I’m working with a lot of binary data, where my data can be a mix of bit-widths.
I want to print or log my data correctly formatted by bit-width, all in a single statement.

Simple example showing only 16 and 32 bit data:

import numpy as np

numbers_32_bit = np.array((0x1000_0000, 0x2000_0000, 0x3000_0000, 0x4000_0000),dtype=np.uint32)
numbers_16_bit = np.array((0x5000, 0x6000, 0x7000, 0x8000), dtype=np.uint16)

np_formatter = {
    # Doesn't throw an error, but has no effect, as expected as per numpy docs
    "uint32": lambda x: f"{x:08X}",
    "uint16": lambda x: f"{x:04X}",
    
    # Applies to all ints
    "int": lambda x: f"0x{x:08X}"
}

np.set_printoptions(formatter=np_formatter)
print(f"32 bits: {numbers_32_bit}, 16 bits: {numbers_16_bit}")

Print output:

32 bits: [0x10000000 0x20000000 0x30000000 0x40000000], 16 bits: [0x00005000 0x00006000 0x00007000 0x00008000]

As you can see, the 16 bit numbers still have the 32-bit formatter applied, as it uses the int formatter, rather than uint16.

This works on a statement-by-statement basis, but isn’t great when you have lots of print/log statements throughout a project:

def format_bitwidths(n8=0, n16=0, n32=0, n64=0):
    np.set_printoptions(formatter={"int": lambda x: f"0x{x:02X}"})
    s8 = f"{n8}"
    np.set_printoptions(formatter={"int": lambda x: f"0x{x:04X}"})
    s16 = f"{n16}"
    np.set_printoptions(formatter={"int": lambda x: f"0x{x:08X}"})
    s32 = f"{n32}"
    np.set_printoptions(formatter={"int": lambda x: f"0x{x:016X}"})
    s64 = f"{n64}"
    return s8, s16, s32, s64


_, str_16_bits, str_32_bits, _ = format_bitwidths(n16=numbers_16_bit, n32=numbers_32_bit)
print(f"32 bits: {str_32_bits}, 16 bits: {str_16_bits}")

Print output:

32 bits: [0x10000000 0x20000000 0x30000000 0x40000000], 16 bits: [0x5000 0x6000 0x7000 0x8000]

Is there a way to set this once without having to call np.set_printoptions or format_bitwidths before each print/log?

If not, is there some other method which I can set once, and apply it across an entire script?