Version française
Home     About     Download     Resources     Contact us    
Browse thread
Feature proposal: improved compare
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: Tom <tom.primozic@g...>
Subject: Feature proposal: improved compare
I believe the compare function from standard library should be extended to
allow total ordering of functional types (= closures). I suggest something
like...

let old_compare = compare;;

let rec compare a b =
      let ar = Obj.repr a in
      let br = Obj.repr b in
      if Obj.tag ar = Obj.closure_tag && Obj.tag br = Obj.closure_tag then
              (* Field 0 of a closure is a pointer of the function code.
Continue if pointers match *)
          let d = (-) (Obj.obj (Obj.field ar 0)) (Obj.obj (Obj.field br 0))
in
          if d <> 0 then d else
              (* now match every other field of the closures - these are the
arguments to partially applied functions *)
          let rec f x = if x = Obj.size ar then 0 else
              let d = (-) (Obj.obj (Obj.field ar x)) (Obj.obj (Obj.field br
x)) in
              if d <> 0 then d else f (x+1)
          in
          f 0
        (* if the two values are not closures, call the old compare. This in
fact is incorrect behaviour, as old compare will fail if it meets
(different) closure values. All the match cases in old compare should be
included in the new compare if the behaviour is to be correct. *)
      else old_compare a b;;



This way, one could use for example association lists in order to store
functional values and use List.assoc with them. The improvement over the old
compare is this:

# old_compare (f 0) (f 0);;
Exception: Invalid_argument "equal: functional value".

# compare (f 0) (f 0);;
- : int = 0

- Tom