Provided by: libhtml-encoding-perl_0.61-1_all 
NAME
HTML::Encoding - Determine the encoding of HTML/XML/XHTML documents
SYNOPSIS
use HTML::Encoding 'encoding_from_http_message';
use LWP::UserAgent;
use Encode;
my $resp = LWP::UserAgent->new->get('http://www.example.org');
my $enco = encoding_from_http_message($resp);
my $utf8 = decode($enco => $resp->content);
WARNING
The interface and implementation are guranteed to change before this module reaches
version 1.00! Please send feedback to the author of this module.
DESCRIPTION
HTML::Encoding helps to determine the encoding of HTML and XML/XHTML documents...
DEFAULT ENCODINGS
Most routines need to know some suspected character encodings which can be provided
through the "encodings" option. This option always defaults to the
$HTML::Encoding::DEFAULT_ENCODINGS array reference which means the following encodings are
considered by default:
* ISO-8859-1
* UTF-16LE
* UTF-16BE
* UTF-32LE
* UTF-32BE
* UTF-8
If you change the values or pass custom values to the routines note that Encode must
support them in order for this module to work correctly.
ENCODING SOURCES
"encoding_from_xml_document", "encoding_from_html_document", and
"encoding_from_http_message" return in list context the encoding source and the encoding
name, possible encoding sources are
* protocol (Content-Type: text/html;charset=encoding)
* bom (leading U+FEFF)
* xml (<?xml version='1.0' encoding='encoding'?>)
* meta (<meta http-equiv=...)
* default (default fallback value)
* protocol_default (protocol default)
ROUTINES
Routines exported by this module at user option. By default, nothing is exported.
encoding_from_content_type($content_type)
Takes a byte string and uses HTTP::Headers::Util to extract the charset parameter from
the "Content-Type" header value and returns its value or "undef" (or an empty list in
list context) if there is no such value. Only the first component will be examined
(HTTP/1.1 only allows for one component), any backslash escapes in strings will be
unescaped, all leading and trailing quote marks and white-space characters will be
removed, all white-space will be collapsed to a single space, empty charset values will
be ignored and no case folding is performed.
Examples:
+-----------------------------------------+-----------+
| encoding_from_content_type(...) | returns |
+-----------------------------------------+-----------+
| "text/html" | undef |
| "text/html,text/plain;charset=utf-8" | undef |
| "text/html;charset=" | undef |
| "text/html;charset=\"\\u\\t\\f\\-\\8\"" | 'utf-8' |
| "text/html;charset=utf\\-8" | 'utf\\-8' |
| "text/html;charset='utf-8'" | 'utf-8' |
| "text/html;charset=\" UTF-8 \"" | 'UTF-8' |
+-----------------------------------------+-----------+
If you pass a string with the UTF-8 flag turned on the string will be converted to bytes
before it is passed to HTTP::Headers::Util. The return value will thus never have the
UTF-8 flag turned on (this might change in future versions).
encoding_from_byte_order_mark($octets [, %options])
Takes a sequence of octets and attempts to read a byte order mark at the beginning of
the octet sequence. It will go through the list of $options{encodings} or the list of
default encodings if no encodings are specified and match the beginning of the string
against any byte order mark octet sequence found.
The result can be ambiguous, for example qq(\xFF\xFE\x00\x00) could be both, a complete
BOM in UTF-32LE or a UTF-16LE BOM followed by a U+0000 character. It is also possible
that $octets starts with something that looks like a byte order mark but actually is
not.
encoding_from_byte_order_mark sorts the list of possible encodings by the length of
their BOM octet sequence and returns in scalar context only the encoding with the
longest match, and all encodings ordered by length of their BOM octet sequence in list
context.
Examples:
+-------------------------+------------+-----------------------+
| Input | Encodings | Result |
+-------------------------+------------+-----------------------+
| "\xFF\xFE\x00\x00" | default | qw(UTF-32LE) |
| "\xFF\xFE\x00\x00" | default | qw(UTF-32LE UTF-16LE) |
| "\xEF\xBB\xBF" | default | qw(UTF-8) |
| "Hello World!" | default | undef |
| "\xDD\x73\x66\x73" | default | undef |
| "\xDD\x73\x66\x73" | UTF-EBCDIC | qw(UTF-EBCDIC) |
| "\x2B\x2F\x76\x38\x2D" | default | undef |
| "\x2B\x2F\x76\x38\x2D" | UTF-7 | qw(UTF-7) |
+-------------------------+------------+-----------------------+
Note however that for UTF-7 it is in theory possible that the U+FEFF combines with other
characters in which case such detection would fail, for example consider:
+--------------------------------------+-----------+-----------+
| Input | Encodings | Result |
+--------------------------------------+-----------+-----------+
| "\x2B\x2F\x76\x38\x41\x39\x67\x2D" | default | undef |
| "\x2B\x2F\x76\x38\x41\x39\x67\x2D" | UTF-7 | undef |
+--------------------------------------+-----------+-----------+
This might change in future versions, although this is not very relevant for most
applications as there should never be need to use UTF-7 in the encoding list for
existing documents.
If no BOM can be found it returns "undef" in scalar context and an empty list in list
context. This routine should not be used with strings with the UTF-8 flag turned on.
encoding_from_xml_declaration($declaration)
Attempts to extract the value of the encoding pseudo-attribute in an XML declaration or
text declaration in the character string $declaration. If there does not appear to be
such a value it returns nothing. This would typically be used with the return values of
xml_declaration_from_octets. Normalizes whitespaces like encoding_from_content_type.
Examples:
+-------------------------------------------+---------+
| encoding_from_xml_declaration(...) | Result |
+-------------------------------------------+---------+
| "<?xml version='1.0' encoding='utf-8'?>" | 'utf-8' |
| "<?xml encoding='utf-8'?>" | 'utf-8' |
| "<?xml encoding=\"utf-8\"?>" | 'utf-8' |
| "<?xml foo='bar' encoding='utf-8'?>" | 'utf-8' |
| "<?xml encoding='a' encoding='b'?>" | 'a' |
| "<?xml encoding=' a b '?>" | 'a b' |
| "<?xml-stylesheet encoding='utf-8'?>" | undef |
| " <?xml encoding='utf-8'?>" | undef |
| "<?xml encoding =\x{2028}'utf-8'?>" | 'utf-8' |
| "<?xml version='1.0' encoding=utf-8?>" | undef |
| "<?xml x='encoding=\"a\"' encoding='b'?>" | 'a' |
+-------------------------------------------+---------+
Note that encoding_from_xml_declaration() determines the encoding even if the XML
declaration is not well-formed or violates other requirements of the relevant XML
specification as long as it can find an encoding pseudo-attribute in the provided
string. This means XML processors must apply further checks to determine whether the
entity is well-formed, etc.
xml_declaration_from_octets($octets [, %options])
Attempts to find a ">" character in the byte string $octets using the encodings in
$encodings and upon success attempts to find a preceding "<" character. Returns all the
strings found this way in the order of number of successful matches in list context and
the best match in scalar context. Should probably be combined with the only user of this
routine, encoding_from_xml_declaration... You can modify the list of suspected encodings
using $options{encodings};
encoding_from_first_chars($octets [, %options])
Assuming that documents start with "<" optionally preceded by whitespace characters,
encoding_from_first_chars attempts to determine an encoding by matching $octets against
something like /^[@{$options{whitespace}}]*</ in the various suspected
$options{encodings}.
This is useful to distinguish e.g. UTF-16LE from UTF-8 if the byte string does not start
with a byte order mark nor an XML declaration (e.g. if the document is a HTML document)
to get at least a base encoding which can be used to decode enough of the document to
find <meta> elements using encoding_from_meta_element. $options{whitespace} defaults to
qw/CR LF SP TB/. Returns nothing if unsuccessful. Returns the matching encodings in
order of the number of octets matched in list context and the best match in scalar
context.
Examples:
+---------------+----------+---------------------+
| String | Encoding | Result |
+---------------+----------+---------------------+
| '<!DOCTYPE ' | UTF-16LE | UTF-16LE |
| ' <!DOCTYPE ' | UTF-16LE | UTF-16LE |
| '...' | UTF-16LE | undef |
| '...<' | UTF-16LE | undef |
| '<' | UTF-8 | ISO-8859-1 or UTF-8 |
| "<!--\xF6-->" | UTF-8 | ISO-8859-1 or UTF-8 |
+---------------+----------+---------------------+
encoding_from_meta_element($octets, $encname [, %options])
Attempts to find <meta> elements in the document using HTML::Parser. It will attempt to
decode chunks of the byte string using $encname to characters before passing the data to
HTML::Parser. An optional %options hash can be provided which will be passed to the
HTML::Parser constructor. It will stop processing the document if it encounters
* </head>
* encoding errors
* the end of the input
* ... (see todo)
If relevant <meta> elements, i.e. something like
<meta http-equiv=Content-Type content='...'>
are found, uses encoding_from_content_type to extract the charset parameter. It returns
all such encodings it could find in document order in list context or the first encoding
in scalar context (it will currently look for others regardless of calling context) or
nothing if that fails for some reason.
Note that there are many edge cases where this does not yield in "proper" results
depending on the capabilities of the HTML::Parser version and the options you pass for
it, for example,
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN" [
<!ENTITY content_type "text/html;charset=utf-8">
]>
<meta http-equiv="Content-Type" content="&content_type;">
<title></title>
<p>...</p>
This would likely not detect the "utf-8" value if HTML::Parser does not resolve the
entity. This should however only be a concern for documents specifically crafted to
break the encoding detection.
encoding_from_xml_document($octets, [, %options])
Uses encoding_from_byte_order_mark to detect the encoding using a byte order mark in the
byte string and returns the return value of that routine if it succeeds. Uses
xml_declaration_from_octets and encoding_from_xml_declaration and returns the encoding
for which the latter routine found most matches in scalar context, and all encodings
ordered by number of occurences in list context. It does not return a value of neither
byte order mark not inbound declarations declare a character encoding.
Examples:
+----------------------------+----------+-----------+----------+
| Input | Encoding | Encodings | Result |
+----------------------------+----------+-----------+----------+
| "<?xml?>" | UTF-16 | default | UTF-16BE |
| "<?xml?>" | UTF-16LE | default | undef |
| "<?xml encoding='utf-8'?>" | UTF-16LE | default | utf-8 |
| "<?xml encoding='utf-8'?>" | UTF-16 | default | UTF-16BE |
| "<?xml encoding='cp37'?>" | CP37 | default | undef |
| "<?xml encoding='cp37'?>" | CP37 | CP37 | cp37 |
+----------------------------+----------+-----------+----------+
Lacking a return value from this routine and higher-level protocol information (such as
protocol encoding defaults) processors would be required to assume that the document is
UTF-8 encoded.
Note however that the return value depends on the set of suspected encodings you pass to
it. For example, by default, EBCDIC encodings would not be considered and thus for
<?xml version='1.0' encoding='cp37'?>
this routine would return the undefined value. You can modify the list of suspected
encodings using $options{encodings}.
encoding_from_html_document($octets, [, %options])
Uses encoding_from_xml_document and encoding_from_meta_element to determine the encoding
of HTML documents. If $options{xhtml} is set to a false value uses
encoding_from_byte_order_mark and encoding_from_meta_element to determine the encoding.
The xhtml option is on by default. The $options{encodings} can be used to modify the
suspected encodings and $options{parser_options} can be used to modify the HTML::Parser
options in encoding_from_meta_element (see the relevant documentation).
Returns nothing if no declaration could be found, the winning declaration in scalar
context and a list of encoding source and encoding name in list context, see ENCODING
SOURCES.
...
Other problems arise from differences between HTML and XHTML syntax and encoding
detection rules, for example, the input could be
Content-Type: text/html
<?xml version='1.0' encoding='utf-8'?>
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN"
"http://www.w3.org/TR/html4/strict.dtd">
<meta http-equiv = "Content-Type"
content = "text/html;charset=iso-8859-2">
<title></title>
<p>...</p>
This is a perfectly legal HTML 4.01 document and implementations might be expected to
consider the document ISO-8859-2 encoded as XML rules for encoding detection do not
apply to HTML documents. This module attempts to avoid making decisions which rules
apply for a specific document and would thus by default return 'utf-8' for this input.
On the other hand, if the input omits the encoding declaration,
Content-Type: text/html
<?xml version='1.0'?>
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN"
"http://www.w3.org/TR/html4/strict.dtd">
<meta http-equiv = "Content-Type"
content = "text/html;charset=iso-8859-2">
<title></title>
<p>...</p>
It would return 'iso-8859-2'. Similar problems would arise from other differences
between HTML and XHTML, for example consider
Content-Type: text/html
<?foo >
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html ...
?>
...
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN">
...
If this is processed using HTML rules, the first > will end the processing instruction
and the XHTML document type declaration would be the relevant declaration for the
document, if it is processed using XHTML rules, the ?> will end the processing
instruction and the HTML document type declaration would be the relevant declaration.
IOW, an application would need to assume a certain character encoding (family) to
process enough of the document to determine whether it is XHTML or HTML and the result
of this detection would depend on which processing rules are assumed in order to process
it. It is thus in essence not possible to write a "perfect" detection algorithm, which
is why this routine attempts to avoid making any decisions on this matter.
encoding_from_http_message($message [, %options])
Determines the encoding of HTML / XML / XHTML documents enclosed in HTTP message.
$message is an object compatible to HTTP::Message, e.g. a HTTP::Response object.
%options is a hash with the following possible entries:
encodings
array references of suspected character encodings, defaults to
$HTML::Encoding::DEFAULT_ENCODINGS.
is_html
Regular expression matched against the content_type of the message to determine
whether to use HTML rules for the entity body, defaults to "qr{^text/html$}i".
is_xml
Regular expression matched against the content_type of the message to determine
whether to use XML rules for the entity body, defaults to "qr{^.+/(?:.+\+)?xml$}i".
is_text_xml
Regular expression matched against the content_type of the message to determine
whether to use text/html rules for the message, defaults to
"qr{^text/(?:.+\+)?xml$}i". This will only be checked if is_xml matches aswell.
html_default
Default encoding for documents determined (by is_html) as HTML, defaults to
"ISO-8859-1".
xml_default
Default encoding for documents determined (by is_xml) as XML, defaults to "UTF-8".
text_xml_default
Default encoding for documents determined (by is_text_xml) as text/xml, defaults to
"undef" in which case the default is ignored. This should be set to "US-ASCII" if
desired as this module is by default inconsistent with RFC 3023 which requires that
for text/xml documents without a charset parameter in the HTTP header "US-ASCII" is
assumed.
This requirement is inconsistent with RFC 2616 (HTTP/1.1) which requires to assume
"ISO-8859-1", has been widely ignored and is thus disabled by default.
xhtml
Whether the routine should look for an encoding declaration in the XML declaration of
the document (if any), defaults to 1.
default
Whether the relevant default value should be returned when no other information can be
determined, defaults to 1.
This is furhter possibly inconsistent with XML MIME types that differ in other ways from
application/xml, for example if the MIME Type does not allow for a charset parameter in
which case applications might be expected to ignore the charset parameter if erroneously
provided.
EBCDIC SUPPORT
By default, this module does not support EBCDIC encodings. To enable support for EBCDIC
encodings you can either change the $HTML::Encodings::DEFAULT_ENCODINGS array reference or
pass the encodings to the routines you use using the encodings option, for example
my @try = qw/UTF-8 UTF-16LE cp500 posix-bc .../;
my $enc = encoding_from_xml_document($doc, encodings => \@try);
Note that there are some subtle differences between various EBCDIC encodings, for example
"!" is mapped to 0x5A in "posix-bc" and to 0x4F in "cp500"; these differences might affect
processing in yet undetermined ways.
TODO
* bundle with test suite
* optimize some routines to give up once successful
* avoid transcoding for HTML::Parser if e.g. ISO-8859-1
* consider adding a "HTML5" modus of operation?
SEE ALSO
* http://www.w3.org/TR/REC-xml/#charencoding
* http://www.w3.org/TR/REC-xml/#sec-guessing
* http://www.w3.org/TR/xml11/#charencoding
* http://www.w3.org/TR/xml11/#sec-guessing
* http://www.w3.org/TR/html4/charset.html#h-5.2.2
* http://www.w3.org/TR/xhtml1/#C_9
* http://www.ietf.org/rfc/rfc2616.txt
* http://www.ietf.org/rfc/rfc2854.txt
* http://www.ietf.org/rfc/rfc3023.txt
* perlunicode
* Encode
* HTML::Parser
AUTHOR / COPYRIGHT / LICENSE
Copyright (c) 2004-2008 Bjoern Hoehrmann <bjoern@hoehrmann.de>.
This module is licensed under the same terms as Perl itself.