Chapter 8 Language extensions
This chapter describes language extensions and convenience features that are implemented in OCaml, but not described in the OCaml reference manual.
12 Attributes
(Introduced in OCaml 4.02, infix notations for constructs other than expressions added in 4.03)
Attributes are “decorations” of the syntax tree which are mostly ignored by the type-checker but can be used by external tools. An attribute is made of an identifier and a payload, which can be a structure, a type expression (prefixed with :), a signature (prefixed with :) or a pattern (prefixed with ?) optionally followed by a when clause:
|
The first form of attributes is attached with a postfix notation on “algebraic” categories:
|
This form of attributes can also be inserted after the `tag-name in polymorphic variant type expressions (tag-spec-first, tag-spec, tag-spec-full) or after the method-name in method-type.
The same syntactic form is also used to attach attributes to labels and constructors in type declarations:
|
Note: when a label declaration is followed by a semi-colon, attributes can also be put after the semi-colon (in which case they are merged to those specified before).
The second form of attributes are attached to “blocks” such as type declarations, class fields, etc:
A third form of attributes appears as stand-alone structure or signature items in the module or class sub-languages. They are not attached to any specific node in the syntax tree:
|
(Note: contrary to what the grammar above describes, item-attributes cannot be attached to these floating attributes in class-field-spec and class-field.)
It is also possible to specify attributes using an infix syntax. For instance:
let[@foo] x = 2 in x + 1 === (let x = 2 [@@foo] in x + 1) begin[@foo][@bar x] ... end === (begin ... end)[@foo][@@bar x] module[@foo] M = ... === module M = ... [@@foo] type[@foo] t = T === type t = T [@@foo] method[@foo] m = ... === method m = ... [@@foo]
For let, the attributes are applied to each bindings:
let[@foo] x = 2 and y = 3 in x + y === (let x = 2 [@@foo] and y = 3 in x + y) let[@foo] x = 2 and[@bar] y = 3 in x + y === (let x = 2 [@@foo] and y = 3 [@bar] in x + y)
12.1 Built-in attributes
Some attributes are understood by the type-checker:
- “ocaml.warning” or “warning”, with a string literal payload. This can be used as floating attributes in a signature/structure/object/object type. The string is parsed and has the same effect as the -w command-line option, in the scope between the attribute and the end of the current signature/structure/object/object type. The attribute can also be attached to any kind of syntactic item which support attributes (such as an expression, or a type expression) in which case its scope is limited to that item. Note that it is not well-defined which scope is used for a specific warning. This is implementation dependent and can change between versions. Some warnings are even completely outside the control of “ocaml.warning” (for instance, warnings 1, 2, 14, 29 and 50).
- “ocaml.warnerror” or “warnerror”, with a string literal payload. Same as “ocaml.warning”, for the -warn-error command-line option.
- “ocaml.alert” or “alert”: see section 8.21.
- “ocaml.deprecated” or “deprecated”: alias for the “deprecated” alert, see section 8.21.
- “ocaml.deprecated_mutable” or “deprecated_mutable”. Can be applied to a mutable record label. If the label is later used to modify the field (with “expr.l <- expr”), the “deprecated” alert will be triggered. If the payload of the attribute is a string literal, the alert message includes this text.
- “ocaml.ppwarning” or “ppwarning”, in any context, with a string literal payload. The text is reported as warning (22) by the compiler (currently, the warning location is the location of the string payload). This is mostly useful for preprocessors which need to communicate warnings to the user. This could also be used to mark explicitly some code location for further inspection.
- “ocaml.warn_on_literal_pattern” or “warn_on_literal_pattern” annotate constructors in type definition. A warning (52) is then emitted when this constructor is pattern matched with a constant literal as argument. This attribute denotes constructors whose argument is purely informative and may change in the future. Therefore, pattern matching on this argument with a constant literal is unreliable. For instance, all built-in exception constructors are marked as “warn_on_literal_pattern”. Note that, due to an implementation limitation, this warning (52) is only triggered for single argument constructor.
- “ocaml.tailcall” or “tailcall” can be applied to function application in order to check that the call is tailcall optimized. If it it not the case, a warning (51) is emitted.
- “ocaml.inline” or “inline” take either “never”, “always” or nothing as payload on a function or functor definition. If no payload is provided, the default value is “always”. This payload controls when applications of the annotated functions should be inlined.
- “ocaml.inlined” or “inlined” can be applied to any function or functor application to check that the call is inlined by the compiler. If the call is not inlined, a warning (55) is emitted.
- “ocaml.noalloc”, “ocaml.unboxed”and “ocaml.untagged” or “noalloc”, “unboxed” and “untagged” can be used on external definitions to obtain finer control over the C-to-OCaml interface. See 20.11 for more details.
- “ocaml.immediate” or “immediate” applied on an abstract type mark the type as having a non-pointer implementation (e.g. “int”, “bool”, “char” or enumerated types). Mutation of these immediate types does not activate the garbage collector’s write barrier, which can significantly boost performance in programs relying heavily on mutable state.
- “ocaml.immediate64” or “immediate64” applied on an abstract type mark the type as having a non-pointer implementation on 64 bit platforms. No assumption is made on other platforms. In order to produce a type with the “immediate64“ attribute, one must use “Sys.Immediate64.Make“ functor.
- ocaml.unboxed or unboxed can be used on a type definition if the type is a single-field record or a concrete type with a single constructor that has a single argument. It tells the compiler to optimize the representation of the type by removing the block that represents the record or the constructor (i.e. a value of this type is physically equal to its argument). In the case of GADTs, an additional restriction applies: the argument must not be an existential variable, represented by an existential type variable, or an abstract type constructor applied to an existential type variable.
- ocaml.boxed or boxed can be used on type definitions to mean the opposite of ocaml.unboxed: keep the unoptimized representation of the type. When there is no annotation, the default is currently boxed but it may change in the future.
- ocaml.local or local take either never, always, maybe or nothing as payload on a function definition. If no payload is provided, the default is always. The attribute controls an optimization which consists in compiling a function into a static continuation. Contrary to inlining, this optimization does not duplicate the function’s body. This is possible when all references to the function are full applications, all sharing the same continuation (for instance, the returned value of several branches of a pattern matching). never disables the optimization, always asserts that the optimization applies (otherwise a warning 55 is emitted) and maybe lets the optimization apply when possible (this is the default behavior when the attribute is not specified). The optimization is implicitly disabled when using the bytecode compiler in debug mode (-g), and for functions marked with an ocaml.inline always or ocaml.unrolled attribute which supersede ocaml.local.