good-old-32-bits

-- To make things interesting, I decided to assume a 32-bit machine.
-- Because invalid 32-bit numbers don't lie on nice boundaries, you 
-- have to be careful about doing validity checks before multiplying.
-- That is, you can't calculate out a value of 2147483648 and then
-- check if that's bigger than the 32-bit boundary (2147483647) because 
-- on a 32-bit machine, the value you'd be checking would be way negative.

toInt "1" => 1
toInt "-1" => -1

-- strings with too many characters are rejected (but there's more to the story)
toInt  "1000000000" => 1000000000 ## must fit within 2^32
toInt "10000000000" => error ## definitely does not fit

-- There are positive boundaries
toInt "2147483647" => 2147483647
toInt "2147483648" => error

-- negative boundaries
toInt "-2147483648" => -2147483648 ## one more valid negative than positive number
toInt "-2147483649" => error

-- can add an explicit plus sign (note this would screw up a naive length check)
toInt "+2147483647" => 2147483647

toInt "" -> error
toInt "-" -> error
toInt "+" -> error
toInt "--5" -> error
toInt "5a" -> error
toInt "a6" -> error
toInt " 6" -> error
toInt "6 " -> error