Hi all,
Discussion over this issue has died down. Yitzchak and others have
made good arguments for the type signatures
readHistory :: String -> IO Bool
writeHistory :: String -> IO Bool
which reflect the non-exceptional nature of failure of those functions.
Does anyone have any problems with implementing the following patch
from Alex (along with the above change)?
http://www.haskell.org/pipermail/libraries/2008-January/009007.html
For reference, this is in relation to the following proposal:
http://hackage.haskell.org/trac/ghc/ticket/2053
If no objections are raised by next week (say, Feb. 27), we should
make the above changes to readline, which will let us implement #2050
(persistent history file in ghci).
Thanks,
-Judah
On Wed, Feb 6, 2008 at 1:09 PM, Yitzchak Gale
Judah Jacobson wrote:
- readFile... writeFile...
I think of readHistory and writeHistory as analogues to the above functions.
I don't think so. I don't really care about the file itself - that just happens to be the way readline implements its persistence. Non-existence of the file is the normal way that readline represents the fact that there isn't any history yet. The semantics of the API call are: load history, if any exists yet, and report the result. Nothing exceptional about that.
All that being said, this is a relatively minor issue
Agreed. All the more so with the MonadIO problem now factored out.
Regards, Yitz