ruby/doc/encodings.rdoc

= Encodings

== The Basics

A {character encoding}[https://en.wikipedia.org/wiki/Character_encoding],
often shortened to _encoding_, is a mapping between:

- A sequence of 8-bit bytes (each byte in the range <tt>0..255</tt>).
- Characters in a specific character set.

Some character sets contain only 1-byte characters;
{US-ASCII}[https://en.wikipedia.org/wiki/ASCII], for example, has 256 1-byte characters.
This string, encoded in US-ASCII, has six characters that are stored as six bytes:

  s = 'Hello!'.encode('US-ASCII')  # => "Hello!"
  s.encoding                       # => #<Encoding:US-ASCII>
  s.bytes                          # => [72, 101, 108, 108, 111, 33]

Other encodings may involve multi-byte characters.
{UTF-8}[https://en.wikipedia.org/wiki/UTF-8], for example,
encodes more than one million characters, encoding each in one to four bytes.
The lowest-valued of these characters correspond to ASCII characters,
and so are 1-byte characters:

  s = 'Hello!' # => "Hello!"
  s.bytes      # => [72, 101, 108, 108, 111, 33]

Other characters, such as the Euro symbol, are multi-byte:

  s = "\u20ac" # => "€"
  s.bytes      # => [226, 130, 172]

== The \Encoding \Class

=== \Encoding Objects

Ruby encodings are defined by constants in class \Encoding.
There can be only one instance of \Encoding for each of these constants.
\Method Encoding.list returns an array of \Encoding objects (one for each constant):

  Encoding.list.size        # => 103
  Encoding.list.first.class # => Encoding
  Encoding.list.take(3)
  # => [#<Encoding:ASCII-8BIT>, #<Encoding:UTF-8>, #<Encoding:US-ASCII>]

=== Names and Aliases

\Method Encoding#name returns the name of an \Encoding:

  Encoding::ASCII_8BIT.name  # => "ASCII-8BIT"
  Encoding::WINDOWS_31J.name # => "Windows-31J"

An \Encoding object has zero or more aliases;
method Encoding#names returns an array containing the name and all aliases:

  Encoding::ASCII_8BIT.names
  # => ["ASCII-8BIT", "BINARY"]
  Encoding::WINDOWS_31J.names
  #=> ["Windows-31J", "CP932", "csWindows31J", "SJIS", "PCK"]

\Method Encoding.aliases returns a hash of all alias/name pairs:

  Encoding.aliases.size # => 71
  Encoding.aliases.take(3)
  # => [["BINARY", "ASCII-8BIT"], ["CP437", "IBM437"], ["CP720", "IBM720"]]

\Method Encoding.name_list returns an array of all the encoding names and aliases:

  Encoding.name_list.size # => 175
  Encoding.name_list.take(3)
  # => ["ASCII-8BIT", "UTF-8", "US-ASCII"]

\Method +name_list+ returns more entries than method +list+
because it includes both the names and their aliases.

\Method Encoding.find returns the \Encoding for a given name or alias, if it exists:

  Encoding.find("US-ASCII")       # => #<Encoding:US-ASCII>
  Encoding.find("US-ASCII").class # => Encoding

=== Default Encodings

\Method Encoding.find, above, also returns a default \Encoding
for each of these special names:

- +external+: the default external \Encoding:

    Encoding.find("external") # => #<Encoding:UTF-8>

- +internal+: the default internal \Encoding (may be +nil+):

    Encoding.find("internal") # => nil

- +locale+: the default \Encoding for a string from the environment:

    Encoding.find("locale") # => #<Encoding:UTF-8>  # Linux
    Encoding.find("locale") # => #<Encoding:IBM437> # Windows

- +filesystem+: the default \Encoding for a string from the filesystem:

    Encoding.find("filesystem") # => #<Encoding:UTF-8>

\Method Encoding.default_external returns the default external \Encoding:

  Encoding.default_external # => #<Encoding:UTF-8>

\Method Encoding.default_external= sets that value:

  Encoding.default_external = 'US-ASCII' # => "US-ASCII"
  Encoding.default_external              # => #<Encoding:US-ASCII>

\Method Encoding.default_internal returns the default internal \Encoding:

  Encoding.default_internal # => nil

\Method Encoding.default_internal= sets the default internal \Encoding:

  Encoding.default_internal = 'US-ASCII' # => "US-ASCII"
  Encoding.default_internal              # => #<Encoding:US-ASCII>

=== Compatible Encodings

\Method Encoding.compatible? returns whether two given objects are encoding-compatible
(that is, whether they can be concatenated);
returns the \Encoding of the concatenated string, or +nil+ if incompatible:

  rus = "\u{442 435 441 442}"
  eng = 'text'
  Encoding.compatible?(rus, eng) # => #<Encoding:UTF-8>

  s0 = "\xa1\xa1".force_encoding('iso-8859-1') # => "\xA1\xA1"
  s1 = "\xa1\xa1".force_encoding('euc-jp')     # => "\x{A1A1}"
  Encoding.compatible?(s0, s1)                 # => nil

== \String \Encoding

A Ruby String object has an encoding that is an instance of class \Encoding.
The encoding may be retrieved by method String#encoding.

The default encoding for a string literal is the script encoding;
see {Script Encoding}[rdoc-ref:encodings.rdoc@Script+Encoding].

  's'.encoding # => #<Encoding:UTF-8>

The default encoding for a string created with method String.new is:

- For a \String object argument, the encoding of that string.
- For a string literal, the script encoding;
  see {Script Encoding}[rdoc-ref:encodings.rdoc@Script+Encoding].

In either case, any encoding may be specified:

  s = String.new(encoding: 'UTF-8')             # => ""
  s.encoding                                    # => #<Encoding:UTF-8>
  s = String.new('foo', encoding: 'ASCII-8BIT') # => "foo"
  s.encoding                                    # => #<Encoding:ASCII-8BIT>

The encoding for a string may be changed:

  s = "R\xC3\xA9sum\xC3\xA9"     # => "Résumé"
  s.encoding                     # => #<Encoding:UTF-8>
  s.force_encoding('ISO-8859-1') # => "R\xC3\xA9sum\xC3\xA9"
  s.encoding                     # => #<Encoding:ISO-8859-1>

Changing the assigned encoding does not alter the content of the string;
it changes only the way the content is to be interpreted:

  s                         # => "R\xC3\xA9sum\xC3\xA9"
  s.force_encoding('UTF-8') # => "Résumé"

The actual content of a string may also be altered;
see {Transcoding a String}[#label-Transcoding+a+String].

Here are a couple of useful query methods:

  s = "abc".force_encoding("UTF-8")         # => "abc"
  s.ascii_only?                             # => true
  s = "abc\u{6666}".force_encoding("UTF-8") # => "abc晦"
  s.ascii_only?                             # => false

  s = "\xc2\xa1".force_encoding("UTF-8") # => "¡"
  s.valid_encoding?                      # => true
  s = "\xc2".force_encoding("UTF-8")     # => "\xC2"
  s.valid_encoding?                      # => false

== \Symbol and \Regexp Encodings

The string stored in a Symbol or Regexp object also has an encoding;
the encoding may be retrieved by method Symbol#encoding or Regexp#encoding.

The default encoding for these, however, is:

- US-ASCII, if all characters are US-ASCII.
- The script encoding, otherwise;
  see (Script Encoding)[rdoc-ref:encodings.rdoc@Script+Encoding].

== Filesystem \Encoding

The filesystem encoding is the default \Encoding for a string from the filesystem:

    Encoding.find("filesystem") # => #<Encoding:UTF-8>

== Locale \Encoding

The locale encoding is the default encoding for a string from the environment,
other than from the filesystem:

  Encoding.find('locale') # => #<Encoding:IBM437>

== Stream Encodings

Certain stream objects can have two encodings; these objects include instances of:

- IO.
- File.
- ARGF.
- StringIO.

The two encodings are:

- An _external_ _encoding_, which identifies the encoding of the stream.
- An _internal_ _encoding_, which (if not +nil+) specifies the encoding
  to be used for the string constructed from the stream.

=== External \Encoding

The external encoding, which is an \Encoding object, specifies how bytes read
from the stream are to be interpreted as characters.

The default external encoding is:

- UTF-8 for a text stream.
- ASCII-8BIT for a binary stream.

The default external encoding is returned by method Encoding.default_external,
and may be set by:

- Ruby command-line options <tt>--external_encoding</tt> or <tt>-E</tt>.

You can also set the default external encoding using method Encoding.default_external=,
but doing so may cause problems; strings created before and after the change
may have a different encodings.

For an \IO or \File object, the external encoding may be set by:

- Open options +external_encoding+ or +encoding+, when the object is created;
  see {Open Options}[rdoc-ref:IO@Open+Options].

For an \IO, \File, \ARGF, or \StringIO object, the external encoding may be set by:

- \Methods +set_encoding+ or (except for \ARGF) +set_encoding_by_bom+.

=== Internal \Encoding

The internal encoding, which is an \Encoding object or +nil+,
specifies how characters read from the stream
are to be converted to characters in the internal encoding;
those characters become a string whose encoding is set to the internal encoding.

The default internal encoding is +nil+ (no conversion).
It is returned by method Encoding.default_internal,
and may be set by:

- Ruby command-line options <tt>--internal_encoding</tt> or <tt>-E</tt>.

You can also set the default internal encoding using method Encoding.default_internal=,
but doing so may cause problems; strings created before and after the change
may have a different encodings.

For an \IO or \File object, the internal encoding may be set by:

- Open options +internal_encoding+ or +encoding+, when the object is created;
  see {Open Options}[rdoc-ref:IO@Open+Options].

For an \IO, \File, \ARGF, or \StringIO object, the internal encoding may be set by:

- \Method +set_encoding+.

== Script \Encoding

A Ruby script has a script encoding, which may be retrieved by:

  __ENCODING__ # => #<Encoding:UTF-8>

The default script encoding is UTF-8;
a Ruby source file may set its script encoding with a magic comment
on the first line of the file (or second line, if there is a shebang on the first).
The comment must contain the word +coding+ or +encoding+,
followed by a colon, space and the Encoding name or alias:

  # encoding: ISO-8859-1
  __ENCODING__ #=> #<Encoding:ISO-8859-1>

== Transcoding

_Transcoding_ is the process of changing a sequence of characters
from one encoding to another.

As far as possible, the characters remain the same,
but the bytes that represent them may change.

The handling for characters that cannot be represented in the destination encoding
may be specified by @Encoding+Options.

=== Transcoding a \String

Each of these methods transcodes a string:

- String#encode: Transcodes +self+ into a new string
  according to given encodings and options.
- String#encode!: Like String#encode, but transcodes +self+ in place.
- String#scrub: Transcodes +self+ into a new string
  by replacing invalid byte sequences with a given or default replacement string.
- String#scrub!: Like String#scrub, but transcodes +self+ in place.
- String#unicode_normalize: Transcodes +self+ into a new string
  according to Unicode normalization.
- String#unicode_normalize!: Like String#unicode_normalize,
  but transcodes +self+ in place.

== Transcoding a Stream

Each of these methods may transcode a stream;
whether it does so depends on the external and internal encodings:

- IO.foreach: Yields each line of given stream to the block.
- IO.new: Creates and returns a new \IO object for the given integer file descriptor.
- IO.open: Creates a new \IO object.
- IO.pipe: Creates a connected pair of reader and writer \IO objects.
- IO.popen: Creates an \IO object to interact with a subprocess.
- IO.read: Returns a string with all or a subset of bytes from the given stream.
- IO.readlines: Returns an array of strings, which are the lines from the given stream.
- IO.write: Writes a given string to the given stream.

This example writes a string to a file, encoding it as ISO-8859-1,
then reads the file into a new string, encoding it as UTF-8:

  s = "R\u00E9sum\u00E9"
  path = 't.tmp'
  ext_enc = 'ISO-8859-1'
  int_enc = 'UTF-8'

  File.write(path, s, external_encoding: ext_enc)
  raw_text = File.binread(path)

  transcoded_text = File.read(path, external_encoding: ext_enc, internal_encoding: int_enc)

  p raw_text
  p transcoded_text

Output:

  "R\xE9sum\xE9"
  "Résumé"

== \Encoding Options

A number of methods in the Ruby core accept keyword arguments as encoding options.

Some of the options specify or utilize a _replacement_ _string_, to be used
in certain transcoding operations.
A replacement string may be in any encoding that can be converted
to the encoding of the destination string.

These keyword-value pairs specify encoding options:

- For an invalid byte sequence:

  - <tt>:invalid: nil</tt> (default): Raise exception.
  - <tt>:invalid: :replace</tt>: Replace each invalid byte sequence
    with the replacement string.

  Examples:

    s = "\x80foo\x80"
    s.encode('ISO-8859-3') # Raises Encoding::InvalidByteSequenceError.
    s.encode('ISO-8859-3', invalid: :replace) # => "?foo?"

- For an undefined character:

  - <tt>:undef: nil</tt> (default): Raise exception.
  - <tt>:undef: :replace</tt>: Replace each undefined character
    with the replacement string.

  Examples:

    s = "\x80foo\x80"
    "\x80".encode('UTF-8', 'ASCII-8BIT') # Raises Encoding::UndefinedConversionError.
    s.encode('UTF-8', 'ASCII-8BIT', undef: :replace) # => "<22>foo<6F>"


- Replacement string:

  - <tt>:replace: nil</tt> (default): Set replacement string to default value:
    <tt>"\uFFFD"</tt> ("<22>") for a Unicode encoding, <tt>'?'</tt> otherwise.
  - <tt>:replace: _some_string_</tt>: Set replacement string to the given +some_string+;
    overrides +:fallback+.

  Examples:

    s = "\xA5foo\xA5"
    options = {:undef => :replace, :replace => 'xyzzy'}
    s.encode('UTF-8', 'ISO-8859-3', **options) # => "xyzzyfooxyzzy"

- Replacement fallback:

  One of these may be specified:

  - <tt>:fallback: nil</tt> (default): No replacement fallback.
  - <tt>:fallback: _hash_like_object_</tt>: Set replacement fallback to the given
    +hash_like_object+; the replacement string is <tt>_hash_like_object_[X]</tt>.
  - <tt>:fallback: _method_</tt>: Set replacement fallback to the given
    +method+; the replacement string is <tt>_method_(X)</tt>.
  - <tt>:fallback: _proc_</tt>: Set replacement fallback to the given
    +proc+; the replacement string is <tt>_proc_[X]</tt>.

  Examples:

    s = "\u3042foo\u3043"

    hash = {"\u3042" => 'xyzzy'}
    hash.default = 'XYZZY'
    s.encode('ASCII', fallback: hash) # => "xyzzyfooXYZZY"

    def (fallback = "U+%.4X").escape(x)
      self % x.unpack("U")
    end
    "\u{3042}".encode("US-ASCII", fallback: fallback.method(:escape)) # => "U+3042"

    proc = Proc.new {|x| x == "\u3042" ? 'xyzzy' : 'XYZZY' }
    s.encode('ASCII', fallback: proc) # => "XYZZYfooXYZZY"

- XML entities:

  One of these may be specified:

  - <tt>:xml: nil</tt> (default): No handling for XML entities.
  - <tt>:xml: :text</tt>: Treat source text as XML;
    replace each undefined character
    with its upper-case hexdecimal numeric character reference,
    except that:

    - <tt>&</tt> is replaced with <tt>&amp;</tt>.
    - <tt><</tt> is replaced with <tt>&lt;</tt>.
    - <tt>></tt> is replaced with <tt>&gt;</tt>.

  - <tt>:xml: :attr</tt>: Treat source text as XML attribute value;
    replace each undefined character
    with its upper-case hexdecimal numeric character reference,
    except that:

    - The replacement string <tt>r</tt> is double-quoted (<tt>"r"</tt>).
    - Each embedded double-quote is replaced with <tt>&quot;</tt>.
    - <tt>&</tt> is replaced with <tt>&amp;</tt>.
    - <tt><</tt> is replaced with <tt>&lt;</tt>.
    - <tt>></tt> is replaced with <tt>&gt;</tt>.

  Examples:

    s = 'foo"<&>"bar' + "\u3042"
    s.encode('ASCII', xml: :text) # => "foo\"&lt;&amp;&gt;\"bar&#x3042;"
    s.encode('ASCII', xml: :attr) # => "\"foo&quot;&lt;&amp;&gt;&quot;bar&#x3042;\""


- Newlines:

  One of these may be specified:

  - <tt>:cr_newline: true</tt>: Replace each line-feed character (<tt>"\n"</tt>)
    with a carriage-return character (<tt>"\r"</tt>).
  - <tt>:crlf_newline: true</tt>: Replace each line-feed character (<tt>"\n"</tt>)
    with a carriage-return/line-feed string (<tt>"\r\n"</tt>).
  - <tt>:universal_newline: true</tt>: Replace each carriage-return
    character (<tt>"\r"</tt>) and each carriage-return/line-feed string
    (<tt>"\r\n"</tt>) with a line-feed character (<tt>"\n"</tt>).

  Examples:

    s = "\n \r \r\n"                           # => "\n \r \r\n"
    s.encode('ASCII', cr_newline: true)        # => "\r \r \r\r"
    s.encode('ASCII', crlf_newline: true)      # => "\r\n \r \r\r\n"
    s.encode('ASCII', universal_newline: true) # => "\n \n \n"