cdif - word context diff


Version 4.18.2


cdif [option] file1 file2

cdif [option] [diff-data]


        -c, -Cn         context diff
        -u, -Un         unified diff
        -i              ignore case
        -b              ignore space change
        -w              ignore whitespace
        -t              expand tabs
        --rcs           use rcsdiff
        -r<rev>, -q     rcs options

        --diff=command      specify diff command
        --subdiff=command   specify backend diff command
        --stat              show statistical information
        --colormap=s        specify color map
        --unit=s            word, char or mecab     (default word)
        --[no]color         color or not            (default true)
        --[no]256           ANSI 256 color mode     (default true)
        --[no]commandcolor  color for command line  (default true)
        --[no]markcolor     color for diff mark     (default true)
        --[no]textcolor     color for normal text   (default true)
        --[no]unknowncolor  color for unknown text  (default true)
        --[no]old           print old text          (default true)
        --[no]new           print new text          (default true)
        --[no]command       print diff command line (default true)
        --[no]unknown       print unknown line      (default true)
        --[no]mark          print mark or not       (default true)
        --[no]prefix        read git --graph output (default true)
        --prefix-pattern    prefix pattern
        --visible char=?    set visible attributes
        --[no]mecab         use mecab tokenizer     (default false)
        --[no]lenience      suppress unexpected input warning (default true)


cdif is a post-processor of the Unix diff command. It highlights deleted, changed and added words based on word context (--unit=word by default). If you want to compare text character-by-character, use option --unit=char. Option --unit=mecab tells to use external mecab command as a tokenizer for Japanese text.

If single or no file is specified, cdif reads that file or STDIN as a output from diff command.

Lines those don't look like diff output are simply ignored and printed.



Almost same as diff command.

--rcs, -rrev, -q

Use rcsdiff instead of normal diff. Option --rcs is not required when -rrev is supplied.


Specify the comparing unit. Default is word and compare each line word-by-word. Specify char if you want to compare them character-by-character.

When mecab is given as an unit, mecab command is called as a tokenizer for non-ASCII text. ASCII text is compared word-by-word. External mecab command has to been installed.


Shortcut for --unit=mecab.


Specify the diff command to use.


Specify the backend diff command to get word differences. Accept normal and unified diff format.

If you want to use git diff command, don't forget to set -U0 option.

    --subdiff="git diff -U0 --no-index --histogram"

Use ANSI color escape sequence for output.

--colormap=colormap, --cm=colormap

Basic colormap format is :


where the FIELD is one from these :

    COMMAND  Command line
    OMARK    Old mark
    NMARK    New mark
    OTEXT    Old text
    NTEXT    New text
    OCHANGE  Old change part
    NCHANGE  New change part
    APPEND   Appended part
    DELETE   Deleted part
    VISIBLE  Visualized invisible chars

and additional Common and Merged FIELDs for git-diff combined format.

    CMARK    Common mark
    CTEXT    Common text
    MMARK    Merged mark
    MTEXT    Merged text

You can make multiple fields same color joining them by = :


Also wildcard can be used for field name :


Multiple fields can be specified by repeating options

    --cm FILED1=COLOR1 --cm FIELD2=COLOR2 ...

or combined with comma (,) :

    --cm FILED1=COLOR1,FIELD2=COLOR2, ...

Color specification is a combination of single uppercase character representing 8 colors :

    R  Red
    G  Green
    B  Blue
    C  Cyan
    M  Magenta
    Y  Yellow
    K  Black
    W  White

and alternative (usually brighter) colors in lowercase :

    r, g, b, c, m, y, k, w

or RGB values and 24 grey levels if using ANSI 256 or full color terminal :

    (255,255,255)      : 24bit decimal RGB colors
    #000000 .. #FFFFFF : 24bit hex RGB colors
    #000    .. #FFF    : 12bit hex RGB 4096 colors
    000 .. 555         : 6x6x6 RGB 216 colors
    L00 .. L25         : Black (L00), 24 grey levels, White (L25)

or color names enclosed by angle bracket :

    <red> <blue> <green> <cyan> <magenta> <yellow>
    <aliceblue> <honeydue> <hotpink> <mooccasin>

with other special effects :

    D  Double-struck (boldface)
    I  Italic
    U  Underline
    S  Stand-out (reverse video)

Above color spec is simplified summary so if you want complete information, read Getopt::EX::Colormap.

Defaults are :

    COMMAND => "555/222E"
    OMARK   => "CS"
    NMARK   => "MS"
    OTEXT   => "C"
    NTEXT   => "M"
    OCHANGE => "K/445"
    NCHANGE => "K/445"
    DELETE  => "K/544"
    APPEND  => "K/544"

    CMARK   => "GS"
    MMARK   => "YS"
    CTEXT   => "G"
    MTEXT   => "Y"

This is equivalent to :

    cdif --cm 'COMMAND=555/222E,OMARK=CS,NMARK=MS' \
         --cm 'OTEXT=C,NTEXT=M,*CHANGE=BD/445,DELETE=APPEND=RD/544' \
--[no]commandcolor, --cc
--[no]markcolor, --mc
--[no]textcolor, --tc
--[no]unknowncolor, --uc

Enable/Disable using color for the corresponding field.

--[no]old, --[no]new

Print or not old/new text in diff output.


Print or not command lines preceding diff output.


Print or not lines not look like diff output.


Print or not marks at the top of diff output lines. At this point, this option is effective only for unified diff.

Next example produces the output exactly same as new except visual effects.

    cdif -U100 --no-mark --no-old --no-command --no-unknown old new

These options are prepared for watchdiff(1) command.


Understand prefix for diff output including git --graph option. True by default.


Specify prefix pattern in regex. Default pattern is:

    (?:\| )*(?:  )*

This pattern matches git graph style and whitespace indented diff output.

--visible chaname=[0,1]

Set visible attribute for specified characters. Visible character is converted to corresponding Unicode symbol character. Default visible: nul, bel, bs, vt, np, cr, esc, del. Default invisible: ht, nl, sp.

    NAME  CODE  Unicode NAME                      DEFAULT
    ----  ----  --------------------------------  -------
    nul   \000  SYMBOL FOR NULL                   YES
    bel   \007  SYMBOL FOR BELL                   YES
    bs    \010  SYMBOL FOR BACKSPACE              YES
    nl    \012  SYMBOL FOR NEWLINE                NO
    np    \014  SYMBOL FOR FORM FEED              YES
    cr    \015  SYMBOL FOR CARRIAGE RETURN        YES
    esc   \033  SYMBOL FOR ESCAPE                 YES
    sp    \040  SYMBOL FOR SPACE                  NO
    del   \177  SYMBOL FOR DELETE                 YES

Multiple characters can be specified at once, by assembling them by comma (,) like --visible ht=1,sp=1; or connecting them by equal sign (=) like --visible ht=sp=1. Character name accept wildcard; --visible '*=1'.

Colormap name VISIBLE is applied to those characters. Default setting is S, and visible characters are displayed in reverse video. Unlike other colormaps, only special effects can be set to this name. Effect D (double-struck) is exception (see ~ section in Getopt::EX::Colormap).


Print statistical information at the end of output. It shows number of total appended/deleted/changed words in the context of cdif. It's common to have many insertions and deletions of newlines because of text filling process. So normal information is followed by modified number which ignores insert/delete newlines.


Suppress warning message for unexpected input from diff command. True by default.


Environment variable CDIFOPTS is used to set default options.


Kazumasa Utashiro


Copyright 1992-2021 Kazumasa Utashiro

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.


sdif(1), watchdiff(1)



cdif is naturally not very fast because it uses normal diff command as a back-end processor to compare words.