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 [*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
=======

There are no options yet.

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

The output is designed to fit in an 80-column terminal.

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
======

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

**purple**
  Spaces and unprintable characters ("control" characters, newlines, tabs, etc).
  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 with a red foreground, to make them
  stand out. Examples of invalid sequences:

    - 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.

    - Codepoints above U+10FFFF, which are disallowed by RFC 3629.

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 (e.g. xterm).

Known to work: urxvt, xterm, st, xfce4-terminal, gnome-terminal, kitty, 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 (though I'm not sure what terminal was used).

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** doesn't read anything from the environment. 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 should 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.

BUGS
====

**uxd** doesn't check for overlong UTF-8 encodings (e.g. a character
that could be a 1-byte sequence, but is encoded as 2 or more).
Sequences like this really should be colorized in red. Technically,
this means **uxd** supports WTF-8, not UTF-8.

There should be options and/or a config file to change the colors,
rather than baking them into the binary.

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)