How to parse ruby BigDecimal inspect? - ruby

Question

In the following code:
x = BigDecimal(10)
s = x.inspect # "#<BigDecimal:6fe4790,'0.1E2',9(36)>"
Is there a way to parse s and get the original value ? The reason is that I have some text files with BigDecimal written in them using inspect, and I need to parse these values.

Answer 1

You .to_s to get the value in string. .inspect will print the object
x = BigDecimal(10)
x.to_s
# => "0.1E2"

Answer 2

The documentation for BigDecimal#inspect is incomplete. Consider the following:
require 'bigdecimal`
BigDecimal.new("1.2345").inspect
#=> "#<BigDecimal:7fb06a110298,'0.12345E1',18(18)>"
...
BigDecimal.new("1.234567890").inspect
#=> "#<BigDecimal:7fb06a16ab58,'0.123456789E1',18(27)>"
BigDecimal.new("1.2345678901").inspect
#=> "#<BigDecimal:7fb06a14a6a0,'0.1234567890 1E1',27(27)>"
BigDecimal.new("1.23456789012").inspect
#=> "#<BigDecimal:7fb06a1393a0,'0.1234567890 12E1',27(27)>"
BigDecimal.new("1.234567890123").inspect
#=> "#<BigDecimal:7fb06a123780,'0.1234567890 123E1',27(27)>"
It can be seen from the source code for inspect that, if there are more than 10 significant digits, each 10 characters are separate by a space (for readability, presumably):
BigDecimal.new("123.456789012345678901234567").inspect
#=> "#<BigDecimal:7fb06a0ac8b0,'0.1234567890 1234567890 1234567E3',36(36)>"
I suggest retrieving the string representation of the BigDecimal value as follows:
str = "#<BigDecimal:7fb06a14a6a0,'0.1234567890 1E1',27(27)>"
str.delete(' ').split(?')[1]
#=> "0.12345678901E1"
We are not finished. We must still convert the string we extract to a numerical object. We cannot use BigDecimal#to_f, however, if the value is large in absolute value:
"1.23456789012345678".to_f
#=> 1.2345678901234567
The safest course of action is to return a BigDecimal object, using the method BigDecimal::new, which takes two arguments:
the value to be converted to a BigDecimal object, which can be an Integer, Float, Rational, BigDecimal, or String. If a String, which is what we will supply, "spaces are ignored and unrecognized characters terminate the value" (similar to "123.4cat".to_f #=> 123.4).
the number of significant digits. If omitted or zero, the number of significant digits is determined from the value. I will omit this argument. (For example, BigDecimal.new("0.1234E2").precs #=> [18, 18], where the array contains the current and maximum numbers of significant digits.
Note the second argument is required if the first is a Float or Rational, else it is optional.
We therefore can write:
require 'bigdecimal'
def convert(str)
BigDecimal.new(str.delete(' ').split(?')[1])
end
convert "#<BigDecimal:7facd39d7ee8,'0.1234E4',9(18)>"
#=> #<BigDecimal:7facd39c7de0,'0.1234E4',9(18)>
convert "#<BigDecimal:7facd39b7be8,'0.1234E2',18(18)>"
#=> #<BigDecimal:7facd39ae610,'0.1234E2',18(18)>
convert "#<BigDecimal:7facd3990638,'0.1234E0',9(18)>"
#=> #<BigDecimal:7facd3980aa8,'0.1234E0',9(18)>
convert "#<BigDecimal:7facd3970e28,'0.1234E-2',9(18)>"
#=> #<BigDecimal:7facd39625d0,'0.1234E-2',9(18)>
v = convert "#<BigDecimal:7fb06a123780,'0.1234567890 123E1',27(27)>"
#=> #<BigDecimal:7fb069851d78,'0.1234567890 123E1',27(27)>
An easy way to see if the BigDecimal object can be converted to a float without loss of accuracy is:
def convert_bd_to_float(bd)
f = bd.to_f
(bd==BigDecimal.new(f.to_s)) ? f : nil
end
convert_bd_to_float BigDecimal.new('1234567890123456')
#=> 1.234567890123456e+15
convert_bd_to_float BigDecimal.new('12345678901234567')
#=> nil

Answer 3

"#<BigDecimal:6fe4790,'0.1E2',9(36)>"[/(?<=').+(?=')/]
# => "0.1E2"

Answer 4

I don't know which version of Ruby you are using, so I checked some MRI source code for BigDecimal:
2000 /* Returns debugging information about the value as a string of comma-separated
2001 * values in angle brackets with a leading #:
2002 *
2003 * BigDecimal.new("1234.5678").inspect ->
2004 * "#<BigDecimal:b7ea1130,'0.12345678E4',8(12)>"
2005 *
2006 * The first part is the address, the second is the value as a string, and
2007 * the final part ss(mm) is the current number of significant digits and the
2008 * maximum number of significant digits, respectively.
2009 */
2010 static VALUE
2011 BigDecimal_inspect(VALUE self)
2012 {
2013 ENTER(5);
2014 Real *vp;
2015 volatile VALUE obj;
2016 size_t nc;
2017 char *psz, *tmp;
2018
2019 GUARD_OBJ(vp, GetVpValue(self, 1));
2020 nc = VpNumOfChars(vp, "E");
2021 nc += (nc + 9) / 10;
2022
2023 obj = rb_str_new(0, nc+256);
2024 psz = RSTRING_PTR(obj);
2025 sprintf(psz, "#<BigDecimal:%"PRIxVALUE",'", self);
2026 tmp = psz + strlen(psz);
2027 VpToString(vp, tmp, 10, 0);
2028 tmp += strlen(tmp);
2029 sprintf(tmp, "',%"PRIuSIZE"(%"PRIuSIZE")>", VpPrec(vp)*VpBaseFig(), VpMaxPrec(vp)*VpBaseFig());
2030 rb_str_resize(obj, strlen(psz));
2031 return obj;
2032 }
2033
So, what you want seems to be the second part of the inspect-string, 0.1E2 in your case, equal to 10. The comment above is quite clear, this should be the full numeric value of the object. Simple regex will be enough.

Answer 5

another option:
"#<BigDecimal:95915c4,'0.1E2',9(27)>".split(",")[1].tr! "'", ''
=> "0.1E2"