Numeric types’ Read instances use exponential CPU/memory

[andersk]

Computing

read "1e99999999"

Int -- returns 0

read "1e99999999"

Integer -- returns [REDACTED]

read "1e99999999"

Float -- returns Infinity

read "1e99999999"

Double -- returns Infinity

takes rather a lot of CPU time and memory, and adding more digits to the exponent makes it take exponentially more. In some applications, this behavior could be exploitable as a denial of service.

I’m not sure what the right solution is for Integer, but clearly all the other cases should be optimized to produce 0 or Infinity very quickly.

[pcapriotti]

Are you sure you're using 7.4.2? This should be fixed, and indeed I can't reproduce it on 7.4.2. See #5688.

[andersk]

My mistake; I’m still on 7.4.1. Thanks.

[andersk]

Okay hold on, there’s still something wrong in 7.4.2 (I just downloaded ghc-7.4.2-x86_64-unknown-linux.tar.bz2).

> read ("1e" ++ show (2^63 - 2)) :: Double
Infinity  -- Fine.
> read ("1e" ++ show (2^63 - 1)) :: Double
0.0       -- What?
> read ("1e" ++ show (2^64 - 2^10 - 2)) :: Double
0.0       -- What?
> read ("1e" ++ show (2^64 - 2^10 - 1)) :: Double
          -- Memory-eating infinite loop!

[andersk]

Also it looks like GHC itself (also 7.4.2) is not using the improved read, as compiling x = 1e99999999 or typing 1e99999999 into GHCi uses lots of CPU and memory for several seconds, although this is of course a less critical problem.

[pcapriotti]

See #7034 and #5692.