HTML::Encoding(3) User Contributed Perl Documentation HTML::Encoding(3)NAMEHTML::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.
DESCRIPTIONHTML::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.
perl v5.14.1 2011-06-21 HTML::Encoding(3)
