Installation

Just do the customary make && sudo make install. This only needs flex and a standards-compliant C++ compiler.

Usage

First, see the description of the RST format. RST is a very simple markup format, almost like plain text (in fact, you're looking at RST right now, this document is written in it!). You can learn enough of it to start marking up your source in about five minutes.

Second, you'll have to mark up your source comments. pure-doc recognizes comments in RST format by looking at the first non-empty line of the comment. A comment (either /* ... */ or a contiguous sequence of // line comments) is assumed to contain RST format if the first non-empty line starts with :, .. or __. Other comments are taken to be plain text and are ignored by pure-doc.

Notes:

• pure-doc makes no other assumption about the contents of marked up comments, so you can include whatever you want: titles, section headers, fields, admonitions, plain text, whatever. Just make sure that the comment starts with one of the special tokens listed above. (You can always put just .. at the beginning of the comment to force it to be recognized, this will be treated as a comment by the docutils tools.)
• Also, pure-doc makes very few assumptions about the source; in fact, any source files with a C/C++-like comment and string syntax should work. So you could also use it to document your C/C++ programs, or even plain text files like this one, as long as they adhere to these standards.
• Indentation in extracted comments is preserved (assuming tabs = 8 spaces by default, you can change this with the -t option). This is important because indentation conveys document structure in RST.

For instance, here is a sample RST-formatted comment:

/* :Name: ``rand`` - compute random numbers
   :Synopsis: ``rand``
   :Description: Computes a (pseudo) random number. Takes no parameters.
   :Example: Here is how you can call ``rand`` in Pure:
     ::

       > extern int rand();
       > rand;
       1804289383

   :See Also: rand(3) */

This will be rendered as follows:

Name:	rand - compute random numbers
Synopsis:	rand
Description:	Computes a (pseudo) random number. Takes no parameters.
Example:	Here is how you can call rand in Pure: > extern int rand(); > rand; 1804289383
See Also:	rand(3)

Finally, to extract the documentation you run pure-doc on your source files as follows:

pure-doc source-files ...

If no input files are specfied then the source is read from standard input. Otherwise all input files are read and processed in the indicated order. The output is written to stdout, so that you can directly pipe it into one of the docutils programs:

pure-doc source-files ... | rst2html

If you prefer to write the output to a file, you can do that as follows:

pure-doc source-files ... > rst-file

pure-doc also understands the following two options. These must come before any file arguments.

-h	Print a short help message.
-twidth	Set the tab width to the given number of spaces.

There are no other options. By its design pure-doc is just a plain simple "docstring scraping" utility with no formatting knowledge of its own. All actual formatting is handled by the docutils programs which offer plenty of options to change the appearance of the generated output; please refer to the docutils documentation for details.

If docutils isn't enough for you, then you should consider one of the other available RST-based documentation frameworks, such as Sphinx. These provide even more elaborate formatting options, but are usually somewhat Python-centric and harder to use. Also, you should take into account that many Pure users read documentation in a text-based browser while running the interpreter, so including all the latest html gizmos is not going to do them much good and might in fact impair readability.

Literate Programming

pure-doc also recognizes literate code delimited by comments which, besides the comment delimiters and whitespace, contain nothing but the special start and end "tags" >>> and <<<. Code between these delimiters (including all comments) is extracted from the source and output as a RST literal code block.

For instance:

/* ..

   pure-doc supports literate programming, too. */

// >>>

// This is a literate comment.
/* .. This too! */

extern int rand();
rand;

// <<<

This will be rendered as follows:

pure-doc supports literate programming, too.

// This is a literate comment.
/* .. This too! */

extern int rand();
rand;

Try it now! You can scrape all the sample "documentation" from this file and format it as html, as follows:

pure-doc README | rst2html --no-doc-title --no-doc-info > test.html

As already mentioned, this document itself is also in RST format, so you can also do the following to get nicely formatted documentation:

rst2html README pure-doc.html

(You can also just type make html to get this.)

Hyperlink Targets and Index Generation

To supplement the normal hyperlink target processing by the docutils tools, pure-doc recognizes explicit hyperlink targets of the form:

.. _target:

pure-doc automatically creates raw html targets (<a name=...>) for those. This works around the docutils name mangling (which is undesirable if you're indexing, say, function names). It also resolves a quirk with w3m which doesn't pick up all id attributes in the docutils-generated html source.

In addition, you can also have pure-doc generate an index from all explicit targets. To these ends, just add the following special directive at the place where you want the index to appear:

.. makeindex::

The directive will be replaced with a bullet list of references to all targets collected up to that point, sorted alphabetically. This also resets the list of collected targets, so that you can have multiple smaller indices in your document instead of one big one.

It goes without saying that this facility is rather simplistic, and can't compete with a carefully hand-made index. But we decided to toss it in because docutils currently doesn't provide an automatic index facility of its own, and the simple kind of index generated with pure-doc is certainly better than having no index at all. (An alternative to pure-doc's indexing facility would be to introduce a special RST "text role" for indexing purposes and use a custom utility to process this kind of markup. But at present this is beyond pure-doc's scope, so you're on your own if you take that route.)

Generating Documentation

If you're generating some library documentation for which you have to process a bigger collection of source files, then it is often convenient to have a few Makefile rules to automatize the process. To these ends, simply add rules similar to the following to your Makefile:

# The sources. Order matters here. The generated documentation will have the
# comments from each source file in the indicated order.
sources = foo.pure bar.pure

# The basename of the documentation files to be generated.
target = foo

.PHONY: html tex pdf

html: $(target).html
tex: $(target).tex
pdf: $(target).pdf

$(target).txt: $(sources)
        pure-doc $(sources) > $@

# This requires that you have docutils installed.

%.html: %.txt
        rst2html $< $@

%.tex: %.txt
        rst2latex $< $@

# This also requires that you have TeX installed.

%.pdf: %.tex
        pdflatex $<
        rm -f *.aux *.log *.out

clean:
        rm -f *.html *.tex *.pdf

(This assumes GNU make. If you use some other make utility then you will have to adapt the pattern rules accordingly.)

Now you can just type make or make html to generate the documentation in html format, and make tex or make pdf to generate the other formats. The clean target removes the generated files.

Installing Documentation

Now that you generated the documentation for your Pure library, you might want to make the Pure interpreter know about it, so that you can read your documentation with the help command of the interpreter. This isn't difficult either. You just install the html files on the Pure library path. The following Makefile rule automatizes this process. Add this to the Makefile in the previous section:

# Try to guess the installation prefix (this needs GNU make):
prefix = $(patsubst %/bin/pure,%,$(shell which pure 2>/dev/null))
ifeq ($(strip $(prefix)),)
# Fall back to /usr/local.
prefix = /usr/local
endif

libdir = $(prefix)/lib
datadir = $(libdir)/pure

install:
        test -d "$(DESTDIR)$(datadir)" || mkdir -p "$(DESTDIR)$(datadir)"
        cp $(target).html "$(DESTDIR)$(datadir)"

After a make install your documentation should now end up in the Pure library directory, and you can read it in the Pure interpreter using a command like the following:

> help foo#

For a concrete example, you can find all this and more in pure-doc's own Makefile. In fact, after installation this document ends up as pure-doc.html in the Pure library directory, so that you can read it with:

> help pure-doc#

Note the hash character. This tells the help command that this is an auxiliary documentation file, rather than a function name to be looked up in the Pure Library Manual. You can also look up a specific section in the pure-doc manual as follows:

> help pure-doc#literate-programming

Please also refer to the Pure Manual for more information on how to use the interpreter's online help.