\documentclass[12pt,a4paper]{article}

\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}

\usepackage[english]{babel}

\usepackage[margin=1in]{geometry}

\usepackage[svgnames]{xcolor}
\usepackage[colorlinks, linkcolor={blue}, urlcolor={blue}]{hyperref}

\usepackage{tcolorbox}
\tcbuselibrary{listings}

\newcommand{\checkcites}{\texttt{checkcites}}
\newcommand{\email}[1]{\small\texttt{#1}}
\newcommand{\version}{Version 1.0h from March 16, 2012.}

\newenvironment{infoblock}[1]
  {\par\addvspace{\medskipamount}
   \begin{tcolorbox}[colframe=DarkTurquoise,coltitle=black,fonttitle=\bfseries,title=#1]}
  {\end{tcolorbox}\addvspace{\medskipamount}}

\newenvironment{terminal}
  {\par\addvspace{\medskipamount}
   \begin{tcolorbox}[colframe=DarkTurquoise]}
  {\end{tcolorbox}\par\addvspace{\medskipamount}}

\title{The \checkcites\footnote{\version}\ \ script}
\author{%
  Enrico Gregorio\\\email{Enrico dot Gregorio at univr dot it}\\[3ex]
  Paulo Roberto Massa Cereda\\\email{cereda at users dot sf dot net}%
}
\date{}

\begin{document}

\maketitle

\tableofcontents

\section{Introduction}
\label{sec:intro}

\checkcites\ is a Lua script written for the sole purpose of detecting
unused or undefined references from both \LaTeX{} auxiliary or
bibliography files. We use \emph{unused reference} to refer to the
reference present the bibliography file --~with the \verb|.bib|
extension~-- but not cited in the \verb|.tex| file. The
\emph{undefined reference} is exactly the opposite, that is, the items
cited in the \texttt{.tex} file, but not present in the \verb|.bib|
file.

The original idea came from a question posted in the
\href{http://tex.stackexchange.com}{TeX.sx community} about
\href{http://tex.stackexchange.com/questions/43276}{how to check which
bibliography entries were not used}. We decided to write a script to
check references. We opted for Lua, since it's a very straightforward
language and it has an interpreter available on every modern \TeX\
distribution.

\begin{infoblock}{Attention!}
\checkcites\ is known to run with the most recent \verb|texlua| and
\verb|lua| interpreters. Unfortunately, the code is incompatible with
interpreters prior to the Lua 5.1 language specification.
\end{infoblock}

\section{How the script works}
\label{sec:howto}

\checkcites\ uses the generated \verb|.aux| file to start the
analysis. The first step is to extract all citations found, in the
form of \verb|\citation{a}|. For every \verb|\citation| line found,
\checkcites\ will extract the citations and add them to a table, even
for multiple citations separated by commas, like
\verb|\citation{a,b,c}|. Then the citations table has all duplicate
values removed --~in other words, the table becomes a set.  Let's call
$A$ the set of citations.

\begin{infoblock}{Attention!}
If \verb|\citation{*}| is found, \checkcites\ will issue a message
telling that \verb|\nocite{*}| is in the \verb|.tex| document, but
the script will do the check nonetheless.
\end{infoblock}

At the same time \checkcites\ also looks for bibliography data, in the
form of \verb|\bibdata{a}|. Similarly, for every \verb|\bibdata| line
found, the script will extract the bibliography data and add them to a
table, even if they are separated by commas, like
\verb|\bibdata{d,e,f}|. The table has all duplicate values removed.

\begin{infoblock}{Attention!}
If no \verb|\bibdata| command is found, the script ends. There's
nothing to do in this case.
\end{infoblock}

Now, \checkcites\ will extract all entries from the bibliography files
found in the previous step. For every element in the bibliography data
table, the script will look for entries like \verb|@BOOK|,
\verb|@ARTICLE| and so forth --~we actually use pattern matching for
this~-- and add their identifiers to a table. The script treats all
\verb|.bib| files as if they were only one. After all files have been
analyzed and all entries' identifiers extracted, the table has all
duplicate values removed. Let's call $B$ the set of bibliography
entries.

\begin{infoblock}{Attention!}
If \checkcites\ cannot find a certain bibliography file --~that is, a
\verb|.bib| file~-- the script ends. Make sure to put the correct
name of the bibliography file in your \verb|.tex| file.
\end{infoblock}

Now we have both sets $A$ and $B$. In order to get all unused
references in the \verb|.bib| files, we compute the set difference
\[
B - A = \{ x : x \in B, x \notin A \}.
\] 
Similarly, in order to get all undefined references in the \verb|.tex|
file, we compute the set difference
\[
A - B = \{ x : x \in A, x \notin B \}.
\]
If there are either unused or undefined references, \checkcites\ will
print them in a list format. In Section~\ref{sec:usage} there's a more
complete explanation on how to use the script.

\section{Usage}
\label{sec:usage}

\checkcites\ is very easy to use. First of all, let's define two files
that will be used here to explain the script usage. Here's our sample
bibliography file \verb|example.bib|, with five fictional entries.

\begin{tcblisting}{colframe=DarkTurquoise,coltitle=black,listing only,
                   title=Bibliography file, fonttitle=\bfseries,
                   listing options={columns=fullflexible,basicstyle=\ttfamily}}
@BOOK{foo:2012a,
  title = {My Title One},
  publisher = {My Publisher One},
  year = {2012},
  editor = {My Editor One},
  author = {Author One}
}

@BOOK{foo:2012b,
  title = {My Title Two},
  publisher = {My Publisher Two},
  year = {2012},
  editor = {My Editor Two},
  author = {Author Two}
}

@BOOK{foo:2012c,
  title = {My Title Three},
  publisher = {My Publisher Three},
  year = {2012},
  editor = {My Editor Three},
  author = {Author Three}
}

@BOOK{foo:2012d,
  title = {My Title Four},
  publisher = {My Publisher Four},
  year = {2012},
  editor = {My Editor Four},
  author = {Author Four}
}

@BOOK{foo:2012e,
  title = {My Title Five},
  publisher = {My Publisher Five},
  year = {2012},
  editor = {My Editor Five},
  author = {Author Five}
}
\end{tcblisting}

The second file is our main \LaTeX{} document, \verb|document.tex|.

\begin{tcblisting}{colframe=DarkTurquoise,coltitle=black,listing only,
                   title=Main document, fonttitle=\bfseries,
                   listing options={columns=fullflexible,basicstyle=\ttfamily}}
\documentclass{article}

\begin{document}

Hello world \cite{foo:2012a,foo:2012c}, how are you \cite{foo:2012f},
and goodbye \cite{foo:2012d,foo:2012a}.

\bibliographystyle{plain}
\bibliography{example}

\end{document}
\end{tcblisting}

Open a terminal and run \verb|checkcites|:

\begin{terminal}
\begin{verbatim}
$ checkcites
\end{verbatim}
\tcblower
\begin{verbatim}
checkcites.lua -- a reference checker script (v1.0g)
Copyright (c) 2012 Enrico Gregorio, Paulo Roberto Massa Cereda

Usage: checkcites.lua [--all | --unused | --undefined] file.aux

--all         Lists all unused and undefined references.
--unused      Lists only unused references in your 'bib' file.
--undefined   Lists only undefined references in your 'tex' file.

If no flag is provided, '--all' is set by default.
Be sure to have all your 'bib' files in the same directory.
\end{verbatim}
\end{terminal}

If you don't have \checkcites\ installed with your \TeX\ distribution,
you can run the standalone script \verb|checkcites.lua| with either
\verb|texlua| or \verb|lua|. We recommend to use \verb|texlua|, as
it's shipped with all the modern \TeX\ distributions:
\begin{terminal}
\begin{verbatim}
$ texlua checkcites.lua
\end{verbatim}
\end{terminal}

When you run \checkcites\ without providing any argument to it, the
script usage will be printed, as seen in the previous output. The only
required argument is the auxiliary file --~with the \verb|.aux|
extension~-- which is generated when you compile your main \verb|.tex|
file. For example, if your main document is named \verb|foo.tex|, you
probably have a \verb|foo.aux| file too. Let's compile our sample
document \verb|document.tex|:
\begin{terminal}
\begin{verbatim}
$ pdflatex document.tex
\end{verbatim}
\end{terminal}

After running \verb|pdflatex| on our \verb|.tex| file, there's now a
\verb|document.aux| file in our work directory.

\begin{tcblisting}{colframe=DarkTurquoise,coltitle=black,listing only,
                   title=Auxiliary file, fonttitle=\bfseries,
                   listing options={columns=fullflexible,basicstyle=\ttfamily}}
\relax 
\citation{foo:2012a}
\citation{foo:2012c}
\citation{foo:2012f}
\citation{foo:2012d}
\citation{foo:2012a}
\bibstyle{plain}
\bibdata{example}
\end{tcblisting}

Now we can run \checkcites\ on the \verb|document.aux| file:

\begin{terminal}
\begin{verbatim}
$ checkcites document.aux
\end{verbatim}
\tcblower
\begin{verbatim}
checkcites.lua -- a reference checker script (v1.0g)
Copyright (c) 2012 Enrico Gregorio, Paulo Roberto Massa Cereda

I found 4 citation(s).
Great, there's only one 'bib' file. Let me check it.
I found 5 reference(s).

Unused reference(s) in your bibliography file(s): 2
- foo:2012b
- foo:2012e

Undefined reference(s) in your TeX file: 1
- foo:2012f
\end{verbatim}
\end{terminal}

As we can see in the script output, \checkcites\ analyzed both
\verb|.aux| and \verb|.bib| files and found two unused references in
the bibliography file --~\verb|foo:2012b| and \verb|foo:2012e|~-- and
one undefined reference in the document --~\verb|foo:2012f|.

\checkcites\ allows a command line switch that will tell it how to
behave. For example,

\begin{terminal}
\begin{verbatim}
$ checkcites --unused document.aux
\end{verbatim}
\end{terminal}

The \verb|--unused| flag will make the script only look for unused
references in the \verb|.bib| file. The argument order doesn't matter,
you can also run

\begin{terminal}
\begin{verbatim}
$ checkcites document.aux --unused
\end{verbatim}
\end{terminal}

The script will behave the same. Similarly, you can use

\begin{terminal}
\begin{verbatim}
$ checkcites --undefined document.aux
\end{verbatim}
\end{terminal}

The \verb|--undefined| flag will make the script only look for
undefined references in the \verb|.tex| file. If you want \checkcites\
to look for both unused and undefined references, run:

\begin{terminal}
\begin{verbatim}
$ checkcites --all document.aux
\end{verbatim}
\end{terminal}

If no special argument is provided, the \verb|--all| flag is set as default.

\section{License}
\label{sec:license}

This script is licensed under the
\href{http://www.latex-project.org/lppl/}{LaTeX Project Public
License}. If you want to support \LaTeX{} development by a donation,
the best way to do this is donating to the
\href{http://www.tug.org/}{TeX Users Group}.

\begin{infoblock}{Official code repository}
\url{http://github.com/cereda/checkcites}
\end{infoblock}

\end{document}