Working with kernel.eval

[Ian Whitney]

I'm trying to calculate the results of formulas that are stored in
text files.

For example, a file may contain the line: 1+2/3

I want to take that line & get the result of the formula, 1.66 (roughly)

The way to do this seems to be Kernel.eval. But I can't get it to
work correctly. Here's my irb session:

irb(main):001:0> a = "1+2/3"
=> "1+2/3"
irb(main):002:0> b = eval(a)
=> 1
irb(main):004:0> b.class
=> Fixnum

I would think that eval(a) should return a Float, not a Fixnum.

I thought maybe that eval wasn't following the standard order of
operations, which could explain the return value of 1. So then I
tried this:

irb(main):005:0> a = "2/3"
=> "2/3"
irb(main):006:0> b = eval(a)
=> 0
irb(main):007:0> b.class
=> Fixnum


So, it's not an order of operations thing.

Maybe I shouldn't be using eval. But I can't find another option. Any
thoughts?

Thanks,

Ian

 

[Kent Sibilev]

$ irb
irb(main):001:0> 2/3
=3D> 0
irb(main):002:0> 2.0/3
=3D> 0.666666666666667
irb(main):003:0>

Kent.

 

[Joel VanderWerf]

I'm trying to calculate the results of formulas that are stored in text
files.

For example, a file may contain the line: 1+2/3

I want to take that line & get the result of the formula, 1.66 (roughly)

The way to do this seems to be Kernel.eval. But I can't get it to work
correctly. Here's my irb session:

irb(main):001:0> a = "1+2/3"
=> "1+2/3"
irb(main):002:0> b = eval(a)
=> 1
irb(main):004:0> b.class
=> Fixnum

I would think that eval(a) should return a Float, not a Fixnum.

It's nothing to do with eval, but rather that if you start with fixnums,
you do fixnum arithmetic. Try replacing 2 with 2.0. That invokes float
division, and every result depending on that input will be float. In
general, you can use x.to_f if you want to force a value to be treated
as float.

 

[David Vallner]

require "mathn" for a batch of tweaks that will make Ruby maths expression=
=20
more maths-like than C-like.

David Vallner

D=C5=88a Pondelok 13 Febru=C3=A1r 2006 21:07 Ian Whitney nap=C3=ADsal:

 

[James Edward Gray II]

Any thoughts on how I can take the string as it stands, "1+2/3" and
get the right answer?

Well, depending on how many operators you need to support, it's
fairly easy to parse them into an Abstract Syntax Tree and run the
calculation.

James Edward Gray II

 

[Joel VanderWerf]

Any thoughts on how I can take the string as it stands, "1+2/3" and get
the right answer? I could use regex to replace every integer with a
floating point number, I guess. But that seems kludgey.

David's suggestion to require 'mathn' is a good way to go, but be aware
that 2/3 will evaluate to a rational--precise, but in general slower to
compute with.

 

[Ian Whitney]

That does appear to be the best solution. I plugged that in & my unit =20=

tests look good.

Thanks,

Ian