package refl
Install
Dune Dependency
Authors
Maintainers
Sources
sha512=e6e4f4f58c85aeb20db3adbcfcd66fcd1bba71b6a4cf032b85ae79505625bb2fc0c3fe536a9116f4c124a70563a0243bd2e13d0e6472e64d5c5370f67dfd0961
Description
PPX deriver for reflection
Published: 11 May 2020
README
refl
: Efficient and typesafe type reflection for OCaml
refl
provides runtime representations for most OCaml types, and a "one size fits all" deriving plugin. This plugin automatically derives runtime type representations from type declarations, allowing most other deriving plugins to be rewritten as regular OCaml functions parametized by these runtime representations. The library also provides such functions which provides flexible alternatives to ppx_deriving
standard plugins: show
, compare
, eq
, map
, iter
, fold
, enum
, make
.
Once refl
is derived for a given type, values of this type can be used on all (compatible) functions. This is the main motivation behind refl
: instead of having to decide which derivers to use at the type declaration point, it is sufficient to derive only refl
, and then this type can be used with all functions operating on such runtime representations, even functions that are defined after the type declaration.
refl
can be used with dune
by using the preprocess
field.
(executable
...
(preprocess (pps refl.ppx))
(libraries ... refl ...))
Basic usage
The following example declares a type for binary trees along with its runtime representation.
type 'a binary_tree =
 Leaf
 Node of { left : 'a binary_tree; label : 'a; right : 'a binary_tree }
[@@deriving refl]
The function Refl.show
can be used to print such trees, the function Refl.compare
can be used to compare two binary trees, the function Refl.map
can be used to change labels, etc. For instance:
# Refl.show [%refl: string binary_tree] []
(Node { left = Leaf; label = "root"; right = Leaf });;
 : string = "Node { left = Leaf; label = \"root\"; right = Leaf }"
Builtin types can be used directly without prior declaration.
# Refl.show [%refl: (string * int) list] [] ["a", 1; "b", 2];;
 : string = "[(\"a\", 1); (\"b\", 2)]"
When a type contains type variables, custom printers for these variables should be listed in the second argument of Refl.show
.
# Refl.show [%refl: _ list] [Some Format.pp_print_string] ["a"; "b"; "c"];;
 : string = "[a; b; c]"
The custom printers are optional for skipping "shadow" type variables, that is to say type variables that are not used to type subvalues. This will be particularly useful for GADTs. Typically, shadow type variables appear in GADTs as indexes. For instance, consider the following definition for fixedlength vector.
type ('a, 'length) vector =
 [] : ('a, [`Zero]) vector
 (::) : 'a * ('a, 'length) vector > ('a, [`Succ of 'length]) vector
[@@deriving refl]
When applying Refl.show
to vectors, None
can be passed for the printer of the second parameter ('length
), since it will never be used.
# Refl.show [%refl: (_, _) vector] [Some Format.pp_print_string; None]
["a"; "b"; "c"];;
 : string = "[a; b; c]"
Another example of function is Refl.map
, that applies a given function for each type variables.
# Refl.map [%refl: _ binary_tree] [%refl: _ binary_tree] [P string_of_int]
(Node { left = Leaf; label = 1; right = Leaf });;
 : string binary_tree = Node {left = Leaf; label = "1"; right = Leaf}
Note that the reflexive representation keeps track whether type variables occur positively or negatively. The P
constructor above marks functions appliable to type variables that appear positively (outside arrow types or at the righthand side of arrow types). For types that appear negatively, functions in the reverse direction should be provided, marked with N
. For types that appear both positively and negatively, a pair of functions for both directions should be provided, marked with PN
.
Type constraints
Functions can require constraints to be satisfied by the type to be applicable: for instance, the function compare
requires that the type contains no arrows unless they are hidden by an [@opaque]
attribute.
Dependencies (7)
Dev Dependencies
None
Used by (3)

clangmltransforms
= "0.25"

override
>= "0.4.0"

pattern
= "0.2.0"
Conflicts
None