JavaScript Numbers are represented as IEEE 754 double-precision floats. Unfortunately, this means they lose integer precision for values beyond +/- 2^^53. For projects that need to accurately handle 64-bit ints, such as node-thrift, a performant, Number-like class is needed. Int64 is that class.
Int64 instances look and feel much like JS-native Numbers. By way of example ...
> (0x123456789).toString(16)
'123456789'
> (0x123456789abcdef0).toString(16)
'123456789abcdf00'
> Int64 = require('node-int64')
> x = new Int64(0x123456789)
[Int64 value:4886718345 octets:00 00 00 01 23 45 67 89]
> y = new Int64('123456789abcdef0')
[Int64 value:Infinity octets:12 34 56 78 9a bc de f0]
> x + 1
4886718346
> y + 1
Infinity
> 'value: ' + x
'value: 4886718345'
> 'value: ' + y
'value: Infinity'
> x.toString(2)
'100100011010001010110011110001001'
> y.toString(2)
'Infinity'
> isFinite(x)
true
> isFinite(y)
false
> x.toOctetString()
'0000000123456789'
> y.toOctetString()
'123456789abcdef0'
> new Int64(0x12345678, 0x9abcdef0)
[Int64 value:Infinity octets:12 34 56 78 9a bc de f0]
> new Int64(new Buffer([0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0]))
[Int64 value:Infinity octets:12 34 56 78 9a bc de f0]
> new Int64(new Buffer([0,0,0,0,0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0]), 4)
[Int64 value:Infinity octets:12 34 56 78 9a bc de f0]
> new Int64(new Buffer([0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0])).toBuffer()
<Buffer 12 34 56 78 9a bc de f0>
> var buf = new Buffer(1024);
> new Int64(new Buffer([0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0])).copy(buf, 512);