Text.Regex.Posix exposes too little

[Simon Marlow]

On 11 February 2005 21:24, Peter Simons wrote:

> I was wondering why Text.Regex.Posix does not expose the Ptr
> interface to the regular expression library? Is there a
> deeper reason for that?

No deep reason that I'm aware of.

> I am also worried about the fact that the module uses
> finalizers to free-up unused expressions -- rather than
> providing a bracket-style
> 
>   withRegex :: String -> Int -> (Regex -> IO a) -> IO a
> 
> function.

We quite often use it with a regex defined at the top-level:

r = mkRegex "..."

f x = case matchRegex r x of ...

Although in this case there is limited benefit from having the regex
garbage collected, since there can only be a bounded, probably small,
number of these in the program.

Providing withRegex seems like a good idea, though.

> Last but not least I see that the regex flags are
> represented as an 'Int'. I don't feel strongly about this,
> but IMHO a
> 
>   data RegexFlag = IgnoreCase
>                  | Newline
>                  | ...
> 
> data type with an appropriate declaration for Enum would be
> nicer for this purpose.
> 
> Unless there is violent opposition, I'd volunteer to make
> those changes, because as it happens I need them anyway. ;-)

Please go ahead, I'll review and commit your patch.

Cheers,
	Simon