From: Joseph Young <ocaml@optimojoe.com>
To: caml-list@inria.fr
Subject: Re: [Caml-list] Conditionals based on phantom types
Date: Tue, 3 Aug 2010 08:15:28 +0200 (CEST) [thread overview]
Message-ID: <Pine.LNX.4.64.1008030812270.4144@myhome> (raw)
In-Reply-To: <Pine.LNX.4.64.1008030441350.1541@myhome>
[-- Attachment #1: Type: TEXT/PLAIN, Size: 4294 bytes --]
I had someone point out that
add (`Meters, 0.5) (`Feet, 0.5);;
will type check by instantiating 'a as [`Feet | `Meters]. Hiding the type
in a module may fix the example:
type units=[`Feet | `Meters];;
let unit_to_string=function
| `Feet -> "ft"
| `Meters -> "m"
;;
module Units : sig
type t
val to_feet : float -> [`Feet]*t
val to_meters : float -> [`Meters]*t
val add : ([<units]*t as 'a) -> 'a -> 'a
val print : [<units]*t -> unit
end = struct
type t=float
let to_feet x=`Feet,x;;
let to_meters x=`Meters,x;;
let add (u,x) (_,y) = u,x +. y;;
let print (u,x)=Printf.printf "%f (%s)" x (unit_to_string u);;
end
Then, expressions as the one above can not be written.
Joe
On Tue, 3 Aug 2010, Joseph Young wrote:
> I appreciate the reply and the suggestions. After looking over your
> responses, could everything be simplified by removing the phantom types? In
> other words, is there a draw back to using the definition:
>
> type units=[`Feet | `Meters];;
> let to_feet x : [`Feet]*float=`Feet,x;;
> let to_meters x : [`Meters]*float=`Meters,x;;
> let add (u,x:'a*float) (_,y:'a*float) = u,x +. y;;
> let unit_to_string=function
> | `Feet -> "ft"
> | `Meters -> "m"
> ;;
> let print (u,x:[< units]*float) = Printf.printf "%f (%s)" x
> (unit_to_string u);;
>
> which instead of phantom types relies on polymorphic variants that contain
> only a single value?
>
> Joe
>
> On Mon, 2 Aug 2010, bluestorm wrote:
>
>> Two remarks on Lukasz suggestion :
>>
>> > val add : 'a t -> 'a t -> 'a t
>> > [..]
>> > let add (_,x) (_,y) = x +. y
>>
>> This does not typecheck. I suggest the following :
>>
>> let add (ux, x) (uy, y) =
>> assert (ux = uy);
>> (ux, x +. y)
>>
>> While the assertion does not seem necessary at first (correct units
>> are guaranteed by typing !), it may be helpful in case of bug inside
>> the Units module or signature, wich breaks the typing invariant. If
>> you're planning to do relatively elaborate things inside the Units, I
>> strongly recommend to use any kind of dynamic checking available, at
>> least during development. This is something is understood late in my
>> own phantom-type project (Macaque), and would have been very useful
>> for debugging.
>>
>>
>>
>> On Mon, Aug 2, 2010 at 9:49 AM, Lukasz Stafiniak <lukstafi@gmail.com>
>> wrote:
>> > val print : 'a t -> unit
>> > [..]
>> > type 'a t = 'a * float
>> > let print (u,x) = Printf.printf "%f %s" x (to_string u)
>>
>> Lukasz doesn't give a to_string function. Assuming this one, there is
>> a typing problem here.
>>
>> let to_string = function
>> | `Feet -> "feet"
>> | `Meters -> "meters"
>>
>> Values do not match:
>> val print : [< `Feet | `Meters ] * float -> unit
>> is not included in
>> val print : 'a t -> unit
>>
>> The issue is that, with Lukasz definition, 'a is now coupled to the
>> concrete values (type 'a t = 'a * float), and the to_string function
>> is *not* polymorphic in 'a as advertised by the interface. I see two
>> solutions :
>>
>> 1) restrict print to only print the units you directly support :
>>
>> val print : [ `Feet | `Meters ] * float -> unit
>>
>> 2) make 'a a phantom type parameter again by decoupling the type
>> information (the polymoprhic variant) and the runtime information
>> (another, value-level, variant) :
>>
>> odule Units : sig
>> type 'a t
>> val to_feet : float -> [`Feet ] t
>> val to_meters : float -> [`Meters] t
>> val add : 'a t -> 'a t -> 'a t
>> val print : 'a t -> unit
>> end = struct
>> type unit = Feet | Meters
>> let string_of_unit = function
>> | Feet -> "feet"
>> | Meters -> "meters"
>>
>> type 'a t = unit * float
>>
>> let to_feet x = Feet, x
>> let to_meters x = Meters, x
>>
>> let add (ux, x) (uy, y) =
>> assert (ux = uy);
>> (ux, x +. y)
>>
>> let print (u, x) =
>> Printf.printf "%f (%s)" x (string_of_unit u)
>> end;;
>>
>> I would rather go the second way, wich allows for more flexibility in
>> what runtime informations keep, and what type information you expose.
>>
>
next prev parent reply other threads:[~2010-08-03 6:39 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-08-02 7:07 Joseph Young
2010-08-02 7:49 ` [Caml-list] " Lukasz Stafiniak
2010-08-02 8:02 ` bluestorm
2010-08-03 2:46 ` Joseph Young
2010-08-03 6:15 ` Joseph Young [this message]
2010-08-03 6:57 ` bluestorm
2010-08-04 2:41 ` Joseph Young
2010-08-03 15:47 ` Goswin von Brederlow
2010-08-03 15:39 ` Goswin von Brederlow
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Pine.LNX.4.64.1008030812270.4144@myhome \
--to=ocaml@optimojoe.com \
--cc=caml-list@inria.fr \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox