simple fixed-width integers
This module provides fixed-size integer classes which retain their fixed nature across arithmetic operations. It is geared towards users who need to emulate machine integers.
It provides flexible classes for defining integers with a fixed number of bits, as well as predefined classes for common machine integer sizes. These classes can be used as drop-in replacements for int/long, and can be sliced to extract bitfields.
Mutable versions of these integers are provided, enabling usages such as emulation of machine registers.
A collection of predefined fixed-width integers for widths 8, 16, 32 and 64 are available in signed and unsigned varieties. Mutable and immutable versions of each type are provided.
These are named as [Mutable][U]Int<N>, e.g. UInt64 or MutableInt8. Use these classes as you would int; arithmetic operations involving these classes will preserve fixed width. For example:
x = UInt32(0) print(hex(~x)) # prints 0xffffffff
Mutable instances can be modified in-place, preserving their type:
x = MutableUInt32(0) y = x x += 100 print(y) # prints 100
To set a mutable integer without losing its type, use slicing:
x = MutableUInt32(0) x[:] = -1 print(hex(x)) # prints 0xffffffff
FixedInt instances support all arithmetic operators. For binary operators, both operands are converted to plain Python int and then operated on. With a few exceptions, the result will be cast back to a FixedInt large enough to hold either operand, provided one of the operands was a FixedInt. Note that the resulting FixedInt may not be large enough to hold the complete result, in which case the result will be truncated.
The exceptions are as follows:
- divmod returns a tuple of plain int s
- true division returns a float
- **, << and >> will return a FixedInt if the left operand was a FixedInt, and plain int otherwise.
Mutable instances additionally support in-place operations, which will modify the value without altering its type.
Arithmetic operations between two integers of different sizes follow C integer promotion rules when determining the type of the final result. These rules boil down to the following:
- If the operands are both signed, or both unsigned, the wider of the two operand types is chosen.
- Otherwise, if the unsigned operand is wider, the unsigned operand is chosen.
- Otherwise, the signed operand is chosen.
FixedInt instances support slicing. Slicing with a single integer produces a single Boolean value representing the bit at that position. Slicing with a range produces a FixedInt containing the range of bits. Mutable instances additionally support slice assignment. This makes e.g. manipulating a flag register straightforward, without needing to use bitwise operations.
All indexing operations treat the least-significant bit (LSB) as bit 0. Currently, only contiguous bit sections can be obtained; for more flexibility consider using a module such as bitarray.
Getting a slice results in a FixedInt instance with exactly as many bits as the range. This can be used to perform wraparound arithmetic on a bit field.
Slices support two main syntaxes:
The latter syntax is more convenient when dealing with fixed-width fields. Both of the slice arguments may be omitted, in which case they will default to the LSB and MSB of the FixedInt respectively.
FixedInt instances can be converted to and from raw byte representations by using the .to_bytes instance method and the .from_bytes classmethod. The usage of these methods matches that of Python 3.4’s int.to_bytes and int.from_bytes methods.