#!/usr/bin/env python
# -*- encoding: utf-8 -*-
## tmx_converter.py - Extract walkmap, warp, and spawn information from maps.
##
## Copyright © 2012 Ben Longbons <b.r.longbons@gmail.com>
##
## This file is part of The Mana World
##
## This program is free software: you can redistribute it and/or modify
## it under the terms of the GNU General Public License as published by
## the Free Software Foundation, either version 2 of the License, or
## (at your option) any later version.
##
## This program is distributed in the hope that it will be useful,
## but WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with this program. If not, see <http://www.gnu.org/licenses/>.
from __future__ import print_function
import sys
import os
import posixpath
import struct
import xml.sax
import base64
import zlib
dump_all = False # wall of text
check_mobs = True # mob_db.txt
# lower case versions of everything except 'spawn' and 'warp'
other_object_types = set([
'particle_effect',
'npc', # not interpreted by client
'script', # for ManaServ
'fixme', # flag for things that didn't have a type before
'music',
])
# Somebody has put ManaServ fields in our data!
other_spawn_fields = (
'spawn_rate',
)
other_warp_fields = (
)
TILESIZE = 32
SEPARATOR = '|'
MESSAGE = 'This file is generated automatically. All manually changes will be removed when running the Converter.'
CLIENT_MAPS = 'maps'
SERVER_WLK = 'data'
SERVER_NPCS = 'npc'
TMWA_MAP_CONF = 'conf/tmwa-map.conf'
NPC_MOBS = '_mobs.txt'
NPC_WARPS = '_warps.txt'
NPC_IMPORTS = '_import.txt'
NPC_MASTER_IMPORTS = NPC_IMPORTS
class State(object):
pass
State.INITIAL = State()
State.LAYER = State()
State.DATA = State()
State.FINAL = State()
class Object(object):
__slots__ = (
'name',
#'map',
'x', 'y',
'w', 'h',
)
class Mob(Object):
__slots__ = (
'monster_id',
'max_beings',
'ea_spawn',
'ea_death',
) + other_spawn_fields
def __init__(self):
self.max_beings = 1
self.ea_spawn = 0
self.ea_death = 0
class Warp(Object):
__slots__ = (
'dest_map',
'dest_tile_x',
'dest_tile_y',
) + other_warp_fields
class ContentHandler(xml.sax.ContentHandler):
__slots__ = (
'locator', # keeps track of location in document
'out', # open file handle to .wlk
'state', # state of collision info
'tilesets', # first gid of each tileset
'buffer', # characters within a section
'encoding', # encoding of layer data
'compression', # compression of layer data
'width', # width of the collision layer
'height', # height of the collision layer
'base', # base name of current map
'npc_dir', # world/map/npc/<base>
'mobs', # open file to _mobs.txt
'warps', # open file to _warps.txt
'imports', # open file to _import.txt
'name', # name property of the current map
'object', # stores properties of the latest <object> tag
'mob_ids', # set of all mob types that spawn here
)
def __init__(self, out, npc_dir, mobs, warps, imports):
xml.sax.ContentHandler.__init__(self)
self.locator = None
self.out = open(out, 'w')
self.state = State.INITIAL
self.tilesets = set([0]) # consider the null tile as its own tileset
self.buffer = bytearray()
self.encoding = None
self.compression = None
self.width = None
self.height = None
self.base = posixpath.basename(npc_dir)
self.npc_dir = npc_dir
self.mobs = mobs
self.warps = warps
self.imports = imports
self.object = None
self.mob_ids = set()
def setDocumentLocator(self, loc):
self.locator = loc
# this method randomly cuts in the middle of a line; thus funky logic
def characters(self, s):
if not s.strip():
return
if self.state is State.DATA:
self.buffer += s.encode('ascii')
def startDocument(self):
pass
def startElement(self, name, attr):
if dump_all:
attrs = ' '.join('%s="%s"' % (k,v) for k,v in attr.items())
if attrs:
print('<%s %s>' % (name, attrs))
else:
print('<%s>' % name)
if self.state is State.INITIAL:
if name == u'property' and attr[u'name'].lower() == u'name':
self.name = attr[u'value']
self.mobs.write('// %s\n' % MESSAGE)
self.mobs.write('// %s mobs\n\n' % self.name)
self.warps.write('// %s\n' % MESSAGE)
self.warps.write('// %s warps\n\n' % self.name)
if name == u'tileset':
self.tilesets.add(int(attr[u'firstgid']))
if name == u'layer' and attr[u'name'].lower().startswith(u'collision'):
self.width = int(attr[u'width'])
self.height = int(attr[u'height'])
self.out.write(struct.pack('<HH', self.width, self.height))
self.state = State.LAYER
elif self.state is State.LAYER:
if name == u'data':
if attr.get(u'encoding','') not in (u'', u'csv', u'base64', u'xml'):
print('Bad encoding:', attr.get(u'encoding',''))
return
self.encoding = attr.get(u'encoding','')
if attr.get(u'compression','') not in (u'', u'none', u'zlib', u'gzip'):
print('Bad compression:', attr.get(u'compression',''))
return
self.compression = attr.get(u'compression','')
self.state = State.DATA
elif self.state is State.DATA:
self.out.write(chr(int(attr.get(u'gid',0)) not in self.tilesets))
elif self.state is State.FINAL:
if name == u'object':
obj_type = attr[u'type'].lower()
x = int(attr[u'x']) / TILESIZE;
y = int(attr[u'y']) / TILESIZE;
w = int(attr.get(u'width', 0)) / TILESIZE;
h = int(attr.get(u'height', 0)) / TILESIZE;
# I'm not sure exactly what the w/h shrinking is for,
# I just copied it out of the old converter.
# I know that the x += w/2 is to get centers, though.
if obj_type == 'spawn':
self.object = Mob()
if w > 1:
w -= 1
if h > 1:
h -= 1
x += w/2
y += h/2
elif obj_type == 'warp':
self.object = Warp()
x += w/2
y += h/2
w -= 2
h -= 2
else:
if obj_type not in other_object_types:
print('Unknown object type:', obj_type, file=sys.stderr)
self.object = None
return
obj = self.object
obj.x = x
obj.y = y
obj.w = w
obj.h = h
obj.name = attr[u'name']
elif name == u'property':
obj = self.object
if obj is None:
return
key = attr[u'name'].lower()
value = attr[u'value']
# Not true due to defaulting
#assert not hasattr(obj, key)
try:
value = int(value)
except ValueError:
pass
setattr(obj, key, value)
def add_warp_line(self, line):
self.warps.write(line)
def endElement(self, name):
if dump_all:
print('</%s>' % name)
if name == u'object':
obj = self.object
if isinstance(obj, Mob):
mob_id = obj.monster_id
if mob_id < 1000:
mob_id += 1002
if check_mobs:
try:
name = mob_names[mob_id]
except KeyError:
print('Warning: unknown mob ID: %d (%s)' % (mob_id, obj.name))
else:
if name != obj.name:
print('Warning: wrong mob name: %s (!= %s)' % (obj.name, name))
obj.name = name
self.mob_ids.add(mob_id)
self.mobs.write(
SEPARATOR.join([
'%s.gat,%d,%d,%d,%d' % (self.base, obj.x, obj.y, obj.w, obj.h),
'monster',
obj.name,
'%d,%d,%dms,%dms,Mob%s::On%d\n' % (mob_id, obj.max_beings, obj.ea_spawn, obj.ea_death, self.base, mob_id),
])
)
elif isinstance(obj, Warp):
obj_name = obj.name[:23]
if obj.name != obj_name:
print('Warning: warp name truncated: %r -> %r' % (obj.name, obj_name))
self.warps.write(
SEPARATOR.join([
'%s.gat,%d,%d' % (self.base, obj.x, obj.y),
'warp',
obj_name,
'%d,%d,%s.gat,%d,%d\n' % (obj.w, obj.h, obj.dest_map, obj.dest_tile_x, obj.dest_tile_y),
])
)
if name == u'data':
if self.state is State.DATA:
if self.encoding == u'csv':
for x in self.buffer.split(','):
self.out.write(chr(int(x) not in self.tilesets))
elif self.encoding == u'base64':
data = base64.b64decode(str(self.buffer))
if self.compression == u'zlib':
data = zlib.decompress(data)
elif self.compression == u'gzip':
data = zlib.decompressobj().decompress('x\x9c' + data[10:-8])
for i in range(self.width*self.height):
self.out.write(chr(int(struct.unpack('<I',data[i*4:i*4+4])[0]) not in self.tilesets))
self.state = State.FINAL
def endDocument(self):
self.mobs.write('\n\n%s.gat,0,0,0|script|Mob%s|-1\n{\n end;\n' % (self.base, self.base))
for mob_id in sorted(self.mob_ids):
self.mobs.write('\nOn%d:\n set @mobID, %d;\n callfunc "MobPoints";\n end;\n' % (mob_id, mob_id))
self.mobs.write('}\n')
self.imports.write('// Map %s: %s\n' % (self.base, self.name))
self.imports.write('// %s\n' % MESSAGE)
self.imports.write('map: %s.gat\n' % self.base)
npcs = os.listdir(self.npc_dir)
npcs.sort()
for x in npcs:
if x == NPC_IMPORTS:
continue
if x.startswith('.'):
continue
if x.endswith('.txt'):
self.imports.write('npc: %s\n' % posixpath.join(SERVER_NPCS, self.base, x))
pass
def main(argv):
_, client_data, server_data = argv
tmx_dir = posixpath.join(client_data, CLIENT_MAPS)
wlk_dir = posixpath.join(server_data, SERVER_WLK)
npc_dir = posixpath.join(server_data, SERVER_NPCS)
if check_mobs:
global mob_names
mob_names = {}
with open(posixpath.join(server_data, TMWA_MAP_CONF)) as mob_dbs:
for mob_db_line in mob_dbs:
if mob_db_line.startswith('mob_db:'):
with open(posixpath.join(server_data, mob_db_line.split(':')[1].strip())) as mob_db:
for line in mob_db:
if not line.strip():
continue
if line.startswith('//'):
continue
k, v, _ = line.split(',', 2)
mob_names[int(k)] = v.strip()
npc_master = []
map_basenames = []
for arg in os.listdir(tmx_dir):
base, ext = posixpath.splitext(arg)
if ext == '.tmx':
map_basenames.append(base)
tmx = posixpath.join(tmx_dir, arg)
wlk = posixpath.join(wlk_dir, base + '.wlk')
this_map_npc_dir = posixpath.join(npc_dir, base)
os.path.isdir(this_map_npc_dir) or os.mkdir(this_map_npc_dir)
print('Converting %s to %s' % (tmx, wlk))
with open(posixpath.join(this_map_npc_dir, NPC_MOBS), 'w') as mobs:
with open(posixpath.join(this_map_npc_dir, NPC_WARPS), 'w') as warps:
with open(posixpath.join(this_map_npc_dir, NPC_IMPORTS), 'w') as imports:
xml.sax.parse(tmx, ContentHandler(wlk, this_map_npc_dir, mobs, warps, imports))
npc_master.append('import: %s\n' % posixpath.join(SERVER_NPCS, base, NPC_IMPORTS))
with open(posixpath.join(wlk_dir, 'resnametable.txt'), 'w') as resname:
for base in sorted(map_basenames):
resname.write('%s.gat#%s.wlk#\n' % (base, base))
with open(posixpath.join(npc_dir, NPC_MASTER_IMPORTS), 'w') as out:
out.write('// %s\n\n' % MESSAGE)
npc_master.sort()
for line in npc_master:
out.write(line)
if __name__ == '__main__':
main(sys.argv)