Question

EDIT: A working regex (take the second group):

(^|[ ,\t\n]+)([0-9\\.]+)($|[ ,\t\n]+)

Original post:

I’m new to Haskell, trying to use Text.Regex (from regex-compat) to extract float values from a string. I want my regex to match any series of numbers and periods that is buffered by at least one separator character to the left and the right. This is what I wrote:

regex = "[^ \t\n,]+([0-9\\.])+[$ \t\n,]+"

EDIT: I originally thought this worked properly in Scala, but I now believe I simply got lucky with my test strings. This does not work in Haskell. An example:

matchRegexAll (mkRegex regex) " 12.34 "

yields

Just (" ","12.34 ","",["4"])

when it seems to me it should yield

Just (""," 12.34 ","",["12.34"])

Another example:

matchRegexAll (mkRegex regex) "12.34"

yields

Nothing

when it I think it should yield

Just ("","12.34","",["12.34"])

I’m guessing the parser treats “^” and “$” differently that does the Scala parser, but that’s all I’ve got.

Answer 1

Inside a character class like [^ \t\n,], normal regex metacharacters (such as ^ and $) lose their special meaning; they match themselves instead.¹ Something like (^|[ \t\n,]) should do what you want.

I’m surprised your regex works in Scala; I’ve never seen a regex implementation that doesn’t behave in this manner.

¹ Although as FlopCoder points out, ^ at the start of a character class actually negates it.