PPrint is an OCaml library for pretty-printing textual documents. It takes care of indentation and line breaks, and is typically used to pretty-print code.

API Reference

An experienced user may wish to jump directly to a section of the API documentation:

Core Combinators

At the heart of PPrint is a little domain-specific language of documents. This language has a well-defined semantics, which the printing engine implements. The language and its semantics rest upon a small number of fundamental concepts.

There are combinators for creating atomic documents. For instance, the string combinator turns an OCaml string (which must not contain any newline character) into a document. Thus,

string "hello"

is a simple, unbreakable document. The utf8string combinator is analogous, and should be preferred when working with non-ASCII strings. The utf8format combinator provides a convenient sprintf-style API for constructing a complex string and turning it into an atomic document.

There is a concatenation operator (^^), which joins two documents. For instance,

string "hello" ^^ string "world"

is a composite document. It is in fact equivalent to string "helloworld".

A somewhat more interesting combinator is the breakable blank combinator break. This combinator expects a nonnegative integer argument, the width of the desired breakable blank. If break n is printed in flat mode, it produces n blank characters; if it is printed in normal mode, it produces one newline character.

As suggested by the previous sentence, there are two printing modes, namely flat mode and normal mode. The printing engine goes back and forth between these two modes. Exactly where and how the printing engine switches from one mode to the other is controlled by the next combinator.

The grouping combinator, group, introduces a choice between flat mode and normal mode. It is a document transformer: if d is a document, then group d is a document. When the printing engine encounters group d, two possibilities arise. The first possibility is to print all of d on a single line. This is known as flat mode. The engine tries this first (ignoring all group combinators inside d). If it succeeds, great. If it fails, by lack of space on the current line, then the engine reverts to the second possibility, which is to dissolve the group and print the bare document d in normal mode. This has subtle consequences: there might be further groups inside d, and each of these groups gives rise to further choices.

At each group, the choice is resolved in an efficient way. No backtracking is required. The ideal width of every document is computed (in a bottom-up manner) when documents are constructed. This allows every choice to be resolved in constant time. The time complexity of building and rendering documents is linear in the size of the document.


The interplay of break and group gives rise to an interesting language, where group is used to indicate a choice point, and the appearance of break is dependent upon the choice points that appear higher up in the hierarchical structure of the document. For instance, the document:

group (string "This" ^^ break 1 ^^ string "is" ^^ break 1 ^^ string "pretty.")

is printed either on a single line, if it fits, or on three lines. It cannot be printed on two lines: there is just one choice point, so either the two breakable blanks are broken, or none of them is. By the way, this document can be abbreviated as follows:

group (string "This" ^/^ string "is" ^/^ string "pretty.")

On the other hand, the document:

string "This" ^^
group (break 1 ^^ string "is") ^^
group (break 1 ^^ string "pretty.")

can be printed on one, two, or three lines. There are two choice points, each of which influences one of the two breakable blanks. The two choices are independent of one another. Each of the words in the sentence This is pretty. is printed on the current line if it fits, and on a new line otherwise. By the way, this document can be abbreviated as follows:

flow (break 1) [
  string "This";
  string "is";
  string "pretty."

or as follows:

flow_map (break 1) string [ "This"; "is"; "pretty." ]

More Core Combinators

As noted earlier, the string that is supplied to string, utf8string, or utf8format must not contain any newline characters. If one wishes to impose a line break, one must use the forced newline combinator hardline.

Whereas group introduces a choice between flat mode and normal mode, the conditional construct ifflat allows testing whether the printing engine is currently in flat mode or in normal mode. The document ifflat doc1 doc2 is rendered as doc1 if the engine is currently in flat mode, and as doc2 if the engine is currently in normal mode.

The blank combinator blank is analogous to break, but produces non-breakable blank characters. A blank character is like an ordinary ASCII space character string " ", except that blank characters at the end of a line are automatically suppressed. Thus, the printing engine guarantees that no trailing blank characters are ever produced.

To illustrate the power of these combinators, let us reveal that break is in reality not a primitive combinator: it is defined in terms of hardline, blank, and ifflat. A possible definition of break 1 is ifflat (blank 1) hardline.

The nesting combinator nest deals with indentation. At every time, the printing engine maintains a current indentation level, which is a nonnegative integer. The current indentation level is initially zero. To render the document nest 2 d, the printing engine temporarily increases the current indentation level by 2, renders the document d, then restores the previous indentation level. The effect of the current indentation level is as follows: every time a newline character is emitted, it is immediately followed by n blank characters, where n is the current indentation level.

To illustrate the use of indentation, let us look at this document:

group (
  string "begin" ^^
  nest 2 (break 1 ^^ string "work") ^^
  break 1 ^^ string "end"

Although this document looks somewhat complicated, understanding its behavior is relatively easy, because there is only one group combinator in it. This document can be printed in one of two ways. If it fits on the current line, then the content of the group is rendered in flat mode: break 1 becomes equivalent to blank 1, and (because no newline characters are emitted) nest 2 has no effect. The document is then rendered as follows:

begin work end

If the document does not fit on the current line, then the group is dissolved, and break 1 becomes equivalent to hardline. Thus, the document becomes equivalent to:

string "begin" ^^
nest 2 (hardline ^^ string "work") ^^
hardline ^^ string "end"

Thanks to the nest combinator, the first hardline is immediately followed with two blank characters, whereas the second hardline is not. The document is then rendered as follows:


The alignment combinator align can be used to change the current indentation level in a more subtle way. The effect of this combinator is to set the current indentation level to the current column. To understand what this means, let us look at this document:

string "please" ^/^ align (group (string "align" ^/^ string "here"))

If this document fits on the current line, then neither align nor group have any effect, so the document is rendered as follows:

please align here

If the document does not fit on the current line, then the group is dissolved. The second concatenation operator (^/^) inserts a breakable blank break 1, which is in this case is equivalent to hardline. Because the current indentation level is set by align to the column that follows "please ", the document is rendered as follows:

please align

This concludes our review of PPrint's core combinators. Not every combinator has been mentioned here; for further details, please consult the complete list of the core combinators for building documents.

On top of the core combinators, it is up to the user of the library to define higher-level combinators that are more convenient or better suited to a particular use case. PPrint itself comes with a collection of high-level combinators, and the submodule PPrint.OCaml offers a collection of combinators for printing OCaml values. These collections are not as complete and thoughtfully designed as they could be. They are subject to change in the future.

Rendering Documents

The submodules PPrint.ToChannel, PPrint.ToBuffer, and PPrint.ToFormatter give access to the printing engine, and send their output respectively to an output channel of type out_channel, to a buffer of type Buffer.t, and to a formatter channel of type Format.formatter.

Each of these submodules offers a choice between two printing engines. The pretty printing engine should be preferred in most situations; it attempts to respects the maximum line width and ribbon width specified by the user. The compact printing engine can be used when the readability of the output does not matter: it assumes a maximum line width of zero (so it never flattens a group) and does not emit any indentation characters.

Defining Custom Documents

It is possible to extend PPrint with custom document constructors, provided they meet the expectations of the printing engine. In short, the custom document combinator custom expects an object of class custom. This object must provide three methods. The method requirement must compute the ideal width of the custom document. The methods pretty and compact must render the custom document. For this purpose, they have access to the output channel and to the state of the printing engine. For more details, see Defining Custom Documents.

History and Acknowledgements

The document language and the printing engine are inspired by Daan Leijen's wl-pprint library, which itself is based on the ideas developed by Philip Wadler in the paper A Prettier Printer. This Haskell library exploits laziness to achieve a very low memory requirement: the entire document never needs to reside in memory. PPrint achieves greater simplicity and possibly higher throughput by requiring the entire document to be built in memory before it is printed.

PPrint was written by François Pottier and Nicolas Pouillard, with contributions by Yann Régis-Gianas, Gabriel Scherer, Jonathan Protzenko, and Thomas Refis.