Generic makefile for document processing with docutils

This file provides typical rules for processing documents with docutils. By default it provides common targets (all, pdf, clean, etc.) for processing all *.txt files in this directory.

You should be able to customize just about everything by setting variables defined here (they all have defaults and are documented in comments). You can set them on make's command line or by building your own makefile that sets them before including this one. For simplest cases, you'd want to give DOCS and perhaps ALL.

Some things can be customized per document with make's per-target/per-pattern variable assignment syntax, e.g. FLAGS += --no-doc-title. Unfortunately this won't work for any variables that affect rules sources/targets (in particular, extension variables and variables that enable pre/postprocessing).

If you want deeper customization or to integrate this into your makefile(s) for other things, you can make it very include-friendly by flipping a few settings:

This file should be self-documenting; read it file for the full list of customizable variables and other gory details...

GNU Make is required! Some targets use common unix utils.

The latest version of this file lives at


All variables defined in this file can be optionally prefixed with anything you want (e.g. DOCUTILS.) to avoid namespace pollution. To enable this, set the variable docutils. (note the period in the variable name) to the desired prefix. Note that then, you would have to use the prefix when setting any variables that affect this file...

docutils. ?=
##docutils. ?= DOCUTILS.

Pattern rules - very include-friendly makefile portion

This part is designed to be included in makefiles. It supplies implicit pattern rules for docutils' tools. All that is left to you is to request building of files with extensions of the output formats (.html, .tex, .dvi, .ps, .pdf). You should have (or be able to build) input files with .txt extension. Almost everything (including extensions) is configurable by setting variables defined in this file. You should set these variables before including this file to avoid problems.

If you also want typical explicit rules for targets like html, dvi and clean to build associated files, you will find them in later sections...

Tool names

TOOLS lists the tools / processing stages that are supported and have [at least some] associated variables according to the conventions below. It provides some introspection ablity, currently used in the docutils.clean target for computing the list of output extensions.

                     MATHHACK IMGMATHHACK                               \

As accustomed in makefiles, tool names can be customized by setting variables named after the tools. Here they are slightly more systemathic than their current names. is expected to be renamed to soon.

ifneq "$(shell which" ""
$(docutils.)RST2HTML ?=
$(docutils.)RST2HTML ?=

$(docutils.)RST2LATEX ?= is the RST PEP -> HTML converter tool; also supports old PEP format and does extra PEP services.

$(docutils.)RSTPEP2HTML ?=

$(docutils.)PEPTOOL ?=

In this makefile I use only PEPTOOL (too bad RSTPEP2HTML doesn't support the old pep format, it has a better command-line interface).

More such variables appear in the pre/postprocessing sections below.

@@@ TODO: Support other tools, e.g. RST2XML.


For simplicity, only one source extension is supported at a time. If you need to allow several (e.g. .txt and .rst), you should be able to call make recursively several times, setting SRCEXT to a different value each time...

$(docutils.)SRCEXT ?= .txt

There are various reasons to preprocess/postprocess docutils. This makefile directly supports some ways and should play nice with others. It's more convenient to use intermediate files than pipes (because then you can easily inspect them for debugging). So we need a way to insert intermediate files into the dependency chain. Solution: variables (<toolname>.SRCEXT and <toolname>.TRGEXT) that control the source/target extensions of each rule. These variables must be set before the rules that use them.

$(docutils.)PEPTOOL.SRCEXT ?= $($(docutils.)SRCEXT)
$(docutils.)PEPTOOL.TRGEXT ?= .html
$(docutils.)RST2HTML.SRCEXT ?= $($(docutils.)SRCEXT)
$(docutils.)RST2HTML.TRGEXT ?= .html
$(docutils.)RST2LATEX.SRCEXT ?= $($(docutils.)SRCEXT)
$(docutils.)RST2LATEX.TRGEXT ?= .tex

PDFLaTeX might need different processing and extension (in particular because no image format is supported by both it and LaTeX). This is sort of a hack: pretend we have a separate RST2PDFLATEX tool.

$(docutils.)RST2PDFLATEX.SRCEXT ?= $($(docutils.)RST2LATEX.SRCEXT)
$(docutils.)RST2PDFLATEX.TRGEXT ?= .pdftex

Specific preprocessing/postprocessing support

The extension manipulation is subtle. It's important to use simple variables (set with :=) to allow chaining. Since each processing stage can be enabled independently of others, they can only be attached to the main tool (e.g. RST2HTML).

Note that PEPTOOL is not supported here because non-standard processing is not a good idea in PEPs anyway. Won't be a big problem to add if needed...

@@@ TODO: Support as many external/sandbox tools as possible.

sandbox/cben/rolehack math preprocessing hacks

This makefile provides rules that support preprocessing by ../rolehack/ and ../rolehack/ Set ENABLE_MATHHACK to 1 to run them before and

$(docutils.)ENABLE_MATHHACK ?= 0

$(docutils.)MATHHACK.SRCEXT := $($(docutils.)RST2LATEX.SRCEXT)
ifeq "$(origin $(docutils.)MATHHACK.TRGEXT)" "undefined"
$(docutils.)MATHHACK.TRGEXT := $($(docutils.)RST2LATEX.SRCEXT).mathhack

In theory, IMGMATHHACK is applicable to all tools. It's not very useful with anything but RST2HTML, so that's the only one we attach to currently...

$(docutils.)IMGMATHHACK.SRCEXT := $($(docutils.)RST2HTML.SRCEXT)
ifeq "$(origin $(docutils.)IMGMATHHACK.TRGEXT)" "undefined"
$(docutils.)IMGMATHHACK.TRGEXT := $($(docutils.)RST2HTML.SRCEXT).imgmathhack

ifeq "$($(docutils.)ENABLE_MATHHACK)" "1"
$(docutils.)RST2LATEX.SRCEXT := $($(docutils.)MATHHACK.TRGEXT)
$(docutils.)RST2HTML.SRCEXT := $($(docutils.)IMGMATHHACK.TRGEXT)

If any preprocessing has been applied, it's good to tell docutils the name of the original file. But in current docutils, --source-url forces on --source-link. So you can control this option, per tool.

ifeq "$($(docutils.)RST2LATEX.ENABLE_SOURCE_URL)" "1"
$(docutils.)RST2LATEX.FLAGS += --source-url=$*$($(docutils.)SRCEXT)

ifeq "$($(docutils.)RST2HTML.ENABLE_SOURCE_URL)" "1"
$(docutils.)RST2HTML.FLAGS += --source-url=$*$($(docutils.)SRCEXT)

You can set MATHHACK_DIR to provide an explicit location for these scripts (you must include a trailing slash!). You can also change the tool name variables completely...

$(docutils.)MATHHACK_DIR ?=
$(docutils.)MATHHACK ?= $($(docutils.)MATHHACK_DIR)
$(docutils.)IMGMATHHACK ?= $($(docutils.)MATHHACK_DIR)

%$($(docutils.)MATHHACK.TRGEXT): %$($(docutils.)MATHHACK.SRCEXT)
        $($(docutils.)MATHHACK) $< $@

%$($(docutils.)IMGMATHHACK.TRGEXT): %$($(docutils.)IMGMATHHACK.SRCEXT)
        $($(docutils.)IMGMATHHACK) $< $@
Rudimentary adaptation of .* file extensions to output format

Docutils currently doesn't support referencing different files depending on on the output format (e.g. LaTeX requires EPS images but HTML supports everything else). Here are simple sed scripts, converting .* to one extension for images and one for links, different for each tool. Be warned that the regexps are not bullet-proof and could be confused!

Set ENABLE_WILD_EXTS to 1 if you want to enable them.

$(docutils.)ENABLE_WILD_EXTS ?= 0

This is an arbitrary guess, quite probable that you would want to change it.

$(docutils.)HTML_WILD_EXTS.IMGEXT ?= .png

.htm is also widespread but here we mostly talk about local files produced with docutils.

$(docutils.)HTML_WILD_EXTS.LINKEXT ?= .html

LaTeX only undestands Encapsulated PostScript.

$(docutils.)LATEX_WILD_EXTS.IMGEXT ?= .eps

LaTeX is just as well convetible to PS but only PDF supports links, so this is a good guess...

$(docutils.)LATEX_WILD_EXTS.LINKEXT ?= .pdf

PDFLaTeX understands common formats like PNG but doesn't understand EPS!

$(docutils.)PDFLATEX_WILD_EXTS.IMGEXT ?= .png

Again, PDFLaTeX supports links.

$(docutils.)PDFLATEX_WILD_EXTS.LINKEXT ?= .pdf

Intermediate extensions:

$(docutils.)HTML_WILD_EXTS.SRCEXT ?= .html.wild_exts
$(docutils.)LATEX_WILD_EXTS.SRCEXT ?= .tex.wild_exts
$(docutils.)PDFLATEX_WILD_EXTS.SRCEXT ?= .pdftex.wild_exts

Chain to the tools.

$(docutils.)HTML_WILD_EXTS.TRGEXT := $($(docutils.)RST2HTML.TRGEXT)
$(docutils.)LATEX_WILD_EXTS.TRGEXT := $($(docutils.)RST2LATEX.TRGEXT)

ifeq "$($(docutils.)ENABLE_WILD_EXTS)" "1"
$(docutils.)RST2HTML.TRGEXT := $($(docutils.)HTML_WILD_EXTS.SRCEXT)
$(docutils.)RST2LATEX.TRGEXT := $($(docutils.)LATEX_WILD_EXTS.SRCEXT)

%$($(docutils.)HTML_WILD_EXTS.TRGEXT): %$($(docutils.)HTML_WILD_EXTS.SRCEXT)
        sed -e 's/\( [sS][rR][cC]="[^"]*\)\.\*"/\1$($(docutils.)HTML_WILD_EXTS.IMGEXT)"/g' -e "s/\( [sS][rR][cC]='[^']*\)\.\*'/\1$($(docutils.)HTML_WILD_EXTS.IMGEXT)'/g" -e 's/\( [hH][rR][eE][fF]="[^"]*\)\.\*"/\1$($(docutils.)HTML_WILD_EXTS.LINKEXT)"/g' -e "s/\( [hH][rR][eE][fF]='[^']*\)\.\*'/\1$($(docutils.)HTML_WILD_EXTS.LINKEXT)'/g" $< > $@

%$($(docutils.)LATEX_WILD_EXTS.TRGEXT): %$($(docutils.)LATEX_WILD_EXTS.SRCEXT)
        sed -e 's/\(\\includegraphics{[^}]*\)\.\*}/\1$($(docutils.)LATEX_WILD_EXTS.IMGEXT)}/g' -e 's/\(\\href{[^}]*\)\.\*}/\1$($(docutils.)LATEX_WILD_EXTS.LINKEXT)}/g' $< > $@

        sed -e 's/\(\\includegraphics{[^}]*\)\.\*}/\1$($(docutils.)PDFLATEX_WILD_EXTS.IMGEXT)}/g' -e 's/\(\\href{[^}]*\)\.\*}/\1$($(docutils.)PDFLATEX_WILD_EXTS.LINKEXT)}/g' $< > $@

Config files and extra flags

Consider using config files instead of flags for most tasks. To specify config file(s) other than the default, set CONFIG to a space-separated list of them. Later files override earlier ones (and all override the DOCUTILSCONFIG environment variable, which you should not touch in a Makefile, it's for the user). Non-existant config files are ignored.

Relative file names in CONFIG are relative to the current directory, independently of the target name. If you use % in CONFIG, it will be replaced with the target name (without extension) and it will be interpretted relative to the target's directory -- giving you per-file configs.

$(docutils.)CONFIG ?=

I found absolutely no way to implement in make per-directory config file dependencies that don't depend on the target basename. You can just do:

$(docutils.)FLAGS += $(*D)/

but it won't be considered a dependency (which is not such a big issue). Or you can always run make in the directory of the target...

You can provide extra docutils options by setting variables named <toolname>.FLAGS. FLAGS gives default flags for all tools.

$(docutils.)FLAGS += $(foreach c,$($(docutils.)CONFIG),--config=$(subst %,$(*D)/,$(findstring %,$(c)))$(subst %,$(*F),$(c)))
$(docutils.)RST2HTML.FLAGS += $($(docutils.)FLAGS)
$(docutils.)RST2LATEX.FLAGS += $($(docutils.)FLAGS)
$(docutils.)PEPTOOL.FLAGS += $($(docutils.)FLAGS)

Extra dependencies

When docutils config files change, it usually affects docutils' so it should be rebuilt. So we add them as dependencies.


Deletion of a config file will not cause a rebuild.

Figure out the default config files from the environment:

DOCUTILSCONFIG ?= /etc/docutils.conf:./docutils.conf:$(HOME)/.docutils

You can set CONFIG_DEPS to override the depencies on config files without changing the actual config files used by docutils. It's space-separated.

$(docutils.)CONFIG_DEPS ?= $(subst :, ,$(DOCUTILSCONFIG)) $($(docutils.)CONFIG)

EXTRA_DEPS specifies the actual extra dependencies.

Config files that exist are automatically included (unexistant ones would cause make to complain that it doesn't know how to build them).

$(docutuls.)EXTRA_DEPS += $(wildcard $($(docutils.)CONFIG_DEPS))

Note also that you can supply extra dependencies for pattern rules by yourself, as long as you don't provide commands, e.g.

foo.html: extra_deps

Typical extra depencies you'd want to add: HTML stylesheets (if you embed them), files appearing in .. include::.

HTML output

pep-%.html must be before the %.html rule to get higher priority.

pep-%$($(docutils.)PEPTOOL.TRGEXT): pep-%$($(docutils.)PEPTOOL.SRCEXT) \
        $($(docutils.)PEPTOOL) $($(docutils.)PEPTOOL.FLAGS) $<
ifneq "$($(docutils.)PEPTOOL.TRGEXT)" ".html"
        # $($(docutils.)PEPTOOL) lacks output file name control, work around.
        mv $(basename $<).html $<

%$($(docutils.)RST2HTML.TRGEXT): %$($(docutils.)RST2HTML.SRCEXT) \
        $($(docutils.)RST2HTML) $($(docutils.)RST2HTML.FLAGS) $< $@

LaTeX output

%$($(docutils.)RST2LATEX.TRGEXT): %$($(docutils.)RST2LATEX.SRCEXT) \
        $($(docutils.)RST2LATEX) $($(docutils.)RST2LATEX.FLAGS) $< $@

LaTeX processing rules

This part is not fully parametrized; it's not related to docutils so if you want to get fancy in this respect, provide your own rules. You can disable the rules here, by setting ENABLE_LATEX_RULES to 0.

$(docutils.)ENABLE_LATEX_RULES ?= 1

PDFLaTeX does a better job than LaTeX->DVIPDF (e.g. it supports hyperlinks and generates a PDF outline) but you usually can't run the same file through both LaTeX and PDFLaTeX because they don't understand a single common format. Solution: there is a separate preprocessing stage PDFLATEX_WILD_EXTS. Set this to 0 to go through DVIPDF.

$(docutils.)ENABLE_PDFLATEX ?= 1

ifeq "$($(docutils.)ENABLE_LATEX_RULES)" "1"

These variables are common-practice but it's still cleaner to put them in the "namespace".

$(docutils.)LATEX ?= latex
$(docutils.)PDFLATEX ?= pdflatex
$(docutils.)DVIPS ?= dvips
$(docutils.)DVIPDF ?= dvipdf

%.dvi: %.tex
        cd $(*D); $(LATEX) $(*F) %.dvi
        $(DVIPS) $< -o $@

ifeq "$($(docutils.)ENABLE_PDFLATEX)" "1"
%.pdf: %.pdftex
        cd $(*D); $(PDFLATEX) $(*F).pdftex
%.pdf: %.dvi
        $(DVIPDF) $< $@


Fallback when you don't provide for different PDFLaTeX processing.

%$($(docutils.)RST2PDFLATEX.TRGEXT): %$($(docutils.)RST2LATEX.TRGEXT)
        cp $< $@

Typical standard targets

This section provides typical rules for processing documents with docutils. Since they might clash with your rules, they are optional and can be disabled by setting ENABLE_COMMON_TARGETS to 0.

$(docutils.)ENABLE_COMMON_TARGETS ?= 1

DOCS gives the documents to process. You should typically set this (defaults to all files with the source extension in current directory).

$(docutils.)DOCS ?= $(wildcard *$($(docutils.)SRCEXT))

STEMS is the docs list without the source extension, (makes rules shorter).

$(docutils.)STEMS ?= $($(docutils.)DOCS:$($(docutils.)SRCEXT)=)

Set ALL to change the things to build by default.

$(docutils.)ALL ?= html ps pdf

ifeq "$($(docutils.)ENABLE_COMMON_TARGETS)" "1"

docutils.all: $($(docutils.)ALL)

Targets like clean weren't directly used. Rather they don't have commands but have targets like docutils.clean (that do have commands) as prerequisites. This allows you to give your commands for them. You can inhibit these alias dependencies by setting ENABLE_SHORT_TARGETS to 0.

$(docutils.)ENABLE_SHORT_TARGETS ?= 1

ifeq "$($(docutils.)ENABLE_SHORT_TARGETS)" "1"
all: docutils.all
safeclean: docutils.safeclean
mostlyclean: docutils.mostlyclean
clean: docutils.clean

html: $($(docutils.)STEMS:=.html)
tex: $($(docutils.)STEMS:=.tex)
dvi: $($(docutils.)STEMS:=.dvi)
ps: $($(docutils.)
pdf: $($(docutils.)STEMS:=.pdf)

Extensions of unneeded byproducts (specifically, LaTeX byproducts).

$(docutils.)SAFECLEAN_EXTS += .log .aux .out .toc

Extensions that are direct outputs of this makefile.

$(docutils.)CLEAN_EXTS += $(foreach tool,$($(docutils.)TOOLS),$($(docutils.)$(tool).TRGEXT))
$(docutils.)CLEAN_EXTS += .dvi .ps .pdf

safeclean cleans only unneeded byproducts.

        -rm -f $(foreach ext,$($(docutils.)SAFECLEAN_EXTS),$($(docutils.)STEMS:=$(ext)))

mostlyclean cleans almost everything.

docutils.mostlyclean: docutils.safeclean
        -rm -f $(foreach ext,$($(docutils.)CLEAN_EXTS),$($(docutils.)STEMS:=$(ext)))

clean also cleans imgmathhack images that take a lot of time to rebuild.

docutils.clean: docutils.mostlyclean
ifeq "$($(docutils.)ENABLE_MATHHACK)" "1"
        -rm -rf ./imgmath/