Re: [Caml-list] Recursive record initialisation

[Gabriel Scherer <gabriel.scherer@gmail.com>]

The fact that the declaration is legal does not imply that you can
build useful values at this type. (It may be that trying to reject it
would naturally lead to reject other types that *are* more
interesting, for example.)

In this case you can build a value using value recursion, but it is
not very interesting:

# let rec a = { b }
       and b = { a };;
val a : atype = {b = {a = <cycle>}}
val b : btype = {a = {b = <cycle>}}

( Recursive value definitions:
http://caml.inria.fr/pub/docs/manual-ocaml/extn.html#s%3Aletrecvalues
)

Recursive value definitions are somewhat arcane and rarely used, but
still occasionally useful. One downside is that they prevent a purely
inductive interpretation of datatypes: if a recursive function is
structurally decreasing on one of its argument, there still exists
valid OCaml inputs on which it may not terminate. For example,
consider List.length on (let rec x = 1 :: 2 :: x).

On Mon, Apr 16, 2018 at 11:57 PM, Angelo Corsaro <angelo@icorsaro.net> wrote:
> Hello,
>
> I have a question concerning the initialisation of mutually recursive record
> types. As a trivial example consider the following type declaration:
>
> type atype = { b : btype; }
> and btype = { a : atype; }
>
> This is a legal type declaration (it type-checks w/o a blink), but I can’t
> see how such a type would be initialised? As both “a" and “b” are values
> there is an apparent bootstrapping issue. But as this is a legal type
> declaration I guess there is a way around, wondering what the right way is.
>
> Thanks very much in advance for any clarification.
>
> —Angelo
>
>
>
>
> —
> “Simplicity is the ultimate sophistication." ~ Leonardo da Vinci
>

-- 
Caml-list mailing list.  Subscription management and archives:
https://sympa.inria.fr/sympa/arc/caml-list
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
Bug reports: http://caml.inria.fr/bin/caml-bugs