civicaml/util.mli

(** Utility functions used by multiple phases. *)

open Types

(** {2 Generating variables} *)

(** Generate a new identifier from a given base, e.g. ["foo"] becomes
    ["_foo_1"]. Uses an [int] reference defined in the implementation file as a
    counter to assert uniqueness of the generated variable. *)
val fresh_id : string -> string

(** Generate a new constant (a variable that is known to be assigned only once
    in total) from a given id, e.g. ["foo"] becomes ["_foo_1_"]. Uses the same
    counter as {!fresh_id}. *)
val fresh_const : string -> string

(** Check if an identifier is generated by the compiler. *)
val is_generated_id : string -> bool

(** Check if an identifier is a constant generated by the compiler. *)
val is_const_id : string -> bool

(** Generate an identifier from a base and a number. E.g., [generate_id "foo"
    1] returns ["_foo_1"]*)
val generate_id : string -> int -> string

(** {2 AST traversal} *)

(** Default transformation traversal for AST nodes of arbitrary constructor:
    [traverse unit_type fold_results visit_node node]. For each constructor that
    has one or more children of type {!Types.node}, the children are transformed
    with the given transformation function. A [(node, result)] tuple is
    returned, where the secondary result has the type of the given unit type.
    [visit_node] must also return such a tuple. [fold_results] is used to reduce
    results of multiple children to a single result.

    For [Program] nodes, {!flatten_blocks} is called on the resulting
    declaration list. This allows the transformation function to return multiple
    nodes as a replacement, in the form of a [Block] node with multiple
    children. *)
val traverse : 'a -> ('a -> 'a -> 'a) -> (node -> (node * 'a)) -> node ->
               (node * 'a)

(** Wrapper for {!traverse} that ignores the secondary traversal result. Returns
    {!Types.node} instead of a tuple. *)
val traverse_unit : (node -> node) -> node -> node

(** Wrapper for {!traverse} for list results. Defined as:
    [let traverse_list visit = traverse [] (@) visit] *)
val traverse_list : (node -> (node * 'a list)) -> node -> (node * 'a list)

(** Flatten [Block] nodes into containing node lists.
    E.g., [[A; Block [B; Block [C; D]]; E] -> [A; B; C; D; E]].
    Traverses into elements of the list recursively first, so this is
    essentially a traversal, as can be seen in the example. Therefore the best
    practice is to call this function once on the children of a [Program] node
    after any traversal that may generate [Block] nodes that need to be
    flattened. *)
val flatten_blocks : node list -> node list

(** Extract the list of child nodes from a [Block] node. *)
val block_body : node -> node list

(** {2 AST node annotations} *)

(** Add a single annotation to a node. *)
val annotate : annotation -> node -> node

(** Extract annotations from a node of arbitrary constructor. *)
val annof : node -> annotation list

(** Get the value of the [Loc] annotation of a node, or {!noloc} if no location
    can be found. *)
val locof : node -> location

(** Empty node location: [("", 0, 0, 0, 0)]. Used then location of a node is
    unknown (if the annotation was removed at some point) or
    non-existant/irrelevant (for generated nodes on which no errors will occur --
    hopefully...). *)
val noloc : location

(** Get the value of the [Depth] annotation of a node. Raises
    {!Types.InvalidNode} if the annotation can not be found. *)
val depthof : node -> int

(** Get the value of the [Index] annotation of a node. Raises
    {!Types.InvalidNode} if the annotation can not be found. *)
val indexof : node -> int

(** Get the value of the [Type] annotation of a node. Some node types
    do not need to be annotated since they have inherent types. For example, a
    [VarDec] node has a type attribute, and a [Dim] node is always an [Int]
    (because array dimensions are integers). All nodes which have inherent types
    are [VarDec], [Param], [FunDec], [FunDef], [GlobalDec], [GlobalDef],
    [TypeCast], and [Dim]. Raises a {!Types.InvalidNode} if the annotation can
    not be found, and the type has no inherent type. *)
val typeof : node -> ctype

(** Get the value of the [LabelName] annotation of a node. Raises
    {!Types.InvalidNode} if the annotation can not be found. *)
val labelof : node -> string

(** Get the basic type of a declaration, removing array dimensions *)
val basetypeof : node -> ctype

(** Get the value of the name attribute from a variable or function declaration
    (similar to {!typeof} for nodes that have types attributes). Raises
    {!Types.InvalidNode} if the node is not one of [GlobalDec], [GlobalDef],
    [FunDec], [FunDef], [VarDec], [Param], or [Dim]. *)
val nameof : node -> string

(** Get the CiviC data type of a constant value. *)
val const_type : const -> ctype

(** Check if a constant value is eligible for creating optimized assembly
    instructions. E.g. [Intval (-1)] is eligible because the instruction
    [iloadc_m1] exists. Used to decide on which {!Types.instr} constructor to
    use during the assembly phases. This function always returns [false] when
    optimizations are disabled (when {!Globals.args}[.optimize = false]). *)
val is_immediate_const : const -> bool

(** Check if a node has an array type. I.e., [Array] or [ArrayDims]. *)
val is_array : node -> bool

(** {2 Logging} *)

(** Horizontal line of ['-'] characters, used to separate output sections. *)
val hline : string

(** Print the stringification of a node to [stderr] (uses
    {!Stringify.node2str}). *)
val prt_node : node -> unit

(** Output a line to stderr if the verbosity level in {!Globals.args} is at
    least as high as the specified verbosity level. The line is indented with a
    number of spaces to match the longest phase identifier (so that logged lines
    align with ideitifiers logged by {!Main.main}). A newline is added
    automatically. *)
val log_line : int -> string -> unit

(** Print a line to [stderr] without indent (but do add a newline). *)
val log_plain_line : int -> string -> unit

(** Same as {!log_line}, but prints a node stringification instead of a literal
    string. *)
val log_node : int -> node -> unit

(** Generate an location tuple from Lexing data structures *)
val loc_from_lexpos : Lexing.position -> Lexing.position -> location

(** Print location tuple to stderr. Produces
    something like the following: {v
int foo;
    ^^^ v}
    The location tuple is likely to originate from a node annotation, extracted
    using {!locof}.*)
val prerr_loc : location -> unit

(** Print location tuple to stderr, along with an error message. Produces
    something like the following: {v
File "foo.cvc", line 10, characters 8-11:
<message>
    int foo;
        ^^^ v} *)
val prerr_loc_msg : location -> string -> unit

(** Print an error message for a node. Calls {!prerr_loc_msg}. *)
val node_error : node -> string -> unit

(** Print a warning message for a node. Calls {!prerr_loc_msg}. *)
val node_warning : node -> string -> unit

(** Print an error message of type {!Types.error_msg}. *)
val print_error : error_msg -> unit

(** Raise a {!Types.FatalError} if the given error list is not empty, and
    print the errors before quitting. *)
val quit_on_error : node -> error_msg list -> node

(** {2 String utilities} *)

(** [repeat s n] returns a new string of [n] times [s]. *)
val repeat : string -> int -> string

(** [expand n s] adds spaces to [s] until the resulting string is at least [n]
    characters long. *)
val expand : int -> string -> string

(** {2 List utilities} *)

(** [optmap f opt] maps [f] to the list value of [opt] if [opt] exists, and
    [None] otherwise. *)
val optmap : ('a -> 'b) -> 'a list option -> 'b list option

(** Same as {!optmap}, but returns the list value instead, or an empty list if
    [opt] is [None]. *)
val optmapl : ('a -> 'b) -> 'a list option -> 'b list

(** [List.mapi] clone (only available in OCaml version >= 4.00. Maps a function
    to a list like [List.map] does, but the iterator function is called with the
    element's index as an additional argument. *)
val mapi : (int -> 'a -> 'b) -> 'a list -> 'b list