.0e-1

That disambiguates them from methods.

70e-1

=> 7.0

70.0e-1

=> 7.0

70.e-1

NoMethodError: undefined method `e' for 70:Fixnum
  from (irb):2
  from /usr/bin/irb:12:in `<main>'

···

On Jan 29, 2014, at 18:31, Vladimir, Chalupecky <chalupecky@jp.fujitsu.com> wrote:

3

I see. Thanks for the help.
Cheers,
Vladimir

···

-----Original Message-----
From: ruby-talk [mailto:ruby-talk-bounces@ruby-lang.org] On Behalf Of Ryan
Davis
Sent: Thursday, January 30, 2014 8:20 PM
To: Ruby Talk
Subject: Re: to_f and the format of E notation

On Jan 29, 2014, at 18:31, Vladimir, Chalupecky <chalupecky@jp.fujitsu.com> > wrote:

> irb(main):001:0> "70.-1".sub(/([\d\.])[eE]?([+-])/, '\1e\2') =>
> "70.e-1"

Remember, we use "." for many things, the most important being method
dispatch.

That string looks like sending a method named 'e' to the integer 70: 70.e() - 1

Full fledged floats have both an integer part and a fractional part, or they're
integers with scientific notation:

70e-1
70.0e-1

That disambiguates them from methods.

>> 70e-1
=> 7.0
>> 70.0e-1
=> 7.0
>> 70.e-1
NoMethodError: undefined method `e' for 70:Fixnum
  from (irb):2
  from /usr/bin/irb:12:in `<main>'