Converting Data.Text to Int in Haskell

Question

In another question, one of the comments says, "[Data.]Text is becoming the de-facto textual implementation. String is still around for legacy reasons and for simple things, but for serious textual manipulation you should be using Text."

What is the easiest way to convert a Data.Text to an Int? read will not work because the read function always takes a String.

The best that I can come up with is:

let fortyTwo = Data.Text.pack "42"
read $ Data.Text.unpack fortyTwo :: Int

Is there a better way?

Solution 1

Looking at the text package, I see a module called Data.Text.Read. It seems to work:

λ> decimal (T.pack "99 bottles")
Right (99," bottles")
λ> decimal (T.pack "a digit")
Left "input does not start with a digit"

Solution 2

In other words, you want a parser that can consume Text. There are many parsers on hackage that can consume Text, I suggest you try attoparsec.

import Data.Attoparsec.Text
parseInt = parseOnly (signed decimal)

Comments

[Ralph]

I'm not sure that will be shorter (or better), since I will still have to deal with the Either. That might be a reasonable replacement for reads.

[shachaf]

Well, you have to deal with that anyway, for most uses of read. read will crash your program if it sees a malformed string, which is pretty bad (it would probably be better if its type was Read a => a -> Maybe String). If you're sure you want to crash your program, you can always define foo f = fst . either error id . f, and then use foo decimal instead of decimal.

[Ralph]

In my case, I know the string is a number; it comes from a regular expression match ("\d{2}").

[ДМИТРИЙ МАЛИКОВ]

What is at mean «know the string is a number»? If you know that something is a number, that it is a number. These types are not equal, that why not every String could be converted to Int. Maybe something like either (const 0) fst will be useful, which is kinda fromMaybe 0.

[shachaf]

Then feel free to do what I said. Crashing your program shouldn't be the default. :-) An alternative might be to combine the regular expression match and the reading into one function that returns Either/Maybe. It depends on what your code is like.

[Ralph]

Please don't misunderstand. I like your answer.