Packed Data¶ ↑

Certain Ruby core methods deal with packing and unpacking data:

Method Array#pack: Formats each element in array self into a binary string; returns that string.
Method String#unpack: Extracts data from string self, forming objects that become the elements of a new array; returns that array.
Method String#unpack1: Does the same, but unpacks and returns only the first extracted object.

Each of these methods accepts a string template, consisting of zero or more directive characters, each followed by zero or more modifier characters.

Examples (directive 'C' specifies ‘unsigned character’):

[65].pack('C')      # => "A"  # One element, one directive.
[65, 66].pack('CC') # => "AB" # Two elements, two directives.
[65, 66].pack('C')  # => "A"  # Extra element is ignored.
[65].pack('')       # => ""   # No directives.
[65].pack('CC')               # Extra directive raises ArgumentError.

'A'.unpack('C')   # => [65]      # One character, one directive.
'AB'.unpack('CC') # => [65, 66]  # Two characters, two directives.
'AB'.unpack('C')  # => [65]      # Extra character is ignored.
'A'.unpack('CC')  # => [65, nil] # Extra directive generates nil.
'AB'.unpack('')   # => []        # No directives.

The string template may contain any mixture of valid directives (directive 'c' specifies ‘signed character’):

[65, -1].pack('cC')  # => "A\xFF"
"A\xFF".unpack('cC') # => [65, 255]

The string template may contain whitespace (which is ignored) and comments, each of which begins with character '#' and continues up to and including the next following newline:

[0,1].pack("  C  #foo \n  C  ")    # => "\x00\x01"
"\0\1".unpack("  C  #foo \n  C  ") # => [0, 1]

Any directive may be followed by either of these modifiers:

'*' - The directive is to be applied as many times as needed:

[65, 66].pack('C*') # => "AB"
'AB'.unpack('C*')   # => [65, 66]

Integer count - The directive is to be applied count times:

[65, 66].pack('C2') # => "AB"
[65, 66].pack('C3') # Raises ArgumentError.
'AB'.unpack('C2')   # => [65, 66]
'AB'.unpack('C3')   # => [65, 66, nil]

Note: Directives in %w[A a Z m] use count differently; see String Directives.

If elements don’t fit the provided directive, only least significant bits are encoded:

[257].pack("C").unpack("C") # => [1]

Packing Method¶ ↑

Method Array#pack accepts optional keyword argument buffer that specifies the target string (instead of a new string):

[65, 66].pack('C*', buffer: 'foo') # => "fooAB"

The method can accept a block:

# Packed string is passed to the block.
[65, 66].pack('C*') {|s| p s }    # => "AB"

Unpacking Methods¶ ↑

Methods String#unpack and String#unpack1 each accept an optional keyword argument offset that specifies an offset into the string:

'ABC'.unpack('C*', offset: 1)  # => [66, 67]
'ABC'.unpack1('C*', offset: 1) # => 66

Both methods can accept a block:

# Each unpacked object is passed to the block.
ret = []
"ABCD".unpack("C*") {|c| ret << c }
ret # => [65, 66, 67, 68]

# The single unpacked object is passed to the block.
'AB'.unpack1('C*') {|ele| p ele } # => 65

Integer Directives¶ ↑

Each integer directive specifies the packing or unpacking for one element in the input or output array.

8-Bit Integer Directives¶ ↑

'c' - 8-bit signed integer (like C signed char):

[0, 1, 255].pack('c*')  # => "\x00\x01\xFF"
s = [0, 1, -1].pack('c*') # => "\x00\x01\xFF"
s.unpack('c*') # => [0, 1, -1]

'C' - 8-bit unsigned integer (like C unsigned char):

[0, 1, 255].pack('C*')    # => "\x00\x01\xFF"
s = [0, 1, -1].pack('C*') # => "\x00\x01\xFF"
s.unpack('C*')            # => [0, 1, 255]

16-Bit Integer Directives¶ ↑

's' - 16-bit signed integer, native-endian (like C int16_t):

[513, -514].pack('s*')      # => "\x01\x02\xFE\xFD"
s = [513, 65022].pack('s*') # => "\x01\x02\xFE\xFD"
s.unpack('s*')              # => [513, -514]

'S' - 16-bit unsigned integer, native-endian (like C uint16_t):

[513, -514].pack('S*')      # => "\x01\x02\xFE\xFD"
s = [513, 65022].pack('S*') # => "\x01\x02\xFE\xFD"
s.unpack('S*')              # => [513, 65022]

'n' - 16-bit network integer, big-endian:

s = [0, 1, -1, 32767, -32768, 65535].pack('n*')
# => "\x00\x00\x00\x01\xFF\xFF\x7F\xFF\x80\x00\xFF\xFF"
s.unpack('n*')
# => [0, 1, 65535, 32767, 32768, 65535]

'v' - 16-bit VAX integer, little-endian:

s = [0, 1, -1, 32767, -32768, 65535].pack('v*')
# => "\x00\x00\x01\x00\xFF\xFF\xFF\x7F\x00\x80\xFF\xFF"
s.unpack('v*')
# => [0, 1, 65535, 32767, 32768, 65535]

32-Bit Integer Directives¶ ↑

'l' - 32-bit signed integer, native-endian (like C int32_t):

s = [67305985, -50462977].pack('l*')
# => "\x01\x02\x03\x04\xFF\xFE\xFD\xFC"
s.unpack('l*')
# => [67305985, -50462977]

'L' - 32-bit unsigned integer, native-endian (like C uint32_t):

s = [67305985, 4244504319].pack('L*')
# => "\x01\x02\x03\x04\xFF\xFE\xFD\xFC"
s.unpack('L*')
# => [67305985, 4244504319]

'N' - 32-bit network integer, big-endian:

s = [0,1,-1].pack('N*')
# => "\x00\x00\x00\x00\x00\x00\x00\x01\xFF\xFF\xFF\xFF"
s.unpack('N*')
# => [0, 1, 4294967295]

'V' - 32-bit VAX integer, little-endian:

s = [0,1,-1].pack('V*')
# => "\x00\x00\x00\x00\x01\x00\x00\x00\xFF\xFF\xFF\xFF"
s.unpack('v*')
# => [0, 0, 1, 0, 65535, 65535]

64-Bit Integer Directives¶ ↑

'q' - 64-bit signed integer, native-endian (like C int64_t):

s = [578437695752307201, -506097522914230529].pack('q*')
# => "\x01\x02\x03\x04\x05\x06\a\b\xFF\xFE\xFD\xFC\xFB\xFA\xF9\xF8"
s.unpack('q*')
# => [578437695752307201, -506097522914230529]

'Q' - 64-bit unsigned integer, native-endian (like C uint64_t):

s = [578437695752307201, 17940646550795321087].pack('Q*')
# => "\x01\x02\x03\x04\x05\x06\a\b\xFF\xFE\xFD\xFC\xFB\xFA\xF9\xF8"
s.unpack('Q*')
# => [578437695752307201, 17940646550795321087]

Platform-Dependent Integer Directives¶ ↑

'i' - Platform-dependent width signed integer, native-endian (like C int):

s = [67305985, -50462977].pack('i*')
# => "\x01\x02\x03\x04\xFF\xFE\xFD\xFC"
s.unpack('i*')
# => [67305985, -50462977]

'I' - Platform-dependent width unsigned integer, native-endian (like C unsigned int):

s = [67305985, -50462977].pack('I*')
# => "\x01\x02\x03\x04\xFF\xFE\xFD\xFC"
s.unpack('I*')
# => [67305985, 4244504319]

'j' - Pointer-width signed integer, native-endian (like C intptr_t):

s = [67305985, -50462977].pack('j*')
# => "\x01\x02\x03\x04\x00\x00\x00\x00\xFF\xFE\xFD\xFC\xFF\xFF\xFF\xFF"
s.unpack('j*')
# => [67305985, -50462977]

'J' - Pointer-width unsigned integer, native-endian (like C uintptr_t):

s = [67305985, 4244504319].pack('J*')
# => "\x01\x02\x03\x04\x00\x00\x00\x00\xFF\xFE\xFD\xFC\x00\x00\x00\x00"
s.unpack('J*')
# => [67305985, 4244504319]

Other Integer Directives¶ ↑

'U' - UTF-8 character:

s = [4194304].pack('U*')
# => "\xF8\x90\x80\x80\x80"
s.unpack('U*')
# => [4194304]

'w' - BER-encoded integer (see BER encoding):

s = [1073741823].pack('w*')
# => "\x83\xFF\xFF\xFF\x7F"
s.unpack('w*')
# => [1073741823]

Modifiers for Integer Directives¶ ↑

For the following directives, '!' or '_' modifiers may be suffixed as underlying platform’s native size.

'i', 'I' - C int, always native size.
's', 'S' - C short.
'l', 'L' - C long.
'q', 'Q' - C long long, if available.
'j', 'J' - C intptr_t, always native size.

Native size modifiers are silently ignored for always native size directives.

The endian modifiers also may be suffixed in the directives above:

'>' - Big-endian.
'<' - Little-endian.

Float Directives¶ ↑

Each float directive specifies the packing or unpacking for one element in the input or output array.

Single-Precision Float Directives¶ ↑

'F' or 'f' - Native format:

s = [3.0].pack('F') # => "\x00\x00@@"
s.unpack('F')       # => [3.0]

'e' - Little-endian:

s = [3.0].pack('e') # => "\x00\x00@@"
s.unpack('e')       # => [3.0]

'g' - Big-endian:

s = [3.0].pack('g') # => "@@\x00\x00"
s.unpack('g')       # => [3.0]

Double-Precision Float Directives¶ ↑

'D' or 'd' - Native format:

s = [3.0].pack('D') # => "\x00\x00\x00\x00\x00\x00\b@"
s.unpack('D')       # => [3.0]

'E' - Little-endian:

s = [3.0].pack('E') # => "\x00\x00\x00\x00\x00\x00\b@"
s.unpack('E')       # => [3.0]

'G' - Big-endian:

s = [3.0].pack('G') # => "@\b\x00\x00\x00\x00\x00\x00"
s.unpack('G')       # => [3.0]

A float directive may be infinity or not-a-number:

inf = 1.0/0.0                  # => Infinity
[inf].pack('f')                # => "\x00\x00\x80\x7F"
"\x00\x00\x80\x7F".unpack('f') # => [Infinity]

nan = inf/inf                  # => NaN
[nan].pack('f')                # => "\x00\x00\xC0\x7F"
"\x00\x00\xC0\x7F".unpack('f') # => [NaN]

String Directives¶ ↑

Each string directive specifies the packing or unpacking for one byte in the input or output string.

Binary String Directives¶ ↑

'A' - Arbitrary binary string (space padded; count is width); nil is treated as the empty string:

['foo'].pack('A')    # => "f"
['foo'].pack('A*')   # => "foo"
['foo'].pack('A2')   # => "fo"
['foo'].pack('A4')   # => "foo "
[nil].pack('A')      # => " "
[nil].pack('A*')     # => ""
[nil].pack('A2')     # => "  "
[nil].pack('A4')     # => "    "

"foo\0".unpack('A')      # => ["f"]
"foo\0".unpack('A4')     # => ["foo"]
"foo\0bar".unpack('A10') # => ["foo\x00bar"] # Reads past "\0".
"foo ".unpack('A')       # => ["f"]
"foo ".unpack('A4')      # => ["foo"]
"foo".unpack('A4')       # => ["foo"]

russian = "\u{442 435 441 442}" # => "тест"
russian.size         # => 4
russian.bytesize     # => 8
[russian].pack('A')  # => "\xD1"
[russian].pack('A*') # => "\xD1\x82\xD0\xB5\xD1\x81\xD1\x82"
russian.unpack('A')  # => ["\xD1"]
russian.unpack('A2') # => ["\xD1\x82"]
russian.unpack('A4') # => ["\xD1\x82\xD0\xB5"]
russian.unpack('A*') # => ["\xD1\x82\xD0\xB5\xD1\x81\xD1\x82"]

'a' - Arbitrary binary string (null padded; count is width):

["foo"].pack('a')    # => "f"
["foo"].pack('a*')   # => "foo"
["foo"].pack('a2')   # => "fo"
["foo\0"].pack('a4') # => "foo\x00"
[nil].pack('a')      # => "\x00"
[nil].pack('a*')     # => ""
[nil].pack('a2')     # => "\x00\x00"
[nil].pack('a4')     # => "\x00\x00\x00\x00"

"foo\0".unpack('a')     # => ["f"]
"foo\0".unpack('a4')    # => ["foo\x00"]
"foo ".unpack('a4')     # => ["foo "]
"foo".unpack('a4')      # => ["foo"]
"foo\0bar".unpack('a4') # => ["foo\x00"] # Reads past "\0".

'Z' - Same as 'a', except that null is added or ignored with '*':

["foo"].pack('Z*')   # => "foo\x00"
[nil].pack('Z*')     # => "\x00"

"foo\0".unpack('Z*')    # => ["foo"]
"foo".unpack('Z*')      # => ["foo"]
"foo\0bar".unpack('Z*') # => ["foo"] # Does not read past "\0".

Bit String Directives¶ ↑

'B' - Bit string (high byte first):

['11111111' + '00000000'].pack('B*') # => "\xFF\x00"
['10000000' + '01000000'].pack('B*') # => "\x80@"

['1'].pack('B0') # => ""
['1'].pack('B1') # => "\x80"
['1'].pack('B2') # => "\x80\x00"
['1'].pack('B3') # => "\x80\x00"
['1'].pack('B4') # => "\x80\x00\x00"
['1'].pack('B5') # => "\x80\x00\x00"
['1'].pack('B6') # => "\x80\x00\x00\x00"

"\xff\x00".unpack("B*") # => ["1111111100000000"]
"\x01\x02".unpack("B*") # => ["0000000100000010"]

"".unpack("B0")     # => [""]
"\x80".unpack("B1") # => ["1"]
"\x80".unpack("B2") # => ["10"]
"\x80".unpack("B3") # => ["100"]

'b' - Bit string (low byte first):

['11111111' + '00000000'].pack('b*') # => "\xFF\x00"
['10000000' + '01000000'].pack('b*') # => "\x01\x02"

['1'].pack('b0') # => ""
['1'].pack('b1') # => "\x01"
['1'].pack('b2') # => "\x01\x00"
['1'].pack('b3') # => "\x01\x00"
['1'].pack('b4') # => "\x01\x00\x00"
['1'].pack('b5') # => "\x01\x00\x00"
['1'].pack('b6') # => "\x01\x00\x00\x00"

"\xff\x00".unpack("b*") # => ["1111111100000000"]
"\x01\x02".unpack("b*") # => ["1000000001000000"]

"".unpack("b0")     # => [""]
"\x01".unpack("b1") # => ["1"]
"\x01".unpack("b2") # => ["10"]
"\x01".unpack("b3") # => ["100"]

Hex String Directives¶ ↑

'H' - Hex string (high nibble first):

['10ef'].pack('H*')    # => "\x10\xEF"
['10ef'].pack('H0')    # => ""
['10ef'].pack('H3')    # => "\x10\xE0"
['10ef'].pack('H5')    # => "\x10\xEF\x00"

['fff'].pack('H3')    # => "\xFF\xF0"
['fff'].pack('H4')    # => "\xFF\xF0"
['fff'].pack('H5')    # => "\xFF\xF0\x00"
['fff'].pack('H6')    # => "\xFF\xF0\x00"
['fff'].pack('H7')    # => "\xFF\xF0\x00\x00"
['fff'].pack('H8')    # => "\xFF\xF0\x00\x00"

"\x10\xef".unpack('H*')    # => ["10ef"]
"\x10\xef".unpack('H0')    # => [""]
"\x10\xef".unpack('H1')    # => ["1"]
"\x10\xef".unpack('H2')    # => ["10"]
"\x10\xef".unpack('H3')    # => ["10e"]
"\x10\xef".unpack('H4')    # => ["10ef"]
"\x10\xef".unpack('H5')    # => ["10ef"]

'h' - Hex string (low nibble first):

['10ef'].pack('h*') # => "\x01\xFE"
['10ef'].pack('h0') # => ""
['10ef'].pack('h3') # => "\x01\x0E"
['10ef'].pack('h5') # => "\x01\xFE\x00"

['fff'].pack('h3') # => "\xFF\x0F"
['fff'].pack('h4') # => "\xFF\x0F"
['fff'].pack('h5') # => "\xFF\x0F\x00"
['fff'].pack('h6') # => "\xFF\x0F\x00"
['fff'].pack('h7') # => "\xFF\x0F\x00\x00"
['fff'].pack('h8') # => "\xFF\x0F\x00\x00"

"\x01\xfe".unpack('h*') # => ["10ef"]
"\x01\xfe".unpack('h0') # => [""]
"\x01\xfe".unpack('h1') # => ["1"]
"\x01\xfe".unpack('h2') # => ["10"]
"\x01\xfe".unpack('h3') # => ["10e"]
"\x01\xfe".unpack('h4') # => ["10ef"]
"\x01\xfe".unpack('h5') # => ["10ef"]

Pointer String Directives¶ ↑

'P' - Pointer to a structure (fixed-length string):

s = ['abc'].pack('P')  # => "\xE0O\x7F\xE5\xA1\x01\x00\x00"
s.unpack('P*')         # => ["abc"]
".".unpack("P")        # => []
("\0" * 8).unpack("P") # => [nil]
[nil].pack("P")        # => "\x00\x00\x00\x00\x00\x00\x00\x00"

'p' - Pointer to a null-terminated string:

s = ['abc'].pack('p')  # => "(\xE4u\xE5\xA1\x01\x00\x00"
s.unpack('p*')         # => ["abc"]
".".unpack("p")        # => []
("\0" * 8).unpack("p") # => [nil]
[nil].pack("p")        # => "\x00\x00\x00\x00\x00\x00\x00\x00"

Other String Directives¶ ↑

'M' - Quoted printable, MIME encoding; text mode, but input must use LF and output LF; (see RFC 2045):

["a b c\td \ne"].pack('M') # => "a b c\td =\n\ne=\n"
["\0"].pack('M')           # => "=00=\n"

["a"*1023].pack('M') == ("a"*73+"=\n")*14+"a=\n"     # => true
("a"*73+"=\na=\n").unpack('M') == ["a"*74]           # => true
(("a"*73+"=\n")*14+"a=\n").unpack('M') == ["a"*1023] # => true

"a b c\td =\n\ne=\n".unpack('M')    # => ["a b c\td \ne"]
"=00=\n".unpack('M')    # => ["\x00"]

"pre=31=32=33after".unpack('M') # => ["pre123after"]
"pre=\nafter".unpack('M')       # => ["preafter"]
"pre=\r\nafter".unpack('M')     # => ["preafter"]
"pre=".unpack('M')              # => ["pre="]
"pre=\r".unpack('M')            # => ["pre=\r"]
"pre=hoge".unpack('M')          # => ["pre=hoge"]
"pre==31after".unpack('M')      # => ["pre==31after"]
"pre===31after".unpack('M')     # => ["pre===31after"]

'm' - Base64 encoded string; count specifies input bytes between each newline, rounded down to nearest multiple of 3; if count is zero, no newlines are added; (see RFC 4648):

[""].pack('m')             # => ""
["\0"].pack('m')           # => "AA==\n"
["\0\0"].pack('m')         # => "AAA=\n"
["\0\0\0"].pack('m')       # => "AAAA\n"
["\377"].pack('m')         # => "/w==\n"
["\377\377"].pack('m')     # => "//8=\n"
["\377\377\377"].pack('m') # => "////\n"

"".unpack('m')       # => [""]
"AA==\n".unpack('m') # => ["\x00"]
"AAA=\n".unpack('m') # => ["\x00\x00"]
"AAAA\n".unpack('m') # => ["\x00\x00\x00"]
"/w==\n".unpack('m') # => ["\xFF"]
"//8=\n".unpack('m') # => ["\xFF\xFF"]
"////\n".unpack('m') # => ["\xFF\xFF\xFF"]
"A\n".unpack('m')    # => [""]
"AA\n".unpack('m')   # => ["\x00"]
"AA=\n".unpack('m')  # => ["\x00"]
"AAA\n".unpack('m')  # => ["\x00\x00"]

[""].pack('m0')             # => ""
["\0"].pack('m0')           # => "AA=="
["\0\0"].pack('m0')         # => "AAA="
["\0\0\0"].pack('m0')       # => "AAAA"
["\377"].pack('m0')         # => "/w=="
["\377\377"].pack('m0')     # => "//8="
["\377\377\377"].pack('m0') # => "////"

"".unpack('m0')     # => [""]
"AA==".unpack('m0') # => ["\x00"]
"AAA=".unpack('m0') # => ["\x00\x00"]
"AAAA".unpack('m0') # => ["\x00\x00\x00"]
"/w==".unpack('m0') # => ["\xFF"]
"//8=".unpack('m0') # => ["\xFF\xFF"]
"////".unpack('m0') # => ["\xFF\xFF\xFF"]

'u' - UU-encoded string:

[""].pack("u")        # => ""
["a"].pack("u")       # => "!80``\n"
["aaa"].pack("u")     # => "#86%A\n"

"".unpack("u")        # => [""]
"#86)C\n".unpack("u") # => ["abc"]

Offset Directives¶ ↑

'@' - Begin packing at the given byte offset; for packing, null fill or shrink if necessary:

[1, 2].pack("C@0C")     # => "\x02"
[1, 2].pack("C@1C")     # => "\x01\x02"
[1, 2].pack("C@5C")     # => "\x01\x00\x00\x00\x00\x02"
[*1..5].pack("CCCC@2C") # => "\x01\x02\x05"

For unpacking, cannot to move to outside the string:

"\x01\x00\x00\x02".unpack("C@3C") # => [1, 2]
"\x00".unpack("@1C")              # => [nil]
"\x00".unpack("@2C")              # Raises ArgumentError.

'X' - For packing, shrink for the given byte offset:

[0, 1, 2].pack("CCXC")    # => "\x00\x02"
[0, 1, 2].pack("CCX2C")   # => "\x02"

For unpacking; rewind unpacking position for the given byte offset:

"\x00\x02".unpack("CCXC") # => [0, 2, 2]

Cannot to move to outside the string:

[0, 1, 2].pack("CCX3C")   # Raises ArgumentError.
"\x00\x02".unpack("CX3C") # Raises ArgumentError.

'x' - Begin packing at after the given byte offset; for packing, null fill if necessary:

[].pack("x0")                # => ""
[].pack("x")                 # => "\x00"
[].pack("x8")                # => "\x00\x00\x00\x00\x00\x00\x00\x00"

For unpacking, cannot to move to outside the string:

"\x00\x00\x02".unpack("CxC") # => [0, 2]
"\x00\x00\x02".unpack("x3C") # => [nil]
"\x00\x00\x02".unpack("x4C") # Raises ArgumentError