faf-replay-parser

FAF Replay Parser

A fast library for parsing Supreme Commander Forged Alliance replay files.

Installation

Pre-built packages are available for Linux, MacOS, and Windows. You can install them with:

pip install faf-replay-parser

Documentation

Here are some examples of using the parser. Check help(fafreplay) for more details on available functions.

Gathering basic info

For the most basic uses there are a few special functions that should be preferred over constructing a Parser object. These do only one job, but they do it very quickly.

Current functions:

• body_offset - Used for splitting replay data into header and body.
• body_ticks - Used for extracting the game length.

Example

from datetime import timedelta
from fafreplay import body_offset, body_ticks

# Split replay data into header and body
offset = body_offset(data)
header_data, body_data = data[:offset], data[offset:]

# Get replay length in ticks
ticks = body_ticks(body_data)
print("Game length:", timedelta(milliseconds=ticks*100))

Using the Parser object

The Parser object can be used to get fine grained control over how replay commands are parsed. Generally it's a good idea to parse only the minimum commands needed, as conversion back to python dictionaries is quite expensive.

from datetime import timedelta
from fafreplay import Parser, commands

parser = Parser(
    # Skip all commands except the ones defined here
    commands=[
        commands.Advance,                   # For the tick counter
        commands.VerifyChecksum,            # For desync detection
    ],
    # Throw away commands right after we parse them. Setting this to `True` will
    # significantly increase the parse time.
    save_commands=False,
    limit=None,
    stop_on_desync=False
)
# Or create a parser with default arguments (turn off save_commands though)
# parser = Parser(save_commands=False)

# Read replay to a `bytes` object
with open("12345.scfareplay", "rb") as f:
    data = f.read()

# Parse to a python dictionary. Data must of type `bytes` or `bytearray`
replay = parser.parse(data)
print("Game time:", timedelta(milliseconds=replay["body"]["sim"]["tick"]*100))
if replay["body"]["sim"]["desync_ticks"]:
    print("Replay desynced!")

Benchmark comparison

To see how much faster the basic functions can be, consider this simple example done on replay 8653680 (an almost 50 minute long Seton's game).

>>> len(body_data)
5586339
>>> body_ticks(body_data)
28917
>>> parser = Parser(
...     commands=[commands.Advance],
...     save_commands=False
... )
>>> timeit.timeit("parser.parse_body(body_data)['sim']['tick']", globals=globals(), number=100)
1.4510237049980788
>>> timeit.timeit("body_ticks(body_data)", globals=globals(), number=100)
0.20173147800232982

In this case body_ticks turned out to be more than 7x faster than using a Parser.

Reading .fafreplay files

Replays downloaded from faforever.com use a compressed data format to reduce the size of the files. These can be decompressed to the original .scfareplay data using the extract_scfa function.

from fafreplay import extract_scfa


with open("12345.scfareplay", "rb") as f:
    scfa_data = f.read()

with open("12345.fafreplay", "rb") as f:
    faf_data = extract_scfa(f)

# The extracted data is in the .scfareplay format
assert faf_data == scfa_data

Note that there are several versions of the .fafreplay format. Version 1 uses base64 and zlib compression which are both part of the python standard library. However, version 2 uses zstd which must be installed through a third party package. To ensure that this dependency is installed you can use the faf extra when installing the parser:

pip install "faf-replay-parser[faf]"