Probably in the same way integers are made redundant : data Integer = 0 | 1 | 2 | .... Cheers, Thu 2009/12/8 Lyndon Maydwell :

Aren't symbols made redundant by algebraic data types?

On Tue, Dec 8, 2009 at 4:48 PM, Michael Vanier wrote:

...

jean-christophe mincke wrote:

...

Hello,

Has there already been attempts to introduce lisp like symbols in haskell?

Thank you

Regards

J-C

J-C,

Do you mean symbols as in "interned strings with an O(1) string comparison method"? I would love to have those, but I don't see an easy way to get it without being in the IO or ST monad.

Mike

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Vladimir Zlatanov wrote:

I think stable names can be used where symbols are used in scheme. I think there are better alternatives for different use cases in haskell.

For example, symbols are often used to tag values - haskell has pattern matching on constructors.

Here is a link: http://www.haskell.org/ghc/docs/latest/html/libraries/base/System-Mem-Stable... Cheers Ben

On Tue, Dec 8, 2009 at 1:48 AM, Michael Vanier wrote:

Do you mean symbols as in "interned strings with an O(1) string comparison method"?  I would love to have those, but I don't see an easy way to get it without being in the IO or ST monad.

data Sym = Sym String Hash fromString :: String -> Sym fromString s = Sym s (hash s) instance Eq Sym where Sym _ h == Sym _ h' = h == h' Much as I am uncomfortable with hash-based equality. 1/2^256, despite being very small, is not zero. It is begging for a mysterious failure someday. Luke

I think lisp like symbols could be quite useful in the context of embedded DSL to create ... well... symbols that can be interpreted as variables in that DSL. I can imagine something such as a small relational DSL i.e Without symbols we have several alternatives: 1. Symbols are written as strings (and possibly automatically lifted to Symbol string if necessary, It is possible for nums I do not know for strings). Moreover the system should be able to tell the difference between a string representing a symbol and a string representing a string value "Table_A" 'where' "Att_Name" = "Smith" 'and' "Att_Age" > 10 2. Using algebraic data type: the syntax looks better but the data type must be declared before. In this case we must know in advance all the attributes and table names (including all the names used in 'rename' relational opération - the 'as' in sql). Table_A 'where' Att_Name = "Smith" 'and' Att_Age > 10 3 If we had symbols. Note lisp like symbols might well be values of some predefined type "Symbol". Just like 2,3, 56 are values of the predefined type Integer. Table_A 'where' Att_Name = "Smith" 'and' Att_Age > 10 Here the problem is that the compiler should be able to tell the difference between a symbol and an undefined variable. Lisp solves it by prefixing the symbol with a little '. Smalltalk prefixes it with #. So with unambiguous symbols: 'Table_A 'where' 'Att_Name = "Smith" 'and' 'Att_Age > 10 or #Table_A 'where' #Att_Name = "Smith" 'and' #Att_Age > 10 Regards J-C