[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Axiom-developer] Aldor rep and per: object-oriented programming or cate
[Axiom-developer] Aldor rep and per: object-oriented programming or category theory?
Thu, 1 Sep 2005 06:36:04 -0400
Dear Axiom enthusiasts and category theorists:
Having spent a few days trying to absorb the Aldor programming
philosophy from the point of view of it's evolution from previous
incarnations of the Axiom library compiler, as described here:
I find there is one construct in particular in Aldor that really
stands out in my mind - it's the concept of representation and
it's opposite (abstraction?). In Aldor these are formally denoted
by two very peculiar coercion operations: rep and pre.
In older SPAD terms, rep and per can be written as the following
rep x == x @ % pretend Rep;
per r == r @ Rep pretend %;
The expression rep(x) describes an object x of this domain but
treats it as belonging to a different domain Rep.
The expression per(y) describes an object y from the Rep domain
which is to be treated as belonging to this domain.
In looking through older SPAD code it is clear that these
combinations occur frequently and their encapsulation as rep
and pre makes reading Aldor code much easier than SPAD.
My question is: does anyone know of any formal programming
language research papers that describe the semantics of rep
and per in a general way?
If this sort of mechanism is present in other languages, what
form does it take?
It seems to me that rep and pre constitute the essence of the
peculiar object-oriented programming style pioneered by Axiom
and Aldor. But what are these constructs in formal (categorical)
Below are my first attempts to answer these questions. Your
comments, corrections and opinions would be most welcome.
It seems appropriate to me to think of rep and per as forming
a pair of adjoint functors:
% domain Rep domain
rep is a "forgetful functor" that maps the abstract structure
(objects and operations) of this domain into it's internal
representation. It "forgets" about the abstract relationships.
The objects of the external domain however are viewed only in
terms of the relationships between objects. The image of this
domain under rep in representation has its internal structure
per on the other hand constructs abstract members of this
domain from the underlying representation. It encapsulates
the internal operations of the representation in terms of
the abstract structure of this domain.
domains must be thought of in terms of categories both in the
sense of category theory and in Axiom's sense of "category".
The appropriate categories usually have at least the structure
of a Cartesian closed category which they inherit from
Axiom's basic types constructors, record, union and mapping.
In terms of category theory there is a natural transformation,
called the 'unit', that partitions elements of the Rep domain
into classes (as an abstract quotient of the representation).
x --> rep(per(x))
and rep is right adjoint of per.
In category theoretic terms Rep domain is some Cartesian closed
category, i.e. consisting of products (records), co-products
(unions) and exponentials (functions) over some basic set
of component domains. It's operations are formed freely from
the operations of it's components.
The implementation of the domain in terms of the representation
establishes a natural bi-jection between the operation of the
external domain and the operation of it's internal representation.
x -> rep y (representation)
per x -> y (domain)
The book: "Basic Category Theory for Computer Scientists", by
Benjamin Pierce, MIT Press, 1991, is a good reference. See
especially section 2.4. Pierce gives an excellent example in
terms of definition of the 'List' constructor.
Also: "Categories, Types and Structures. An introduction to Category
Theory for the working computer scientist." by Andrea Asperti and
Giuseppe Longo, M.I.T. Press, 1991, which is downloadable from here:
|[Prev in Thread]
||[Next in Thread]|
- [Axiom-developer] Aldor rep and per: object-oriented programming or category theory?,
Page, Bill <=