parse_csgo_cfg.py

# NOTE: For stable output, this should be run on a Python version that
# guarantees dict order (3.7+).
import re
import os.path
import sys
from enum import Flag, auto
from collections import defaultdict
from pprint import pprint

if len(sys.argv) < 3:
	print("Need input and output file names")
	sys.exit(1)
fn = os.path.expanduser(sys.argv[1])
out = os.path.expanduser(sys.argv[2])

print("Building %s from %s" % (out, fn))

# In a Source config file (does the format have a name?), data consists
# of alternating keys and values. Keys are always strings; values are
# either strings or mappings. A string starts with a double quote, ends
# with a double quote, and... what happens if it contains one? TODO.
# A mapping starts with an open brace, contains one or more (TODO: can
# it contain zero?) key+value pairs, and ends with an open brace.
# Between any two values, any amount of whitespace is found, possibly
# including comments, which start with "//" and end at EOL.
# 20210521: People call it "VDF", I guess that works. And yes, a mapping
# can contain zero kv pairs, which should be fine for this code. Still
# don't know about strings containing quotes. Other notes on parsing this
# format can be seen in my shed - parsevdf.pike and its grammar.

# Skip whitespace and comments
RE_SKIP = re.compile(r'(\s*//[^\n]*\n)*\s*')
# Read a single string
RE_STRING = re.compile(r'"([^"\n]*)"')

def merge_mappings(m1, m2, path):
	"""Recursively merge the contents of m2 into m1"""
	if type(m1) is not type(m2): raise ValueError("Cannot merge %s and %s" % (type(m1), type(m2)))
	if type(m1) is str:
		if m1 == m2: return m1 # Attempting to merge "foo" into "foo" just produces "foo"
		# Actually.... the Valve-provided files have dirty data in them.
		# We can't make assertions like this. Sigh.
		# raise ValueError("Cannot merge different strings %r and %r --> %s" % (m1, m2, path))
		return m1 # Keep the one from m1... I think??
	for k,v in m2.items():
		if k in m1: merge_mappings(m1[k], v, path + "." + k)
		else: m1[k] = v

def parse_cfg(data):
	pos = 0
	def skip_ws():
		nonlocal pos
		pos = RE_SKIP.match(data, pos).end()
	def parse_str():
		nonlocal pos
		m = RE_STRING.match(data, pos)
		if not m: raise ValueError("Unable to parse string at pos %d" % pos)
		pos = m.end()
		return m.group(1)
	def parse_mapping(path):
		nonlocal pos
		pos += 1 # Skip the initial open brace
		ret = {}
		while "moar stuffo":
			skip_ws()
			if data[pos] == '}': break
			key = parse_str()
			value = parse_value(path + "." + key)
			if key in ret:
				# Sometimes there are duplicates. I don't know what the deal is.
				merge_mappings(value, ret[key], path + "." + key)
			ret[key] = value
		pos += 1 # Skip the final close brace
		return ret
	def parse_value(path):
		skip_ws()
		if data[pos] == '"': return parse_str()
		if data[pos] == '{': return parse_mapping(path)
		raise ValueError("Unexpected glyph '%s' at pos %d" % (data[pos], pos))
	skip_ws()
	assert data[pos] == '"' # The file should always start with a string
	title = parse_str()
	return parse_value(title)

with open(fn) as f: data = f.read()
info = parse_cfg(data)

class Cat(Flag):
	Pistol = auto()
	Shotgun = auto()
	SMG = auto()
	AR = auto()
	Sniper = auto()
	LMG = auto()
	Grenade = auto() # Not currently being listed
	Equipment = auto() # Not currently being listed
	# ----- The first eight categories define the weapon type. Others are flags. Note that
	# the number of categories above this line is hard-coded as WEAPON_TYPE_CATEGORIES below.
	Automatic: "fully Automatic gun" = auto()
	Scoped: "Scoped weapon" = auto()
	Starter: "Starter pistol" = auto()
	NonDamaging = auto() # Not currently detected (will only be on grenade/equip)
	# Create some aliases used by the weapon_type lookup
	SubMachinegun = SMG
	Rifle = AR
	SniperRifle = Sniper
	Machinegun = LMG

demo_items = dict(
	# - "How many total Shotguns do I have here?" -- just count 'em (7)
	nova=1,
	mag7=3,
	sawedoff=3,
	xm1014=0,
	# - "Find my largest magazine fully Automatic gun. How many shots till I reload?" -- it's a Galil (35)
	galilar=1,
	bizon=0,
	p90=0,
	m249=0,
	negev=0,
	# - "How many distinct Pistols do I have here?" -- count unique items (5)
	# The actual selection here is arbitrary and could be randomized.
	deagle=-3,
	elite=-3,
	fiveseven=-3,
	glock=0,
	hkp2000=-3,
	p250=0,
	cz75a=0,
	tec9=-3,
	usp_silencer=0,
	revolver=0,
	# - "This is my SMG. There are none quite like it. How well does it penetrate armor?" -- it's an MP9 (60)
	mp9=1,
	mac10=-2,
	mp7=-2,
	mp5sd=-2,
	ump45=-2,
	# - "This is my Shotgun. There are none quite like it. How many shots till I reload?" -- it's a Nova (8)
	# (covered above)
)

arrays = defaultdict(list)
arrays["categories"] = [c.name for c in Cat]
# FIXME: Does not work on modern Pythons, find a better way to do this (pref not w/ annotations at all)
arrays["category_descr"] = [c.__annotations__.get(c.name, c.name) for c in Cat]
for weapon, data in info["prefabs"].items():
	if data.get("prefab") == "grenade":
		print("Got a nade:", weapon)
	if "item_class" not in data or "attributes" not in data: continue
	# This is a sneaky way to restrict it to just "normal weapons", since
	# you can't apply a sticker to your fists or your tablet :)
	if "stickers" not in data: continue
	weap = weapon.replace("_prefab", "")
	arrays["item_name"].append(weap) # NOTE: This isn't always the same as the weapon_class (cf CZ75a).
	weap = weap.replace("weapon_", "")
	for attr, dflt in {
		"primary clip size": "-1",
		"primary reserve ammo max": "-1",
		"in game price": "-1",
		"kill award": "300",
		"range modifier": "0.98",
	}.items():
		arrays[attr.replace(" ", "_")].append(float(data["attributes"].get(attr, dflt)))
	arrays["armor_pen"].append(float(data["attributes"]["armor ratio"]) * 50)
	# The data file has two speeds available. For scoped weapons, the main
	# speed is unscoped and the alternate is scoped (SG556 has 210 / 150), but
	# for the stupid Revolver, the alternate speed is your real speed, and
	# the "base" speed is how fast you move while charging your shot. Since
	# logically the shot-charging is the alternate, we just pick the higher
	# speed in all cases, so we'll call the SG556 "210" and the R8 "220".
	spd = data["attributes"].get("max player speed", "260")
	spd2 = data["attributes"].get("max player speed alt", "260")
	arrays["max_player_speed"].append(max(float(spd), float(spd2)))
	arrays["demo_quantity"].append(demo_items.get(weap, -1))
	cat = Cat[data["visuals"]["weapon_type"]]
	if int(data["attributes"].get("bullets", "1")) > 1: cat |= Cat.Shotgun # Probably don't actually need this
	if int(data["attributes"].get("is full auto", "0")): cat |= Cat.Automatic
	if int(data["attributes"].get("zoom levels", "0")): cat |= Cat.Scoped
	if data["item_class"] in {"weapon_hkp2000", "weapon_glock"}: cat |= Cat.Starter
	# TODO: Suppressed weapons
	arrays["category"].append(cat.value)
	# Get a quick dump of which weapons are in which categories
	# for c in Cat:
		# if cat & c: arrays["cat_" + c.name].append(weap)
# pprint(list(info["prefabs"]))

with open(out, "w") as f:
	print("//Autogenerated file, do not edit", file=f)
	print("#define WEAPON_TYPE_CATEGORIES 8", file=f)
	for name, arr in arrays.items():
		if name in {"item_name", "categories", "category_descr"} or name.startswith("cat_"): # String fields
			print(f"char weapondata_{name}[][] = {{", file=f)
			for val in arr:
				print(f'\t"{val}",', file=f) # Don't have quotes in them. K?
		else: # Numeric fields
			t = {"category": "int", "demo_quantity": "int"}.get(name, "float")
			print(f"{t} weapondata_{name}[] = {{", file=f)
			for val in arr:
				print(f"\t{val},", file=f)
		print("};", file=f)
	print(f"//Autogenerated from {fn}", file=f)