Bridge

[ZiggerZZ]

Hello!

This is the implementation of #1117. It is not ready to be merged yet - I need to change all binvecs in bridge(use_double_dummy_result=false).txt to reflect the change of the order of the tricks (from [previous, current] to [current, previous]). As suggested by @elkhrt, I didn't introduce absolute_order parameter.
If this approach is fine with you, I'll update the tests and the PR will be ready to be merged.

[ZiggerZZ]

Fixed the tests. Since I swap the current and the previous tricks, I needed to swap those in the binary representations.
Here's the script that takes bridge(use_double_dummy_result=false).txt and modifies hex values for state observations only for the playing phase (hence condition idx > 700). I used this script to patch the tests.

def change_order(hex_value):
    int_value = int(hex_value, 16)  # Convert the hex string to an integer
    binary_str = bin(int_value)[2:]  # Convert the integer to binary and remove the '0b' prefix
    length_required = 571
    binary_str_padded = binary_str.zfill(length_required)
    index_previous_start, index_previous_end = 129, 337
    index_current_start, index_current_end = 337, 545
    new_string_binary = binary_str_padded[:index_previous_start] + binary_str_padded[index_current_start:index_current_end] + binary_str_padded[index_previous_start: index_previous_end] + binary_str_padded[index_current_end:]
    int_value_from_binary = int(new_string_binary, 2)
    hex_str = hex(int_value_from_binary)
    return hex_str

file_path = "./bridge(use_double_dummy_result=false).txt"

with open(file_path, 'r') as file:
    lines = file.readlines()

for idx, line in enumerate(lines):
	# only apply to lines after bidding
    if "binvec" in line and idx > 700:
        hex_value = line.split(', ')[-1].split(')')[0]  # Extract the hex value
        new_hex_value = change_order(hex_value)
        lines[idx] = line.replace(hex_value, new_hex_value)  # Replace old hex value with the new one

with open(file_path, 'w') as file:
    file.writelines(lines)

[ZiggerZZ]

Taken @MattX's review into account

[lanctot]

Thanks for this and to @MattX for the review.

We should be able to pull this in soon. I'll just quickly run it by Ed early next week.

[lanctot]

BTW @ZiggerZZ, you can always simply regenerate those playthroughs given the new implementation using this script: https://github.com/google-deepmind/open_spiel/blob/master/open_spiel/scripts/regenerate_playthroughs.sh

[ZiggerZZ]

Thanks @lanctot! I wanted to verify the correctness of the implementation by writing that logic of modifying the existing observation tensors.