Pavel G. Zhbanov wrote:
If it doesn't have a side effect, why do it anyway? The result 'c' does not depend on a.
myFunc uses IORef and it's (IORef's) result I use afterwards in some other functions.
OK: what is myFunc's type? If it ends in IO alpha, for some alpha, you can say: do a <- myFunc ... etc. If it doesn't, then you should probably re-think its definition.
myFunc :: a -> [b] (a and b are my own types)
Actually, inside myFunc I used unsafePerformIO (didn't want to change the whole programm just because of one function). The purpose of myFunc is to append some value to "some list" lying somewhere (somewhere is "defined" by IORef), store the resulting list and return a copy.
There's a reason why the name unsafePerformIO begins with "unsafe". It
is *not* a magic wand that can simply get rid of the "IO" whenever it
turns out to be inconvenient.
You need to change myFunc's type to:
myFunc :: a -> IO [b]
then use:
do a <- myFunc ...
--
Glynn Clements