- Camlp5 -

Version 6.00

• Introduction
• Transitional and Strict
• Parsing and printing tools

• Parsing tools
  • Stream parsers
  • Stream lexers
  • Functional parsers
  • Backtracking parsers
  • Extensible grammars
• Printing tools
  • Extensible printers
  • Pprintf
  • Pretty print
• Language extensions
  • Locations
  • Syntax tree
  • Syntax tree - transi
  • Syntax tree - strict
  • Quotation kit q_ast.cmo
  • The Pcaml module
  • Directives
  • Extensions of syntax
  • Extensions of printing
  • Redefining OCaml syntax
  • Quotations
  • Revised syntax
  • Scheme syntax
  • Macros
  • Pragma directive
  • Extensible functions
• Appendix
  • Commands and Files
  • Library
  • Camlp5 sources
  • About Camlp5

Syntax tree - transitional mode

This chapter presents the Camlp5 syntax tree when Camlp5 is installed in transitional mode.

Introduction

This syntax tree is defined in the module "MLast" provided by Camlp5. Each node corresponds to a syntactic entity of the corresponding type.

For example, the syntax tree of the statement "if" can be written:

  MLast.ExIfe loc e1 e2 e3

where "loc" is the location in the source, and "e1", "e2" and "e3" are respectively the expression after the "if", the one after the "then" and the one after the "else".

If a program needs to manipulate syntax trees, it can use the nodes defined in the module "MLast". The programmer must know how the concrete syntax is transformed into this abstract syntax.

A simpler solution is to use one of the quotation kit "q_MLast.cmo". It proposes quotations which represent the abstract syntax (the nodes of the module "MLast") into concrete syntax with antiquotations to bind variables inside. The example above can be written:

  <:expr< if $e1$ then $e2$ else $e3$ >>

This representation is very interesting when one wants to manipulate complicated syntax trees. Here is an example taken from the Camlp5 sources themselves:

  <:expr<
    match try Some $f$ with [ Stream.Failure -> None ] with
    [ Some $p$ -> $e$
    | _ -> raise (Stream.Error $e2$) ]
  >>

This example was in a position of a pattern. In abstract syntax, it should have been written:

  MLast.ExMat _
    (MLast.ExTry _ (MLast.ExApp _ (MLast.ExUid _ "Some") f)
       [(MLast.PaAcc _ (MLast.PaUid _ "Stream") (MLast.PaUid _ "Failure"),
         None, MLast.ExUid _ "None")])
    [(MLast.PaApp _ (MLast.PaUid _ "Some") p, None, e);
     (MLast.PaAny _, None,
      MLast.ExApp _ (MLast.ExLid _ "raise")
        (MLast.ExApp _
           (MLast.ExAcc _ (MLast.ExUid _ "Stream") (MLast.ExUid _ "Error"))
           e2))]

Which is less readable and much more complicated to build and update.

Instead of thinking of "a syntax tree", the programmer can think of "a piece of program".

Location

In all syntax tree nodes, the first parameter is the source location of the node.

In expressions

When a quotation is in the context of an expression, the location parameter is "loc" in the node and in all its possible sub-nodes. Example: if we consider the quotation:

  <:sig_item< value foo : int -> bool >>

This quotation, in a context of an expression, is equivalent to:

  MLast.SgVal loc "foo"
    (MLast.TyArr loc (MLast.TyLid loc "int") (MLast.TyLid loc "bool"));

The name "loc" is predefined. However, it is possible to change it, using the argument "-loc" of the Camlp5 shell commands.

Consequently, if there is no variable "loc" defined in the context of the quotation, or if it is not of the correct type, a semantic error occur in the OCaml compiler ("Unbound value loc").

Note that in the extensible grammars, the variable "loc" is bound, in all semantic actions, to the location of the rule.

If the created node has no location, the programmer can define a variable named "loc" equal to "Ploc.dummy".

In patterns

When a quotation is in the context of a pattern, the location parameter of all nodes and possible sub-nodes is set to the wildcard ("_"). The same example above:

  <:sig_item< value foo : int -> bool >>

is equivalent, in a pattern, to:

  MLast.SgVal _ "foo"
    (MLast.TyArr _ (MLast.TyLid _ "int") (MLast.TyLid _ "bool"))

However, it is possible to generate a binding of the variable "loc" on the top node by writing a "colon" before the "less" in the quotation. The same example:

  <:sig_item:< value foo : int -> bool >>

is equivalent to:

  MLast.SgVal loc "foo"
    (MLast.TyArr _ (MLast.TyLid _ "int") (MLast.TyLid _ "bool"))

Antiquotations

The expressions or patterns between dollar ($) characters are called antiquotations. In opposition to quotations which has its own syntax rules, the antiquotation is an area in the syntax of the enclosing context (expression or pattern). See the chapter about quotations.

If a quotation is in the context of an expression, the antiquotation must be an expression. It can be any expression, including function calls. Examples:

  value f e el = <:expr< [$e$ :: $loop False el$] >>;
  value patt_list p pl = <:patt< ( $list:[p::pl]$) >>;

If a quotation is in the context of an pattern, the antiquotation is a pattern. Any pattern is possible, including the wildcard character ("_"). Examples:

   fun [ <:expr< $lid:op$ $_$ $_$ >> -> op ]
   match p with [ <:patt< $_$ | $_$ >> -> Some p ]

Nodes and Quotations

This section describes all nodes defined in the module "MLast" of Camlp5 and how to write them with quotations. Notice that, inside quotations, one is not restricted to these elementary cases, but any complex value can be used, resulting on possibly complex combined nodes.

Variables names give information of their types:

• e, e1, e2, e3: expr
• p, p1, p2, p3: patt
• t, t1, t2, e3: ctyp
• s: string
• b: bool
• me, me1, me2: module_expr
• mt, mt1, mt2: module_type
• le: list expr
• lp: list patt
• lt: list ctyp
• ls: list string
• lse: list (string * expr)
• lpe: list (patt * expr)
• lpp: list (patt * patt)
• lpee: list (patt * option expr * expr)
• op: option patt
• lcsi: list class_str_item
• lcsi: list class_sig_item

expr

Expressions of the language.

  p -> e

  (p, None, e)

  p when e1 -> e

  (p, Some e1, e)

patt

Patterns of the language.

ctyp

Type expressions of the language.

modules...

str_item

Structure items, i.e. phrases in a ".ml" file or "struct" elements.

sig_item

Signature items, i.e. phrases in a ".mli" file or elements inside "sig ... end".

module_expr

module_type

classes...

class_expr

class_type

class_str_item

class_sig_item

other

type_decl

What is after 'type' or 'and' in a type declaration.

with_constr

"With" possibly following a module type.

poly_variant

Polymorphic variants.