Legend:
Library
Module
Module type
Parameter
Class
Class type
Utility functions for pretty-printing. The major features provided by this module are
An fprintf-style interface with support for user-defined printers
The printout is fit to a width by selecting some of the optional newlines
Constructs for alignment and indentation
Print ellipsis starting at a certain nesting depth
Constructs for printing lists and arrays
Pretty-printing occurs in two stages:
Construct a Pretty.doc object that encodes all of the elements to be printed along with alignment specifiers and optional and mandatory newlines
Format the Pretty.doc to a certain width and emit it as a string, to an output stream or pass it to a user-defined function
The formatting algorithm is not optimal but it does a pretty good job while still operating in linear time. The original version was based on a pretty printer by Philip Wadler which turned out to not scale to large jobs.
API
type doc
The type of unformated documents. Elements of this type can be constructed in two ways. Either with a number of constructor shown below, or using the Pretty.dprintf function with a printf-like interface. The Pretty.dprintf method is slightly slower so we do not use it for large jobs such as the output routines for a compiler. But we use it for small jobs such as logging and error messages.
A document that consists of a mandatory newline. This is just like (text
"\n"). The new line will be indented to the current indentation level, unless you use Pretty.leftflush right after this.
A document that consists of either a space or a line break. Also called an optional line break. Such a break will be taken only if necessary to fit the document in a given width. If the break is not taken a space is printed instead.
Mark the current column as the current indentation level. Does not print anything. All taken line breaks will align to this column. The previous alignment level is saved on a stack.
Prints a document as markup. The marked document cannot contain line breaks or alignment constructs.
val seq : sep:doc->doit:('a->doc)->elements:'a list->doc
Formats a sequence. sep is a separator, doit is a function that converts an element to a document.
val docList : ?sep:doc->('a->doc)->unit ->'a list->doc
An alternative function for printing a list. The unit argument is there to make this function more easily usable with the Pretty.dprintf interface. The first argument is a separator, by default a comma.
val d_list : string ->(unit ->'a->doc)->unit ->'a list->doc
sm: Yet another list printer. This one accepts the same kind of printing function that Pretty.dprintf does, and itself works in the dprintf context. Also accepts a string as the separator since that's by far the most common.
val docArray : ?sep:doc->(int ->'a->doc)->unit ->'a array->doc
Formats an array. A separator and a function that prints an array element. The default separator is a comma.
This function provides an alternative method for constructing doc objects. The first argument for this function is a format string argument (of type ('a, unit, doc) format; if you insist on understanding what that means see the module Printf). The format string is like that for the printf function in C, except that it understands a few more formatting controls, all starting with the @ character.
See the gprintf function if you want to pipe the result of dprintf into some other functions.
The following special formatting characters are understood (these do not correspond to arguments of the function):
@^ Inserts a Pretty.leftflush Should be used immediately after @! or "\n".
@@ : inserts a @ character
In addition to the usual printf % formatting characters the following two new characters are supported:
%t Corresponds to an argument of type unit -> doc. This argument is invoked to produce a document
%a Corresponds to two arguments. The first of type unit -> 'a -> doc and the second of type 'a. (The extra unit is do to the peculiarities of the built-in support for format strings in Ocaml. It turns out that it is not a major problem.) Here is an example of how you use this:
val gprintf : (doc->'b)->('a, unit, doc, 'b)format4->'a
Like Pretty.dprintf but more general. It also takes a function that is invoked on the constructed document but before any formatting is done. The type of the format argument means that 'a is the type of the parameters of this function, unit is the type of the first argument to %a and %t formats, doc is the type of the intermediate result, and 'b is the type of the result of gprintf.
If set to true then optional breaks are taken only when the document has exceeded the given width. This means that the printout will looked more ragged but it will be faster