Encoding names are case insensitive. White space in names
is ignored. In addition, an encoding may have aliases.
Each encoding has one ``canonical'' name. The ``canonical''
name is chosen from the names of the encoding by picking
the first in the following sequence (with a few exceptions).
The name used by the Perl community. That includes 'utf8' and 'ascii'.
Unlike aliases, canonical names directly reach the method so such
frequently used words like 'utf8' don't need to do alias lookups.
The MIME name as defined in IETF RFCs. This includes all ``iso-''s.
The name in the IANA registry.
The name used by the organization that defined it.
In case de jure canonical names differ from that of the Encode module, they are always aliased if it ever be implemented. So you can
safely tell if a given encoding is implemented or not just by passing
the canonical name.
Because of all the alias issues, and because in the general case
encodings have state, ``Encode'' uses an encoding object internally
once an operation is in progress.
As of Perl 5.8.0, at least the following encodings are recognized.
Note that unless otherwise specified, they are all case insensitive
(via alias) and all occurrence of spaces are replaced with '-'.
In other words, ``ISO 8859 1'' and ``iso-8859-1'' are identical.
Encodings are categorized and implemented in several different modules
but you don't have to use Encode::XX to make them available for
most cases. Encode.pm will automatically load those modules on demand.
perldoc2tree.cgi: /usr/lib/perl5/5.8.8/Encode/Supported.pod: cannot resolve L in paragraph 24.
null and ascii-ctrl are special. ``null'' fails for all character
so when you set fallback mode to PERLQQ, HTMLCREF or XMLCREF, ALL
CHARACTERS will fall back to character references. Ditto for
``ascii-ctrl'' except for control characters. For fallback modes, see
Encode.
Encode::Byte implements most single-byte encodings except for
Symbols and EBCDIC. The following encodings are based on single-byte
encodings implemented as extended ASCII. Most of them map
\x80-\xff (upper half) to non-ASCII characters.
Since there are so many, they are presented in table format with
languages and corresponding encoding names by vendors. Note that
the table is sorted in order of ISO-8859 and the corresponding vendor
mappings are slightly different from that of ISO. See
http://czyborra.com/charsets/iso8859.html for details.
Lang/Regions ISO/Other Std. DOS Windows Macintosh Others
----------------------------------------------------------------
N. America (ASCII) cp437 AdobeStandardEncoding
cp863 (DOSCanadaF)
W. Europe iso-8859-1 cp850 cp1252 MacRoman nextstep
hp-roman8
cp860 (DOSPortuguese)
Cntrl. Europe iso-8859-2 cp852 cp1250 MacCentralEurRoman
MacCroatian
MacRomanian
MacRumanian
Latin3[1] iso-8859-3
Latin4[2] iso-8859-4
Cyrillics iso-8859-5 cp855 cp1251 MacCyrillic
(See also next section) cp866 MacUkrainian
Arabic iso-8859-6 cp864 cp1256 MacArabic
cp1006 MacFarsi
Greek iso-8859-7 cp737 cp1253 MacGreek
cp869 (DOSGreek2)
Hebrew iso-8859-8 cp862 cp1255 MacHebrew
Turkish iso-8859-9 cp857 cp1254 MacTurkish
Nordics iso-8859-10 cp865
cp861 MacIcelandic
MacSami
Thai iso-8859-11[3] cp874 MacThai
(iso-8859-12 is nonexistent. Reserved for Indics?)
Baltics iso-8859-13 cp775 cp1257
Celtics iso-8859-14
Latin9 [4] iso-8859-15
Latin10 iso-8859-16
Vietnamese viscii cp1258 MacVietnamese
----------------------------------------------------------------
[1] Esperanto, Maltese, and Turkish. Turkish is now on 8859-9.
[2] Baltics. Now on 8859-10, except for Latvian.
[3] TIS 620 + Non-Breaking Space (0xA0 / U+00A0)
[4] Nicknamed Latin0; the Euro sign as well as French and Finnish
letters that are missing from 8859-1 were added.
perldoc2tree.cgi: /usr/lib/perl5/5.8.8/Encode/Supported.pod: cannot resolve L in paragraph 40.
Though ISO-8859 does have ISO-8859-5, the KOI8 series is far more
popular in the Net. Encode comes with the following KOI charsets.
For gory details, see http://czyborra.com/charsets/cyrillic.html
GSM0338 is for GSM handsets. Though it shares alphanumerals with
ASCII, control character ranges and other parts are mapped very
differently, mainly to store Greek characters. There are also escape
sequences (starting with 0x1B) to cover e.g. the Euro sign. Some
special cases like a trailing 0x00 byte or a lone 0x1B byte are not
well-defined and decode() will return an empty string for them.
One possible workaround is
Note that the Encode implementation of GSM0338 does not implement the
reuse of Latin capital letters as Greek capital letters (for example,
the 0x5A is U+005A (LATIN CAPITAL LETTER Z), not U+0396 (GREEK CAPITAL
LETTER ZETA).
The GSM0338 is also covered in Encode::Byte even though it is not
an ``extended ASCII'' encoding.
Note that Vietnamese is listed above. Also read ``Encoding vs Charset''
below. Also note that these are implemented in distinct modules by
countries, due to the size concerns (simplified Chinese is mapped
to 'CN', continental China, while traditional Chinese is mapped to
'TW', Taiwan). Please refer to their respective documentation pages.
Standard DOS/Win Macintosh Comment/Reference
----------------------------------------------------------------
euc-cn [1] MacChineseSimp
(gbk) cp936 [2]
gb12345-raw { GB12345 without CES }
gb2312-raw { GB2312 without CES }
hz
iso-ir-165
----------------------------------------------------------------
[1] GB2312 is aliased to this. See L<Microsoft-related naming mess>
[2] gbk is aliased to this. See L<Microsoft-related naming mess>
Standard DOS/Win Macintosh Comment/Reference
----------------------------------------------------------------
euc-jp
shiftjis cp932 macJapanese
7bit-jis
iso-2022-jp [RFC1468]
iso-2022-jp-1 [RFC2237]
jis0201-raw { JIS X 0201 (roman + halfwidth kana) without CES }
jis0208-raw { JIS X 0208 (Kanji + fullwidth kana) without CES }
jis0212-raw { JIS X 0212 (Extended Kanji) without CES }
----------------------------------------------------------------
Standard DOS/Win Macintosh Comment/Reference
----------------------------------------------------------------
euc-kr MacKorean [RFC1557]
cp949 [1]
iso-2022-kr [RFC1557]
johab [KS X 1001:1998, Annex 3]
ksc5601-raw { KSC5601 without CES }
----------------------------------------------------------------
[1] ks_c_5601-1987, (x-)?windows-949, and uhc are aliased to this.
See below.
Standard DOS/Win Macintosh Comment/Reference
----------------------------------------------------------------
big5-eten cp950 MacChineseTrad {big5 aliased to big5-eten}
big5-hkscs
----------------------------------------------------------------
Due to the size concerns, additional Chinese encodings below are
distributed separately on CPAN, under the name Encode::HanExtra.
Standard DOS/Win Macintosh Comment/Reference
----------------------------------------------------------------
big5ext CMEX's Big5e Extension
big5plus CMEX's Big5+ Extension
cccii Chinese Character Code for Information Interchange
euc-tw EUC (Extended Unix Character)
gb18030 GBK with Traditional Characters
----------------------------------------------------------------
Strictly speaking, MIME header encoding documented in RFC 2047 is more
of encapsulation than encoding. However, their support in modern
world is imperative so they are supported.
This one is not a name of encoding but a utility that lets you pick up
the most appropriate encoding for a data out of given suspects. See
the Encode::Guess manpage for details.
The following encodings are not supported as yet; some because they
are rarely used, some because of technical difficulties. They may
be supported by external modules via CPAN in the future, however.
Not very popular yet. Needs Unicode Database or equivalent to
implement encode() (because it includes JIS X 0208/0212, KSC5601, and
GB2312 simultaneously, whose code points in Unicode overlap. So you
need to lookup the database to determine to what character set a given
Unicode character should belong).
Not very popular. Needs CNS 11643-1 and -2 which are not available in
this module. CNS 11643 is supported (via euc-tw) in Encode::HanExtra.
Autrijus Tang may add support for this encoding in his module in future.
None of the Encode team knows Hebrew enough (ISO-8859-8, cp1255 and
MacHebrew are supported because and just because there were mappings
available at http://www.unicode.org/). Contributions welcome.
The maps for the following are available at http://www.unicode.org/
but remain unsupport because those encodings need algorithmical
approach, currently unsupported by enc2xs:
I believe this issue is prevalent not only for Mac Indics but also in
other Indic encodings, but the above were the only Indic encodings
maps that I could find at http://www.unicode.org/ .
We are used to using the term (character) encoding and character
set interchangeably. But just as confusing the terms byte and
character is dangerous and the terms should be differentiated when
needed, we need to differentiate encoding and character set.
To understand that, here is a description of how we make computers
grok our characters.
First we start with which characters to include. We call this
collection of characters character repertoire.
Then we have to give each character a unique ID so your computer can
tell the difference between 'a' and 'A'. This itemized character
repertoire is now a character set.
If your computer can grow the character set without further
processing, you can go ahead and use it. This is called a coded
character set (CCS) or raw character encoding. ASCII is used this
way for most cases.
But in many cases, especially multi-byte CJK encodings, you have to
tweak a little more. Your network connection may not accept any data
with the Most Significant Bit set, and your computer may not be able to
tell if a given byte is a whole character or just half of it. So you
have to encode the character set to use it.
A character encoding scheme (CES) determines how to encode a given
character set, or a set of multiple character sets. 7bit ISO-2022 is
an example of a CES. You switch between character sets via escape
sequences.
Technically, or mathematically, speaking, a character set encoded in
such a CES that maps character by character may form a CCS. EUC is such
an example. The CES of EUC is as follows:
Map ASCII unchanged.
Map such a character set that consists of 94 or 96 powered by N
members by adding 0x80 to each byte.
You can also use 0x8e and 0x8f to indicate that the following sequence of
characters belongs to yet another character set. To each following byte
is added the value 0x80.
perldoc2tree.cgi: /usr/lib/perl5/5.8.8/Encode/Supported.pod: cannot resolve L in paragraph 135.
By carefully looking at the encoded byte sequence, you can find that the
byte sequence conforms a unique number. In that sense, EUC is a CCS
generated by a CES above from up to four CCS (complicated?). UTF-8
falls into this category. See perlUnicode/``UTF-8'' to find out how
UTF-8 maps Unicode to a byte sequence.
You may also have found out by now why 7bit ISO-2022 cannot comprise
a CCS. If you look at a byte sequence \x21\x21, you can't tell if it is two !'s or IDEOGRAPHIC SPACE. EUC maps the latter to \xA1\xA1
so you have no trouble differentiating between ``!!''. and `` ''.
This section tries to classify the supported encodings by their
applicability for information exchange over the Internet and to
choose the most suitable aliases to name them in the context of
such communication.
To (en|de)code encodings marked by (**), you need
Encode::HanExtra, available from CPAN.
are IANA-registered charsets. See [RFC 2781] for details.
Jungshik Shin reports that UTF-16 with a BOM is well accepted
by MS IE 5/6 and NS 4/6. Beware however that
UTF-16 support in any software you're going to be
using/interoperating with has probably been less tested
then UTF-8 support
UTF-8 coded data seamlessly passes traditional
command piping (cat, more, etc.) while UTF-16 coded
data is likely to cause confusion (with its zero bytes,
for example)
it is beyond the power of words to describe the way HTML browsers
encode non-ASCII form data. To get a general impression, visit
http://ppewww.ph.gla.ac.uk/~flavell/charset/form-i18n.html.
While encoding of form data has stabilized for UTF-8 encoded pages
(at least IE 5/6, NS 6, and Opera 6 behave consistently), be sure to
expect fun (and cross-browser discrepancies) with UTF-16 encoded
pages!
The rule of thumb is to use UTF-8 unless you know what
you're doing and unless you really benefit from using UTF-16.
are totally valid encodings but not registered at IANA.
The names under which they are listed here are probably the
most widely-known names for these encodings and are recommended
names.
GB2312 has been registered in the EUC-CN meaning at
IANA. This has partially repaired the situation: Microsoft's
GB2312 has become a superset of the official GB2312.
Encode aliases GB2312 to euc-cn in full agreement with
IANA registration. cp936 is supported separately.
RawGB_2312-80encoding is available as gb2312-raw.
JIS has not endorsed the full Microsoft standard however.
The official Shift_JIS includes only JIS X 0201 and JIS X 0208
character sets, while Microsoft has always used Shift_JIS
to encode a wider character repertoire. See IANA registration for
Windows-31J.
As a historical predecessor, Microsoft's variant
probably has more rights for the name, though it may be objected
that Microsoft shouldn't have used JIS as part of the name
in the first place.
Unambiguous name: CP932. IANA name (also used by Mozilla, and
provided as an alias by Encode): Windows-31J.
An algorithm to map a character set to a byte sequence. You don't
have to be able to tell which character set a given byte sequence
belongs. 7-bit ISO-2022 is a CES but it cannot be a CCS. EUC is an
example of being both a CCS and CES.
has long been used in the meaning of encoding, CES.
While the word combination character set has lost this meaning
in MIME context since [RFC 2130], the charset abbreviation has
retained it. This is how [RFC 2277] and [RFC 2278] bless charset:
This document uses the term "charset" to mean a set of rules for
mapping from a sequence of octets to a sequence of characters, such
as the combination of a coded character set and a character encoding scheme; this is also what is used as an identifier in MIME "charset="
parameters, and registered in the IANA charset registry ... (Note
that this is NOT a term used by other standards bodies, such as ISO).
[RFC 2277]
A CES that was carefully designed to coexist with ASCII. There are a 7
bit version and an 8 bit version.
The 7 bit version switches character set via escape sequence so it
cannot form a CCS. Since this is more difficult to handle in programs
than the 8 bit version, the 7 bit version is not very popular except for
iso-2022-jp, the de facto standard CES for e-mails.
The 8 bit version can form a CCS. EUC and ISO-8859 are two examples
thereof. Pre-5.6 perl could use them as string literals.
A character set that aims to include all character repertoires of the
world. Many character sets in various national as well as industrial
standards have become, in a way, just subsets of Unicode.
A UTF in 16-bit encoding. Can either be in big endian or little
endian. The big endian version is called UTF-16BE (equal to UCS-2 +
surrogate support) and the little endian version is called UTF-16LE.
Most of the canonical names in Encode derive from this list
so you can directly apply the string you have extracted from MIME
header of mails and web pages.
Features a comprehensive coverage of CJKV character sets and
encodings along with many other issues faced by anyone trying
to better support CJKV languages/scripts in all the areas of
information processing.
