Without going too exotic, I see two choices. Keeping the single type Person makes what you are asking for, as far as I know, impossible (dependent types). For human readability you could consider type MensGroup = [Person] type WomensGroup = [Person] If you split Person into two types, Man and Woman, there are repercussions for the aggregate group type type Group = [Either Man Woman] -Mike On Tue, Dec 14, 2010 at 3:30 PM, Russ Abbott wrote:

Now that this is straightened out, I went back to what I was doing in the first place and realized that I haven't solved my problem. Given

data Person =       Man {name :: String, age :: Int, prostateCondition :: Condition}   | Woman {name :: String, age :: Int, ovaryCondition    :: Condition}

I'd like to define something like this.

type MensGroup = [Man]

Is there a way to do something like that? -- Russ

On Tue, Dec 14, 2010 at 1:06 PM, Russ Abbott wrote:

...

That's good. (It's more or less the way I was doing it.)  What I wanted to avoid was this.

getGenderSpecificCondition (  Man _ _ cond) = cond getGenderSpecificCondition (Woman _ _ cond) = cond

I know it seems like a small thing, but I would like to be able to write it like this.

getGenderSpecificCondition p    | p == (Man _ _ cond) = cond    | p == (Woman _ _ cond) = cond

But that's not legal syntax.  A pattern can't appear in that context. But this does the job.

getGenderSpecificCondition :: Person -> Condition getGenderSpecificCondition p    | isMan p = prostateCondition p    | isWoman p = ovaryCondition p

isMan (     Man _ _ cond) = True isMan _ = False isWoman (Woman _ _ cond) = True isWoman _ = False

That works! prostateCondition and ovaryCondition are both defined on Person. (I'm surprised to see that.)

*Person> Group [Man "Harry" 32 OK, Woman "Sally" 29 Good] Harry(32, OK) Sally(29, Good)

Also

*Person> prostateCondition (Woman "Sally" 29 Good) *** Exception: No match in record selector prostateCondition *Person> prostateCondition (Man "Harry" 29 Good) Good

-- Russ

On Tue, Dec 14, 2010 at 12:31 PM, Michael Katelman wrote:

...

Perhaps this?

https://gist.github.com/741048

-Mike

On Tue, Dec 14, 2010 at 2:27 PM, Russ Abbott wrote:

...

What I'm after is a version of my example that compiles.  Can you make one?

-- Russ

On Tue, Dec 14, 2010 at 12:18 PM, Antoine Latter wrote:

...

Sorry, I really don't know enough about what you're after to attempt that.

But you'll need to change you're signatures of the form:

...

function :: Person -> Foo

to something of the form:

...

function :: Person p => p -> Foo

Because again, a type class can not be used as a type.

Antoine

On Tue, Dec 14, 2010 at 2:12 PM, Russ Abbott wrote:

...

What got fouled up is all the adjustments I had to make to the other declarations. Can you complete the example so that it compiles using

class Person p where ...

I'd very much like to see an example that actually compiles.

Thanks. -- Russ

On Tue, Dec 14, 2010 at 11:58 AM, Antoine Latter wrote: > > On Tue, Dec 14, 2010 at 1:52 PM, Russ Abbott > > wrote: > > If gender is a field in a Person type, then a Person must have > > both > > an > > ovaryCondition and a prostateCondition.  That seems awkward. > > Regarding > >      class Person p where > > I started down that path but got completely fouled up. > > How did this get fouled up? Every class declaration must take > arguments - here, 'p' is the argument for the class. > > Thanks, > Antoine

_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners

I often find myself in a similar predicament, become frustrated, and after I've wasted a lot time ultimately decide that I'm making perfect the enemy of good. One possible option that hasn't been mentioned is to try and enforce safety at the module boundary, perhaps by making MensGroup a newtype and instead of exporting the constructor to export some other function that ensures only men wind up in a MensGroup. Sounds cumbersome to me, though. -Mike On Tue, Dec 14, 2010 at 3:53 PM, Russ Abbott wrote:

Ozgur pointed out this notition p@Man{}.  What I would like is something like this. type Man = Person@Man{} type MensGroup = [Man]

-- Russ

On Tue, Dec 14, 2010 at 1:45 PM, Michael Katelman wrote:

...

Without going too exotic, I see two choices. Keeping the single type Person makes what you are asking for, as far as I know, impossible (dependent types). For human readability you could consider

type MensGroup = [Person] type WomensGroup = [Person]

If you split Person into two types, Man and Woman, there are repercussions for the aggregate group type

type Group = [Either Man Woman]

-Mike

On Tue, Dec 14, 2010 at 3:30 PM, Russ Abbott wrote:

...

Now that this is straightened out, I went back to what I was doing in the first place and realized that I haven't solved my problem. Given

data Person =       Man {name :: String, age :: Int, prostateCondition :: Condition}   | Woman {name :: String, age :: Int, ovaryCondition    :: Condition}

I'd like to define something like this.

type MensGroup = [Man]

Is there a way to do something like that? -- Russ

On Tue, Dec 14, 2010 at 1:06 PM, Russ Abbott wrote:

...

That's good. (It's more or less the way I was doing it.)  What I wanted to avoid was this.

getGenderSpecificCondition (  Man _ _ cond) = cond getGenderSpecificCondition (Woman _ _ cond) = cond

I know it seems like a small thing, but I would like to be able to write it like this.

getGenderSpecificCondition p    | p == (Man _ _ cond) = cond    | p == (Woman _ _ cond) = cond

But that's not legal syntax.  A pattern can't appear in that context. But this does the job.

getGenderSpecificCondition :: Person -> Condition getGenderSpecificCondition p    | isMan p = prostateCondition p    | isWoman p = ovaryCondition p

isMan (     Man _ _ cond) = True isMan _ = False isWoman (Woman _ _ cond) = True isWoman _ = False

That works! prostateCondition and ovaryCondition are both defined on Person. (I'm surprised to see that.)

*Person> Group [Man "Harry" 32 OK, Woman "Sally" 29 Good] Harry(32, OK) Sally(29, Good)

Also

*Person> prostateCondition (Woman "Sally" 29 Good) *** Exception: No match in record selector prostateCondition *Person> prostateCondition (Man "Harry" 29 Good) Good

-- Russ

On Tue, Dec 14, 2010 at 12:31 PM, Michael Katelman wrote:

...

Perhaps this?

https://gist.github.com/741048

-Mike

On Tue, Dec 14, 2010 at 2:27 PM, Russ Abbott wrote:

...

What I'm after is a version of my example that compiles.  Can you make one?

-- Russ

On Tue, Dec 14, 2010 at 12:18 PM, Antoine Latter wrote: > > Sorry, I really don't know enough about what you're after to > attempt > that. > > But you'll need to change you're signatures of the form: > > > function :: Person -> Foo > > to something of the form: > > > function :: Person p => p -> Foo > > Because again, a type class can not be used as a type. > > Antoine > > On Tue, Dec 14, 2010 at 2:12 PM, Russ Abbott > > wrote: > > What got fouled up is all the adjustments I had to make to the > > other > > declarations. > > Can you complete the example so that it compiles using > > > > class Person p where ... > > > > I'd very much like to see an example that actually compiles. > > > > Thanks. > > -- Russ > > > > On Tue, Dec 14, 2010 at 11:58 AM, Antoine Latter > > > > wrote: > >> > >> On Tue, Dec 14, 2010 at 1:52 PM, Russ Abbott > >> > >> wrote: > >> > If gender is a field in a Person type, then a Person must have > >> > both > >> > an > >> > ovaryCondition and a prostateCondition.  That seems awkward. > >> > Regarding > >> >      class Person p where > >> > I started down that path but got completely fouled up. > >> > >> How did this get fouled up? Every class declaration must take > >> arguments - here, 'p' is the argument for the class. > >> > >> Thanks, > >> Antoine > > > >

_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners