path: root/uxd.rst

.. include:: ver.rst

.. |date| date::

===
uxd
===

----------------
UTF-8 hex dumper
----------------

:Manual section: 1
:Manual group: Urchlay's Utilities
:Date: |date|
:Version: |version|

SYNOPSIS
========

uxd [**-n**] [**-c** *colors*] [**-d** *data*] [**-l** *length*] [**-o** *offset*] [[**-s** | **-S**] *seekpos*] [-[**1bchijmnprtTuvw**] ...] [*file* | *-*]

DESCRIPTION
===========

**uxd** is a hex dump utility that's aware of UTF-8 multibyte sequence
semantics, and uses colorized output to indicate which byte
sequences go with which human-readable characters.

Input is read from *file*, or standard input if *file* is missing or
given as **-**. The input is treated as UTF-8 encoded Unicode. Since
ASCII is a subset, **uxd** works fine on plain ASCII files too. Other
encodings such as UTF-16, ISO-8859-*, Shift-JIS, etc, can be used, but
**uxd** won't handle these any better than a regular hex-dump utility
such as **xxd**.

Output is written to standard output, which is normally a
terminal. It's assumed that the terminal supports ANSI-style color and
UTF-8. See **TERMINAL SUPPORT** below. If you want to pipe the output
to a pager, try **less -R**.

OPTIONS
=======

These options can be used on the command line, and/or in the
**UXD_OPTS** environment variable. The command line takes precedence
over the environment.

Options can be bundled: **-ubc1234** is the same as **-u** **-b** **-c
1234**. The one exception is the **-n** option, which should appear
by itself.

The options that accept numbers (**-l**, **-o**, **-s**, and **-S**)
allow decimal, hex (with *0x* prefix), or octal (with *0* prefix).
Also, you can use the suffixes *k*, *m*, and *g* for power-of-2 based
kilobytes, megabytes, or gigabytes (e.g. *1k* is 1024 bytes), as well
as *K*, *M*, and *G* for power-of-10 based (e.g. *1K* is 1000 bytes).

.. the comments are turned into the --help message by mkusage.pl.

-1
   Don't alternate colors for normal characters.

.. don't alternate colors for normal characters.

-b
  Bold output. This may be more or less readable, depending on your
  terminal and its color settings. Ignored if **-m** given.

.. bold color output.

-c nnnnn
  Set the colors to use. Must be 2 to 5 digits, from 0 to 7. These are
  standard ANSI colors. The first 2 are the alternating colors for
  normal characters, the 3rd and 4th (optional) are the alternating color for non-printable
  and space characters, and the 5th (optional) is for invalid UTF-8
  sequences. Default: **23561**. This option also disables a prior **-m**
  option.

.. colors (2 to 5 digits, 0 to 7).

-d data
  Dump this data, instead of reading from a file or stdin. If *data*
  contains spaces or shell metacharacters, make sure you remember to
  quote it. Only one **-d** option can be given.

.. dump this data instead of a file.

-h, --help
  Print built-in usage message and exit.

.. print this help message.

-i
  After dumping, print information about the input: number of bytes,
  characters, ASCII (one-byte) characters, multi-byte characters, and
  bad sequences.

.. print number of bytes/chars/ascii/multibyte/bad sequences.

-j
  Java mode (aka MUTF-8). Identical to UTF-8 except the
  overlong **0xc0 0x80** encoding for codepoint U+0000 (aka NUL),
  is highlighted in purple and not counted as an error.
  This may be useful for looking at serialized data created by Java
  programs.

.. java (MUTF-8) mode: allow 0xc0 0x80 for U+0000.

-l length
  Stop dumping after *length* bytes (not characters). If the limit is
  reached in the middle of a multibyte character, the entire character
  will be dumped. Negative *length* doesn't make sense, and
  is an error.

.. stop dumping after <length> bytes (not characters).

-m
  Monochrome mode. Uses underline, bold, reverse video instead of color.
  Use this if you have trouble distinguishing the colors, or if they
  look too much like angry fruit salad. Disables prior **-b**, **-c**
  options.

.. monochrome mode.

-n
  Ignore **UXD_OPTS** environment variable. This option should not be
  bundled with other options (e.g. use **-n -u**, not **-nu**).

.. ignore UXD_OPTS environment variable.

-o offset
  Add this amount to the hex offsets (left column). May be negative,
  if you can think of a reason to want it to be.

.. added to hex offsets (decimal, 0x hex, 0 octal).

-p
  Permissive mode. Turns off error highlighting for overlongs, codepoints
  above **U+10FFFF**, and surrogates. Only malformed sequences will be
  highlighed in red.

-r
  Don't highlight multi-byte sequences in reverse video.

.. don't highlight multi-byte chars in reverse video.

-s pos
  Seek in input before starting to dump. *pos* is bytes, not
  characters. Positive *pos* means seek from the start of the
  input. Negative *pos* only works on files (not standard input);
  it means seek backward from EOF.

.. seek in input before dumping (-pos = seek back from EOF).

-S pos
  Same as **-s**, but the displayed offsets start at 0 rather than the
  position after seeking. **-S 100** is the same as **-s 100 -o -100**.
  Works with negative *pos*, too.

.. like -s, but also sets -o so addresses start at 0.

-t
  Put terminal in UTF-8 mode, if possible. Prints **ESC % G** sequence, which
  may or may not be supported by your terminal (works for **xterm**\(1);
  does not work for **urxvt**\(1)). To avoid surprises, this
  option also takes the terminal out of UTF-8 mode before **uxd** exits,
  on the theory that you won't need this option if the terminal is normally
  running in UTF-8 mode. If this assumption turns out to be wrong, you can
  use the **-T** option instead.

-T
  Same as **-T**, but leaves the terminal in UTF-8 mode on exit. Can be
  used to recover from a previous misuse of **-t**, like so::

    uxd -T -dd

-u
  Use uppercase hex digits *A-F*. Default is lowercase.

.. uppercase hex digits.

-v, --version
  Print version number and exit.

.. print version of uxd.

-w
  WTF-8 mode. Surrogates **U+D800** to **U+D8FF** will be highlighted in
  purple and not counted as errors.

.. WTF-8 mode (allow surrogates).

OUTPUT FORMAT
=============

The output is designed to fit in an 80-column terminal.
If you want HTML output, you might have a look at **aha**\(1).

Each line of output consists of eighteen columns: the offset from the
start of the file (in hex; minimum 4 digits), 16 bytes of hex
data (or empty cells, if the last line of the dump is for fewer than
16 bytes), and the human-readable form of the same data.

The hex bytes and human-readable data are colorized to make it obvious
which bytes make up each character. Since UTF-8 is a variable-width
encoding, this means that one character may be composed of up to
4 bytes.

The hex bytes that make up one character are displayed in the same
color, which alternates between yellow and green for successive
characters. In addition, they have dashes instead of spaces between
them. An example would be **c3-b1** (for an ñ character).

The 16-byte hex display always has an extra "spacer" column in the
center. Normally this is a space, but if a multibyte character spans
it, it will be a dash (so there'll be two dashes: **c3--b1**).

Since the output lines are always 16 hex bytes, multibyte characters
can span two lines. When this happens, the character itself will be
printed on the first line, along with the first byte(s) on hex. The
last hex byte will be followed by a dash, and the next line of hex
dump will have the remaining bytes (in the same color as the first
bytes and character). This sounds complicated, but it's easy to
understand once you see it a few times.

EXAMPLE
=======

It's hard to give a proper example, since man pages don't support
color. You'll have to use your imagination. Also, this section of
the man page requires your man command to support UTF-8 embedded in
the man page. If the examples looks mangled, try viewing the source
(uxd.rst) in a text editor.

Example copied from the Japanese **ls**\(1) man page::

  $ echo デフォル | ./uxd
  0000: e3-83-87 e3-83-95 e3-82--a9 e3-83-ab 0a           デフォル↵
        GGGGGGGG YYYYYYYY GGGGGGGGG YYYYYYYY PP           G Y G Y P

The colors are indicated by G/Y/P, for green, yellow, and purple. The
character above each letter is displayed in that color.

From the colorization, and from the dashes between the bytes, it's
obvious that "e3 83 87" is the hex representation of the first
character, and that the 2nd is represented by "e3 83 95.

The newline is displayed in purple because it's not a regular
printable character. Its human-readable representation is ↵. Note
that if a regular ↵ character appears in the input, it'll be
rendered in either green or yellow (so you can tell it's not just
another newline).

COLORS
======

The colors in this description are the default ones. They can be
changed with the **-c** option (see above).

**green**, **yellow**
  Printable characters (except the space, U+0020) alternate between green and yellow.

**purple**, **cyan**
  Spaces and unprintable characters ("control" characters, newlines,
  tabs, etc) alternate between purple and cyan. These are printed as
  "visible" characters, e.g. ␣ for the space, ↵ for a newline.
  Hopefully this is an improvement over the usual practice of printing
  these as periods, like standard hex dumpers do. The Unicode BOM
  (byte order marker, U+FEFF) is printed as a purple letter B.

**red**
  Invalid UTF-8 sequences. These are rendered as � (U+0FFD) with
  a red background, to make them stand out. Invalid
  sequences are:

    - Prefix bytes (>= 0x80) which are not followed by the correct number of continuation
      bytes (with their high 2 bits set to **10**).

    - Continuation bytes that aren't preceded by a valid prefix byte.

    - Truncated UTF-8 sequence at EOF.

  Also, there are sequences that are valid UTF-8 encodings, but not valid Unicode.
  These are normally rendered with a red background.

    - UTF-16 surrogates (codepoints U+D800 to U+DFFF) [**\***]. Rendered as **S**.

    - Codepoints above U+10FFFF, which are disallowed by the Unicode standard [**\***].
      Rendered as **>**.

    - Overlong encodings (e.g. codepoints U+0000 to U+007F encoded
      as 2 or more bytes) [**\***]. Rendered as **O**.

  Each error-highlighted sequence will increment the "Bad
  sequences" count, if the **-i** option is used.

  For items marked with [**\***], the **-j**, **-p**, and/or **-w**
  options can disable error highlighting for this type of error. They
  will be displayed in purple or cyan rather than red.

TERMINAL SUPPORT
================

**uxd** should work with any modern terminal that supports color,
ANSI-style escape sequences, Unicode, and UTF-8 rendering.

The author's testing is done primarily with **urxvt**\(1). Other
terminals aren't tested as often. Some terminals may need UTF-8
enabled, if it's not on by default (either in the terminal's settings
or using the **-t** option to **uxd**).

Known to work: urxvt, xterm, st, xfce4-terminal, gnome-terminal,
kitty, konsole, the Linux console (but see **FONTS**, below).

Known **not** to work: rxvt (doesn't support Unicode at all), and its
derivatives such as aterm.

**uxd** also builds and runs correctly on a Mac running a recent
version of OSX with Terminal.app.

FONTS
=====

For the human-readable column to display correctly, you'll need a font
with lots of glyphs. Try *Deja Vu Sans Mono*, *Symbola*, or *Quivira*
(although it's not really a terminal font). If you use urxvt, it
searches for glyphs in multiple fonts, so you can use all of the above
at once.

Any glyph your font lacks, you'll see as a dotted box, or perhaps
a solid block. This isn't something **uxd** can do anything about;
you'll have to use a different font, or (if you use urxvt) add another
font to your URxvt*font resource.

The Linux console is capable of rendering UTF-8, but it's incapable
of displaying more than 512 glyphs. Most console fonts only define
256, since using more than 256 means the console won't be able to
do bold. Expect to see lots of solid or dotted boxes. This isn't
specifically a problem with **uxd**.

FILES
=====

**uxd** doesn't read any files other than the input file, and doesn't write to
any files other than standard output. There's no config file.

ENVIRONMENT
===========

**UXD_OPTS**
  If this is set, its value is treated as a set of options, which
  get applied before any command-line options (unless the command-line
  options inclue **-n**).

**NO_COLOR**
  If this is set (to any value), **uxd** runs in monochrome mode, just
  as though the **-m** option were given. This variable is also
  respected by **xxd**.

It's *not* necessary to have a UTF-8 locale set in e.g. **LANG** or
**LC_ALL**. Also, the **TERM** variable is not used.

EXIT STATUS
===========

Zero for success, non-zero for failure.

Failure status will only be returned if **uxd** failed to open the
input file. Invalid input (non-UTF-8) doesn't count as an error;
it'll just have lots of red in the output.

LIMITATIONS
===========

There are not bugs, because they're part of the design.

Only UTF-8 and a couple of variants (WTF-8, MUTF-8) are supported.
There is no support for UTF-16, UTF-32, UTF-EBCDIC, or any other
non-UTF-8 encoding.

There's no support for any number base except hex.

The input is read one byte at a time, so a search or regex match
option would be difficult or impossible to implement.

Seeking backwards from the end of the file is impossible when reading
from standard input. The only way to fake this would be to read the
whole file into memory at startup, which **uxd** doesn't do.

BUGS
====

Combining characters are not handled well. Or at all, really: the 2
characters being combined will have an ANSI color code in between.
urxvt at least ignores the color code, so the composite character
displays in the color of the first (non-combining) character. I'm not
sure what a better solution would be...

COPYRIGHT
=========

Licensed under the WTFPL. See http://www.wtfpl.net/txt/copying/ for details.

AUTHORS
=======

B. Watson <urchlay@slackware.uk>.

SEE ALSO
========

xxd(1), bvi(1), utf-8(7), unicode(7)