Have naturalorder consider decimals and scientific notation

[billdenney]

Currently, naturalorder considers only integers. When decimals or scientific notation is given, it sorts as though all non-digit characters are text-- including period and negative numbers.

> naturalorder(c("2", "1", "0"))
[1] 3 2 1
> naturalorder(c("0.10", "0.020", "0.0030"))
[1] 1 2 3
> > naturalorder(c("-0.10", "-0.010", "-0.0010"))
[1] 3 2 1

Would you be interested in a pull request that would make both of these work as expected from a floating point interpretation? If you wouldn't want it to always apply, perhaps an option on how to interpret the order such as a new option: interpretation=c("natural", "integers", "real", "scientific")

And, a different regular expression would be associated with each.

[kos59125]

I'm interested in your PR.
What happens to vectors such like c("1.0.1", "1.1.0", "1.0") for each interpretation?

[billdenney]

The way I'm thinking of it, the order would be as follows for each interpretation:

• "natural": 2 3 1
• "integer": 2 3 1
  • Same tokenization as "natural" in this case
• "real": 2 3 1
  • Note that the tokenization of each string would have been different than natural here, but the result happens to be the same:
    • 1.0, ".", 1
    • 1.1, ".", 0
    • 1.0
• "scientific": 2 3 1
  • Same tokenization as "real" in this case

Where it would get more complex is something like the c("1.0-1", "1.1-0", "1.0", "1.0-0")

• "natural": 3 4 1 2
• "integer": 2 4 1 3
  • Tokenization:
    • 1, ".", 0, -1
    • 1, ".", 1, -0 (interpreted as just 0)
    • 1, ".", 0
    • 1, ".", 0, -0
• "real": 2 4 1 3
  • Tokenization:
    • 1.0, -1
    • 1.1, -0 (interpreted as just 0)
    • 1.0
    • 1.0, -0
• "scientific": 2 4 1 3
  • Same tokenization as "real" in this case