LATEXDIFF(1)LATEXDIFF(1)NAMElatexdiff - determine and markup differences between two latex files
SYNOPSISlatexdiff [ OPTIONS ] old.tex new.tex > diff.tex
DESCRIPTION
Briefly, latexdiff is a utility program to aid in the management of
revisions of latex documents. It compares two valid latex files, here
called "old.tex" and "new.tex", finds significant differences between
them (i.e., ignoring the number of white spaces and position of line
breaks), and adds special commands to highlight the differences. Where
visual highlighting is not possible, e.g. for changes in the
formatting, the differences are nevertheless marked up in the source.
The program treats the preamble differently from the main document.
Differences between the preambles are found using line-based
differencing (similarly to the Unix diff command, but ignoring white
spaces). A comment, ""%DIF >"" is appended to each added line, i.e. a
line present in "new.tex" but not in "old.tex". Discarded lines
are deactivated by prepending ""%DIF <"". Changed blocks are preceded
by comment lines giving information about line numbers in the original
files. Where there are insignificant differences, the resulting file
"diff.tex" will be similar to "new.tex". At the end of the preamble,
the definitions for latexdiff markup commands are inserted. In
differencing the main body of the text, latexdiff attempts to satisfy
the following guidelines (in order of priority):
1. If both "old.tex" and "new.tex" are valid LaTeX, then the resulting
"diff.tex" should also be valid LateX. (NB If a few plain TeX
commands are used within "old.tex" or "new.tex" then "diff.tex" is
not guaranteed to work but usually will).
2. Significant differences are determined on the level of individual
words. All significant differences, including differences between
comments should be clearly marked in the resulting source code
"diff.tex".
3. If a changed passage contains text or text-producing commands, then
running "diff.tex" through LateX should produce output where added
and discarded passages are highlighted.
4. Where there are insignificant differences, e.g. in the positioning
of line breaks, "diff.tex" should follow the formatting of "new.tex"
For differencing the same algorithm as diff is used but words instead
of lines are compared. An attempt is made to recognize blocks which
are completely changed such that they can be marked up as a unit.
Comments are differenced line by line but the number of spaces within
comments is ignored. Commands including all their arguments are
generally compared as one unit, i.e., no mark-up is inserted into the
arguments of commands. However, for a selected number of commands (for
example, "\caption" and all sectioning commands) the last argument is
known to be text. This text is split into words and differenced just as
ordinary text (use options to show and change the list of text
commands, see below). As the algorithm has no detailed knowledge of
LaTeX, it assumes all pairs of curly braces immediately following a
command (i.e. a sequence of letters beginning with a backslash) are
arguments for that command. As a restriction to condition 1 above it
is thus necessary to surround all arguments with curly braces, and to
not insert extraneous spaces. For example, write
\section{\textem{This is an emphasized section title}}
and not
\section {\textem{This is an emphasized section title}}
or
\section\textem{This is an emphasized section title}
even though all varieties are the same to LaTeX (but see --allow-spaces
option which allows the second variety).
For environments whose content does not conform to standard LaTeX or
where graphical markup does not make sense all markup commands can be
removed by setting the PICTUREENV configuration variable, set by
default to "picture" and "DIFnomarkup" environments; see --config
option). The latter environment ("DIFnomarkup") can be used to protect
parts of the latex file where the markup results in illegal markup. You
have to surround the offending passage in both the old and new file by
"\begin{DIFnomarkup}" and "\end{DIFnomarkup}". You must define the
environment in the preambles of both old and new documents. I prefer to
define it as a null-environment,
"\newenvironment{DIFnomarkup}{}{}"
but the choice is yours. Any markup within the environment will be
removed, and generally everything within the environment will just be
taken from the new file.
It is also possible to difference files which do not have a preamble.
In this case, the file is processed in the main document mode, but the
definitions of the markup commands are not inserted.
All markup commands inserted by latexdiff begin with ""\DIF"". Added
blocks containing words, commands or comments which are in "new.tex"
but not in "old.tex" are marked by "\DIFaddbegin" and "\DIFaddend".
Discarded blocks are marked by "\DIFdelbegin" and "\DIFdelend". Within
added blocks all text is highlighted with "\DIFadd" like this:
"\DIFadd{Added text block}" Selected `safe' commands can be contained
in these text blocks as well (use options to show and change the list
of safe commands, see below). All other commands as well as braces "{"
and "}" are never put within the scope of "\DIFadd". Added comments
are marked by prepending ""%DIF > "".
Within deleted blocks text is highlighted with "\DIFdel". Deleted
comments are marked by prepending ""%DIF < "". Non-safe command and
curly braces within deleted blocks are commented out with
""%DIFDELCMD < "".
OPTIONS
Preamble
The following options determine the visual markup style by adding the
appropriate command definitions to the preamble. See the end of this
section for a description of available styles.
--type=markupstyle or -t markupstyle
Add code to preamble for selected markup style. This option defines
"\DIFadd" and "\DIFdel" commands. Available styles:
"UNDERLINE CTRADITIONAL TRADITIONAL CFONT FONTSTRIKE INVISIBLE
CHANGEBAR CCHANGEBAR CULINECHBAR CFONTCBHBAR BOLD"
[ Default: "UNDERLINE" ]
--subtype=markstyle or -s markstyle
Add code to preamble for selected style for bracketing commands
(e.g. to mark changes in margin). This option defines
"\DIFaddbegin", "\DIFaddend", "\DIFdelbegin" and "\DIFdelend"
commands. Available styles: "SAFE MARGIN COLOR DVIPSCOL ZLABEL
ONLYCHANGEDPAGE (LABEL)*"
[ Default: "SAFE" ] * Subtype "LABEL" is deprecated
--floattype=markstyle or -f markstyle
Add code to preamble for selected style which replace standard
marking and markup commands within floats (e.g., marginal remarks
cause an error within floats so marginal marking can be disabled
thus). This option defines all "\DIF...FL" commands. Available
styles: "FLOATSAFE TRADITIONALSAFE IDENTICAL"
[ Default: "FLOATSAFE" ]
--encoding=enc or -e enc
Specify encoding of old.tex and new.tex. Typical encodings are
"ascii", "utf8", "latin1", "latin9". A list of available encodings
can be obtained by executing
"perl -MEncode -e 'print join ("\n",Encode-"encodings( ":all" )) ;'
>
[Default encoding is utf8 unless the first few lines of the
preamble contain an invocation "\usepackage[..]{inputenc}" in which
case the encoding chosen by this command is asssumed. Note that
ASCII (standard latex) is a subset of utf8]
--preamble=file or -p file
Insert file at end of preamble instead of generating preamble. The
preamble must define the following commands "\DIFaddbegin,
\DIFaddend, \DIFadd{..}, \DIFdelbegin,\DIFdelend,\DIFdel{..}," and
varieties for use within floats "\DIFaddbeginFL, \DIFaddendFL,
\DIFaddFL{..}, \DIFdelbeginFL, \DIFdelendFL, \DIFdelFL{..}" (If
this option is set -t, -s, and -f options are ignored.)
--packages=pkg1,pkg2,..
Tell latexdiff that .tex file is processed with the packages in
list loaded. This is normally not necessary if the .tex file
includes the preamble, as the preamble is automatically scanned for
"\usepackage" commands. Use of the --packages option disables
automatic scanning, so if for any reason package specific parsing
needs to be switched off, use --packages=none. The following
packages trigger special behaviour:
"amsmath"
Configuration variable MATHARRREPL is set to "align*"
(Default: "eqnarray*"). (Note that many of the amsmath
array environments are already recognised by default as
such)
"endfloat"
Ensure that "\begin{figure}" and "\end{figure}" always
appear by themselves on a line.
"hyperref"
Change name of "\DIFadd" and "\DIFdel" commands to
"\DIFaddtex" and "\DIFdeltex" and define new "\DIFadd" and
"\DIFdel" commands, which provide a wrapper for these
commands, using them for the text but not for the link
defining command (where any markup would cause errors).
"apacite"
Redefine the commands recognised as citation commands.
"siunitx"
Treat "\SI" as equivalent to citation commands (i.e.
protect with "\mbox" if markup style uses ulem package.
"cleveref"
Treat "\cref,\Cref", etc as equivalent to citation commands
(i.e. protect with "\mbox" if markup style uses ulem
package.
"glossaries"
Define most of the glossaries commands as safe, protecting
them with \mbox'es where needed
"mhchem"
Treat "\ce" as a safe command, i.e. it will be highlighted
(note that "\cee" will not be highlighted in equations as
this leads to processing errors)
"chemformula" or "chemmacros"
Treat "\ch" as a safe command outside equations, i.e. it
will be highlighted (note that "\ch" will not be
highlighted in equations as this leads to processing
errors)
[ Default: scan the preamble for "\usepackage" commands to
determine
loaded packages. ]
--show-preamble
Print generated or included preamble commands to stdout.
Configuration
--exclude-safecmd=exclude-file or -A exclude-file or
--exclude-safecmd="cmd1,cmd2,..."
--replace-safecmd=replace-file
--append-safecmd=append-file or -a append-file or
--append-safecmd="cmd1,cmd2,..."
Exclude from, replace or append to the list of regular expressions
(RegEx) matching commands which are safe to use within the scope of
a "\DIFadd" or "\DIFdel" command. The file must contain one Perl-
RegEx per line (Comment lines beginning with # or % are ignored).
Note that the RegEx needs to match the whole of the token, i.e.,
/^regex$/ is implied and that the initial "\" of the command is not
included. The --exclude-safecmd and --append-safecmd options can
be combined with the ---replace-safecmd option and can be used
repeatedly to add cumulatively to the lists.
--exclude-safecmd and --append-safecmd can also take a comma
separated list as input. If a comma for one of the regex is
required, escape it thus "\,". In most cases it will be necessary
to protect the comma-separated list from the shell by putting it in
quotation marks.
--exclude-textcmd=exclude-file or -X exclude-file or
--exclude-textcmd="cmd1,cmd2,..."
--replace-textcmd=replace-file
--append-textcmd=append-file or -x append-file or
--append-textcmd="cmd1,cmd2,..."
Exclude from, replace or append to the list of regular expressions
matching commands whose last argument is text. See entry for
--exclude-safecmd directly above for further details.
--replace-context1cmd=replace-file
--append-context1cmd=append-file or =item
--append-context1cmd="cmd1,cmd2,..."
Replace or append to the list of regex matching commands whose last
argument is text but which require a particular context to work,
e.g. \caption will only work within a figure or table. These
commands behave like text commands, except when they occur in a
deleted section, when they are disabled, but their argument is
shown as deleted text.
--replace-context2cmd=replace-file
--append-context2cmd=append-file or =item
--append-context2cmd="cmd1,cmd2,..."
As corresponding commands for context1. The only difference is
that context2 commands are completely disabled in deleted sections,
including their arguments.
--exclude-mboxsafecmd=exclude-file or
--exclude-mboxsafecmd="cmd1,cmd2,..."
--append-mboxsafecmd=append-file or
--append-mboxsafecmd="cmd1,cmd2,..."
Define safe commands, which additionally need to be protected by
encapsulating in an \\mbox{..}. This is sometimes needed to get
around incompatibilities between external packages and the ulem
package, which is used for highlighting in the default style
UNDERLINE as well as CULINECHBAR CFONTSTRIKE
--config var1=val1,var2=val2,... or -c var1=val1,..
-c configfile
Set configuration variables. The option can be repeated to set
different variables (as an alternative to the comma-separated
list). Available variables (see below for further explanations):
"ARRENV" (RegEx)
"COUNTERCMD" (RegEx)
"FLOATENV" (RegEx)
"ITEMCMD" (RegEx)
"LISTENV" (RegEx)
"MATHARRENV" (RegEx)
"MATHARRREPL" (String)
"MATHENV" (RegEx)
"MATHREPL" (String)
"MINWORDSBLOCK" (integer)
"PICTUREENV" (RegEx)
--show-safecmd
Print list of RegEx matching and excluding safe commands.
--show-textcmd
Print list of RegEx matching and excluding commands with text
argument.
--show-config
Show values of configuration variables.
--show-all
Combine all --show commands.
NB For all --show commands, no "old.tex" or "new.tex" file needs to
be specified, and no differencing takes place.
Other configuration options:
--allow-spaces
Allow spaces between bracketed or braced arguments to commands.
Note that this option might have undesirable side effects
(unrelated scope might get lumpeded with preceding commands) so
should only be used if the default produces erroneous results.
(Default requires arguments to directly follow each other without
intervening spaces).
--math-markup=level
Determine granularity of markup in displayed math environments:
Possible values for level are (both numerical and text labels are
acceptable):
"off" or 0: suppress markup for math environments. Deleted
equations will not appear in diff file. This mode can be used if
all the other modes cause invalid latex code.
"whole" or 1: Differencing on the level of whole equations. Even
trivial changes to equations cause the whole equation to be marked
changed. This mode can be used if processing in coarse or fine
mode results in invalid latex code.
"coarse" or 2: Detect changes within equations marked up with a
coarse granularity; changes in equation type (e.g.displaymath to
equation) appear as a change to the complete equation. This mode is
recommended for situations where the content and order of some
equations are still being changed. [Default]
"fine" or 3: Detect small change in equations and mark up at fine
granularity. This mode is most suitable, if only minor changes to
equations are expected, e.g. correction of typos.
--disable-citation-markup or --disable-auto-mbox
Suppress citation markup and markup of other vulnerable commands in
styles using ulem (UNDERLINE,FONTSTRIKE, CULINECHBAR) (the two
options are identical and are simply aliases)
--enable-citation-markup or --enforce-auto-mbox
Protect citation commands and other vulnerable commands in changed
sections with "\mbox" command, i.e. use default behaviour for ulem
package for other packages (the two options are identical and are
simply aliases)
Miscellaneous
--verbose or -V
Output various status information to stderr during processing.
Default is to work silently.
--driver=type
Choose driver for changebar package (only relevant for styles using
changebar: CCHANGEBAR CFONTCHBAR CULINECHBAR CHANGEBAR).
Possible drivers are listed in changebar manual, e.g.
pdftex,dvips,dvitops
[Default: dvips]
--ignore-warnings
Suppress warnings about inconsistencies in length between input and
parsed strings and missing characters. These warning messages are
often related to non-standard latex or latex constructions with a
syntax unknown to "latexdiff" but the resulting difference argument
is often fully functional anyway, particularly if the non-standard
latex only occurs in parts of the text which have not changed.
--label=label or -L label
Sets the labels used to describe the old and new files. The first
use of this option sets the label describing the old file and the
second use of the option sets the label for the new file, i.e. set
both labels like this "-L labelold -L labelnew". [Default: use the
filename and modification dates for the label]
--no-label
Suppress inclusion of old and new file names as comment in output
file
--visble-label
Include old and new filenames (or labels set with "--label" option)
as visible output.
--flatten
Replace "\input" and "\include" commands within body by the content
of the files in their argument. If "\includeonly" is present in
the preamble, only those files are expanded into the document.
However, no recursion is done, i.e. "\input" and "\include"
commands within included sections are not expanded. The included
files are assumed to
be located in the same directories as the old and new master
files, respectively, making it possible to organise files into old
and new directories. --flatten is applied recursively, so inputted
files can contain further "\input" statements.
Use of this option might result in prohibitive processing times for
larger documents, and the resulting difference document no longer
reflects the structure of the input documents.
--help or -h
Show help text
--version
Show version number
Predefined styles
Major types
The major type determine the markup of plain text and some selected
latex commands outside floats by defining the markup commands
"\DIFadd{...}" and "\DIFdel{...}" .
"UNDERLINE"
Added text is wavy-underlined and blue, discarded text is
struck out and red (Requires color and ulem packages).
Overstriking does not work in displayed math equations such
that deleted parts of equation are underlined, not struck out
(this is a shortcoming inherent to the ulem package).
"CTRADITIONAL"
Added text is blue and set in sans-serif, and a red footnote
is created for each discarded piece of text. (Requires color
package)
"TRADITIONAL"
Like "CTRADITIONAL" but without the use of color.
"CFONT" Added text is blue and set in sans-serif, and discarded text
is red and very small size.
"FONTSTRIKE"
Added tex is set in sans-serif, discarded text small and
struck out
"CCHANGEBAR"
Added text is blue, and discarded text is red. Additionally,
the changed text is marked with a bar in the margin (Requires
color and changebar packages).
"CFONTCHBAR"
Like "CFONT" but with additional changebars (Requires color
and changebar packages).
"CULINECHBAR"
Like "UNDERLINE" but with additional changebars (Requires
color, ulem and changebar packages).
"CHANGEBAR"
No mark up of text, but mark margins with changebars
(Requires changebar package).
"INVISIBLE"
No visible markup (but generic markup commands will still be
inserted.
"BOLD" Added text is set in bold face, discarded is not shown.
Subtypes
The subtype defines the commands that are inserted at the begin and end
of added or discarded blocks, irrespectively of whether these blocks
contain text or commands (Defined commands: "\DIFaddbegin, \DIFaddend,
\DIFdelbegin, \DIFdelend")
"SAFE" No additional markup (Recommended choice)
"MARGIN" Mark beginning and end of changed blocks with symbols in the
margin nearby (using the standard "\marginpar" command - note
that this sometimes moves somewhat from the intended
position.
"COLOR" An alternative way of marking added passages in blue, and
deleted ones in red. (It is recommeneded to use instead the
main types to effect colored markup, although in some cases
coloring with dvipscol can be more complete, for example with
citation commands).
"DVIPSCOL"
An alternative way of marking added passages in blue, and
deleted ones in red. Note that "DVIPSCOL" only works with the
dvips converter, e.g. not pdflatex. (it is recommeneded to
use instead the main types to effect colored markup, although
in some cases coloring with dvipscol can be more complete).
"ZLABEL" can be used to highlight only changed pages, but requires
post-processing. It is recommend to not call this option
manually but use "latexdiff-vc" with "--only-changes" option.
Alternatively, use the script given within preamble of diff
files made using this style.
"ONLYCHANGEDPAGE"
also highlights changed pages, without the need for post-
processing, but might not work reliably if there is floating
material (figures, tables).
"LABEL" is similar to "ZLABEL", but does not need the zref package
and works less reliably (deprecated).
Float Types
Some of the markup used in the main text might cause problems when used
within floats (e.g. figures or tables). For this reason alternative
versions of all markup commands are used within floats. The float type
defines these alternative commands.
"FLOATSAFE"
Use identical markup for text as in the main body, but set
all commands marking the begin and end of changed blocks to
null-commands. You have to choose this float type if your
subtype is "MARGIN" as "\marginpar" does not work properly
within floats.
"TRADITIONALSAFE"
Mark additions the same way as in the main text. Deleted
environments are marked by angular brackets \[ and \] and the
deleted text is set in scriptscript size. This float type
should always be used with the "TRADITIONAL" and
"CTRADITIONAL" markup types as the \footnote command does not
work properly in floating environments.
"IDENTICAL"
Make no difference between the main text and floats.
Configuration Variables
"ARRENV" If a match to "ARRENV" is found within an inline math
environment within a deleted or added block, then the inlined
math is surrounded by "\mbox{"..."}". This is necessary as
underlining does not work within inlined array environments.
[ Default: "ARRENV"="(?:array|[pbvBV]matrix)"
"COUNTERCMD"
If a command in a deleted block which is also in the textcmd
list matches "COUNTERCMD" then an additional command
"\addtocounter{"cntcmd"}{-1}", where cntcmd is the matching
command, is appended in the diff file such that the numbering
in the diff file remains synchronized with the numbering in
the new file.
[ Default: "COUNTERCMD"="(?:footnote|part|section|subsection"
...
"|subsubsection|paragraph|subparagraph)" ]
"FLOATENV"
Environments whose name matches the regular expression in
"FLOATENV" are considered floats. Within these environments,
the latexdiff markup commands are replaced by their FL
variaties.
[ Default: "(?:figure|table|plate)[\w\d*@]*" ]
"ITEMCMD" Commands representing new item line with list environments.
[ Default: \"item" ]
"LISTENV" Environments whose name matches the regular expression in
"LISTENV" are list environments.
[ Default: "(?:itemize|enumerate|description)" ]
"MATHENV","MATHREPL"
If both \begin and \end for a math environment (environment
name matching "MATHENV" or \[ and \]) are within the same
deleted block, they are replaced by a \begin and \end
commands for "MATHREPL" rather than being commented out.
[ Default: "MATHENV"="(?:displaymath|equation)" ,
"MATHREPL"="displaymath" ]
"MATHARRENV","MATHARRREPL"
as "MATHENV","MATHREPL" but for equation arrays
[ Default: "MATHARRENV"="eqnarray\*?" ,
"MATHREPL"="eqnarray" ]
"MINWORDSBLOCK"
Minimum number of tokens required to form an independent
block. This value is used in the algorithm to detect changes
of complete blocks by merging identical text parts of less
than "MINWORDSBLOCK" to the preceding added and discarded
parts.
[ Default: 3 ]
"PICTUREENV"
Within environments whose name matches the regular expression
in "PICTUREENV" all latexdiff markup is removed (in
pathologic cases this might lead to inconsistent markup but
this situation should be rare).
[ Default: "(?:picture|DIFnomarkup)[\w\d*@]*" ]
COMMON PROBLEMS AND FAQ
Citations result in overfull boxes
There is an incompatibility between the "ulem" package, which
"latexdiff" uses for underlining and striking out in the
UNDERLINE style, the default style, and the way citations are
generated. In order to be able to mark up citations properly,
they are enclosed with an "\mbox" command. As mboxes cannot
be broken across lines, this procedure frequently results in
overfull boxes, possibly obscuring the content as it extends
beyond the right margin. The same occurs for some other
packages (e.g., siunitx). If this is a problem, you have two
possibilities.
1. Use "CFONT" type markup (option "-t CFONT"): If this
markup is chosen, then changed citations are no longer marked
up with the wavy line (additions) or struck out (deletions),
but are still highlighted in the appropriate color, and
deleted text is shown with a different font. Other styles not
using the "ulem" package will also work.
2. Choose option "--disable-citation-markup" which turns off
the marking up of citations: deleted citations are no longer
shown, and added citations are shown without markup. (This
was the default behaviour of latexdiff at versions 0.6 and
older)
For custom packages you can define the commands which need to
be protected by "\mbox" with "--append-mboxsafecmd" and
"--excludemboxsafecmd" options (submit your lists of command
as feature request at github page to set the default
behaviour of future versions, see section 6)
Changes in complicated mathematical equations result in latex
processing errors
Try options "--math-markup=whole". If even that fails, you
can turn off mark up for equations with "--math-markup=off".
How can I just show the pages where changes had been made
Use options -"-s ZLABEL" (some postprocessing required) or
"-s ONLYCHANGEDPAGE". "latexdiff-vc --ps|--pdf" with
"--only-changes" option takes care of the post-processing for
you (requires zref package to be installed).
BUGS
Option allow-spaces not implemented entirely consistently. It breaks
the rules that number and type of white space does not matter, as
different numbers of inter-argument spaces are treated as significant.
Please submit bug reports using the issue tracker of the github
repository page https://github.com/ftilmann/latexdiff.git, or send them
to tilmann -- AT -- gfz-potsdam.de. Include the version number of
latexdiff (from comments at the top of the source or use --version).
If you come across latex files that are error-free and conform to the
specifications set out above, and whose differencing still does not
result in error-free latex, please send me those files, ideally edited
to only contain the offending passage as long as that still reproduces
the problem. If your file relies on non-standard class files, you must
include those. I will not look at examples where I have trouble to
latex the original files.
SEE ALSO
latexrevise, latexdiff-vc
PORTABILITYlatexdiff does not make use of external commands and thus should run on
any platform supporting Perl 5.6 or higher. If files with encodings
other than ASCII or UTF-8 are processed, Perl 5.8 or higher is
required.
The standard version of latexdiff requires installation of the Perl
package "Algorithm::Diff" (available from www.cpan.org -
http://search.cpan.org/~nedkonz/Algorithm-Diff-1.15) but a stand-alone
version, latexdiff-so, which has this package inlined, is available,
too. latexdiff-fast requires the diff command to be present.
AUTHOR
Version 1.1.1 Copyright (C) 2004-2015 Frederik Tilmann
This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License Version 3
Contributors of fixes and additions: V. Kuhlmann, J. Paisley, N.
Becker, T. Doerges, K. Huebner, T. Connors, Sebastian Gouezel and many
others. Thanks to the many people who sent in bug reports, feature
suggestions, and other feedback.
perl v5.18.2 2015-12-26 LATEXDIFF(1)
