--- -- An implementation of the lua 5.2 bit32 library, in pure Lua -- Note that in Lua, "x % n" is defined such that will always return a number -- between 0 and n-1 for positive n. We take advantage of that a lot here. local bit32 = {} local function checkint( name, argidx, x, level ) local n = tonumber( x ) if not n then error( string.format( "bad argument #%d to '%s' (number expected, got %s)", argidx, name, type( x ) ), level + 1 ) end return math.floor( n ) end local function checkint32( name, argidx, x, level ) local n = tonumber( x ) if not n then error( string.format( "bad argument #%d to '%s' (number expected, got %s)", argidx, name, type( x ) ), level + 1 ) end return math.floor( n ) % 0x100000000 end function bit32.bnot( x ) x = checkint32( 'bnot', 1, x, 2 ) -- In two's complement, -x = not(x) + 1 -- So not(x) = -x - 1 return ( -x - 1 ) % 0x100000000 end --- -- Logic tables for and/or/xor. We do pairs of bits here as a tradeoff between -- table space and speed. If you change the number of bits, also change the -- constants 2 and 4 in comb() below, and the initial value in bit32.band and -- bit32.btest local logic_and = { [0] = { [0] = 0, 0, 0, 0}, [1] = { [0] = 0, 1, 0, 1}, [2] = { [0] = 0, 0, 2, 2}, [3] = { [0] = 0, 1, 2, 3}, } local logic_or = { [0] = { [0] = 0, 1, 2, 3}, [1] = { [0] = 1, 1, 3, 3}, [2] = { [0] = 2, 3, 2, 3}, [3] = { [0] = 3, 3, 3, 3}, } local logic_xor = { [0] = { [0] = 0, 1, 2, 3}, [1] = { [0] = 1, 0, 3, 2}, [2] = { [0] = 2, 3, 0, 1}, [3] = { [0] = 3, 2, 1, 0}, } --- -- @param name string Function name -- @param args table Function args -- @param nargs number Arg count -- @param s number Start value, 0-3 -- @param t table Logic table -- @return number result local function comb( name, args, nargs, s, t ) for i = 1, nargs do args[i] = checkint32( name, i, args[i], 3 ) end local pow = 1 local ret = 0 for b = 0, 31, 2 do local c = s for i = 1, nargs do c = t[c][args[i] % 4] args[i] = math.floor( args[i] / 4 ) end ret = ret + c * pow pow = pow * 4 end return ret end function bit32.band( ... ) return comb( 'band', { ... }, select( '#', ... ), 3, logic_and ) end function bit32.bor( ... ) return comb( 'bor', { ... }, select( '#', ... ), 0, logic_or ) end function bit32.bxor( ... ) return comb( 'bxor', { ... }, select( '#', ... ), 0, logic_xor ) end function bit32.btest( ... ) return comb( 'btest', { ... }, select( '#', ... ), 3, logic_and ) ~= 0 end function bit32.extract( n, field, width ) n = checkint32( 'extract', 1, n, 2 ) field = checkint( 'extract', 2, field, 2 ) width = checkint( 'extract', 3, width or 1, 2 ) if field < 0 then error( "bad argument #2 to 'extract' (field cannot be negative)", 2 ) end if width <= 0 then error( "bad argument #3 to 'extract' (width must be positive)", 2 ) end if field + width > 32 then error( 'trying to access non-existent bits', 2 ) end return math.floor( n / 2^field ) % 2^width end function bit32.replace( n, v, field, width ) n = checkint32( 'replace', 1, n, 2 ) v = checkint32( 'replace', 2, v, 2 ) field = checkint( 'replace', 3, field, 2 ) width = checkint( 'replace', 4, width or 1, 2 ) if field < 0 then error( "bad argument #3 to 'replace' (field cannot be negative)", 2 ) end if width <= 0 then error( "bad argument #4 to 'replace' (width must be positive)", 2 ) end if field + width > 32 then error( 'trying to access non-existent bits', 2 ) end local f = 2^field local w = 2^width local fw = f * w return ( n % f ) + ( v % w ) * f + math.floor( n / fw ) * fw end -- For the shifting functions, anything over 32 is the same as 32 -- and limiting to 32 prevents overflow/underflow local function checkdisp( name, x ) x = checkint( name, 2, x, 3 ) return math.min( math.max( -32, x ), 32 ) end function bit32.lshift( x, disp ) x = checkint32( 'lshift', 1, x, 2 ) disp = checkdisp( 'lshift', disp ) return math.floor( x * 2^disp ) % 0x100000000 end function bit32.rshift( x, disp ) x = checkint32( 'rshift', 1, x, 2 ) disp = checkdisp( 'rshift', disp ) return math.floor( x / 2^disp ) % 0x100000000 end function bit32.arshift( x, disp ) x = checkint32( 'arshift', 1, x, 2 ) disp = checkdisp( 'arshift', disp ) if disp <= 0 then -- Non-positive displacement == left shift -- (since exponent is non-negative, the multipication can never result -- in a fractional part) return ( x * 2^-disp ) % 0x100000000 elseif x < 0x80000000 then -- High bit is 0 == right shift -- (since exponent is positive, the division will never increase x) return math.floor( x / 2^disp ) elseif disp > 31 then -- Shifting off all bits return 0xffffffff else -- 0x100000000 - 2 ^ ( 32 - disp ) creates a number with the high disp -- bits set. So shift right then add that number. return math.floor( x / 2^disp ) + ( 0x100000000 - 2 ^ ( 32 - disp ) ) end end -- For the rotation functions, disp works mod 32. -- Note that lrotate( x, disp ) == rrotate( x, -disp ). function bit32.lrotate( x, disp ) x = checkint32( 'lrotate', 1, x, 2 ) disp = checkint( 'lrotate', 2, disp, 2 ) % 32 local x = x * 2^disp return ( x % 0x100000000 ) + math.floor( x / 0x100000000 ) end function bit32.rrotate( x, disp ) x = checkint32( 'rrotate', 1, x, 2 ) disp = -checkint( 'rrotate', 2, disp, 2 ) % 32 local x = x * 2^disp return ( x % 0x100000000 ) + math.floor( x / 0x100000000 ) end return bit32