Python .REO to .OBJ Converter

Anything to do with Drakan level editing and modifications, post it here! Also this is the place to tell us about your new levels and get player feedback.

Moderators: Arokhs Twin, yangez93

Posts: 17
Joined: Thu Feb 27, 2014 2:46 am

Python .REO to .OBJ Converter

Postby unterbuchse » Mon Aug 08, 2016 2:15 am


I just tested the REO to OBJ converter posted here and it did not work out for me (The generated obj. files were wrong regarding vertex count and face count. Arokhs soul crystal for example had some weird additional vertices and faces) . Furthermore, the whole hassle with binaries, runtimes etc and the "only for windows thingy" somehow annoyed me. So i just wrote a python script for it. Its just quick and dirty so dont expect the holy grail here.
I tested it for 3 REO files and it works (including texture mapping). The bounding spheres / boxes and other additional information is parsed but not yet used in the obj. output.

I will clean it up / enhance it tomorrow, if i can find some motivation.
some screens:
(200.76 KiB) Downloaded 1888 times
(230.89 KiB) Downloaded 1888 times
(205.85 KiB) Downloaded 1888 times
* I only put one of the 3 textures on the dragon head so yeah the side textures are incorrect but their coordinates are correct *

Will put the code on github, but if someone wants to experiment right away:

** UPDATED VERSION FROM 09.August.2016 - now with .mtl conversion **

Code: Select all

''' Created on 08.08.2016 @author: BuXXe ''' import REOClass if __name__ == '__main__': # Filename without extension! (.reo .obj and .mtl will be appended in code) # TODO: Right now we assume the reo file is in the same directory as the filename="InnerTemple03" # read the reo file # blocks: version, name, created by ... on, lighting, transform, materials, faces, vertices, bspheres/bboxes with open(filename+".reo", 'r') as f: read_data = f.readlines() f.close() # TODO: perhaps use only regex in future but for now keep it simple # TODO: perhaps we should use some kind of grammar or state machine to parse the file # iterate through the read_data linecounter = 0 # create new object REO = REOClass.REO() # INFO: Comments and empty lines will be ignored as for now while(linecounter < len(read_data)): if(read_data[linecounter][0] == '#' or read_data[linecounter][0] == '\n'): linecounter += 1 continue # set metadata elif(read_data[linecounter].startswith("version")): REO.version = read_data[linecounter].replace("version ","",1).replace("\n","") elif(read_data[linecounter].startswith("name")): = read_data[linecounter].replace("name ","",1).replace("\n","") elif(read_data[linecounter].startswith("created by")): authoranddate = read_data[linecounter].replace("created by ","",1).replace("\n","") # INFO: might seem complex but is necessary cause a user could have " on " as part of his name # which would perhaps even kill the REO Importer of the Riot Engine # TODO: Right now we assume a fixed length for the date. Perhaps a more flexible way of solving this split would be nice REO.creationDate = authoranddate[-10:] = authoranddate[:-14] elif(read_data[linecounter].startswith("Lighting")): = read_data[linecounter].replace("Lighting ","",1).replace("\n","") elif(read_data[linecounter].startswith("materials")): materialcount = read_data[linecounter].replace("materials ","",1).replace("\n","") # walk over the next #materialcount lines to collect the materials materials = [] for n in xrange (int(materialcount)): # TODO: right now we assume the format: ID TYPE FILENAME (Filenames without any spaces!) # If there is a model with no texture the entry is ex.: 0 Texture(0) # For these definitions the converter would crash due index out of range entry = read_data[linecounter+n+1].replace("\n","").split(" ") materials.append([entry[1],entry[2]]) REO.materials = materials # jump over #materialcount lines linecounter += int(materialcount) elif(read_data[linecounter].startswith("transform")): # INFO: Assumes that after transform keyword we will have 3 lines representing the matrix REO.transform = [read_data[linecounter+1].replace("\n","").split(" "), read_data[linecounter+2].replace("\n","").split(" "), read_data[linecounter+3].replace("\n","").split(" ")] linecounter+=3 elif(read_data[linecounter].startswith("vertices")): vertexcount = read_data[linecounter].replace("vertices ","",1).replace("\n","") # walk over the next #vertexcount lines to collect the vertices vertices = [] for n in xrange (int(vertexcount)): # TODO: right now we assume the format: ID X Y Z # if there are other formats, for example including the W entry, this needs to be changed entry = read_data[linecounter+n+1].replace("\n","").split(" ") vertices.append([entry[1],entry[2],entry[3]]) REO.vertices = vertices # jump over #vertexcount lines linecounter += int(vertexcount) elif(read_data[linecounter].startswith("faces")): facecount = read_data[linecounter].replace("faces ","",1).replace("\n","") # Assumption: There is an empty line after the faces entry. Then we have blocks of 5 or 6 lines per polygon # TODO: The Inner Temple has an entry "fl 2S" in a polygon block which leads to a 6 line block. linecounter+=1 faces = [] # Dictionary of UV-Tex-Coords -> vtid reverseTexDict = {} # VT entries of format [U,V] vtentries = [] # walk over the next #facecount blocks to collect the faces for n in xrange (int(facecount)): # INFO: polygon rows may be ignored as they are only giving structural information vt = read_data[linecounter+n*6+2].replace("\n","") # each face has: used vertices, corresponding UV coords per vertex, used material # face entry consists of: [[vertices],[vtextureids],materialid] face = vt.split(":")[1].split(" ") mat = read_data[linecounter+n*6+3].replace("\n","").split(" ")[1] Utex = read_data[linecounter+n*6+4].replace("\n","").replace("tu ","",1).split(" ") Vtex = read_data[linecounter+n*6+5].replace("\n","").replace("tv ","",1).split(" ") # list holding the corresponding vt ids for the current faces' vertices vtent = [] for i in xrange(len(face)): # If there is no entry in the reverseTexDict for the current UV coords, create # an entry and append UV coords to vtentries list if str([Utex[i],Vtex[i]]) not in reverseTexDict: reverseTexDict[str([Utex[i],Vtex[i]])] = len(vtentries) vtentries.append([Utex[i],Vtex[i]]) # Append the textureid to the list vtent.append(reverseTexDict[str([Utex[i],Vtex[i]])]) # TODO: fl 2S entries are not further investigated yet and will be ignored as for now if(read_data[linecounter+n*6+6].startswith("fl")): linecounter+=1 # add vt entries faces.append([face,vtent,mat]) REO.faces = faces REO.vtentries = vtentries # jump over #facecount * 6 lines linecounter += int(facecount)*6 # TODO: We cannot use the bounding data in the obj format. Right now we will just ignore them. # In future approach, we could parse the bounding spheres / boxes as model data. elif(read_data[linecounter].startswith("bspheres")): bspherecount = read_data[linecounter].replace("bspheres ","",1).replace("\n","") # Assumption again: Seems to have an empty line before the bsphere block like in face def above linecounter+=1 bspheres = [] # walk over the next #bspherecount blocks to collect the bspheres for n in xrange (int(bspherecount)): # Assumption: 3 lines per bsphere bspheres.append([read_data[linecounter+n*4+1].replace("\n",""),read_data[linecounter+n*4+2].replace("\n",""),read_data[linecounter+n*4+3].replace("\n","")]) REO.bspheres = bspheres # jump over #bspherecount * 4 lines linecounter += int(bspherecount)*4 elif(read_data[linecounter].startswith("bboxes")): bboxcount = read_data[linecounter].replace("bboxes ","",1).replace("\n","") # Assumption again: Seems to have an empty line before the bboxes block like in face and bsphere def above linecounter+=1 bboxes = [] # walk over the next #bboxcount blocks to collect the bboxes for n in xrange (int(bboxcount)): # Assumption: 6 lines per bbox bbox =[] for i in xrange(1,7): bbox.append(read_data[linecounter+n*7+i].replace("\n","")) bboxes.append(bbox) REO.bboxes = bboxes # jump over #bboxcount * 7 lines linecounter += int(bboxcount)*7 linecounter += 1 # print REO.__dict__ withTextures = True if(withTextures): # Build the mtl file with open(filename+".mtl", 'w') as f: f.write("# Material file for REO file: ""\n") f.write("\n") # Assumption: Only entries of id Texutre Filename in reo file for index,entry in enumerate(REO.materials): f.write("newmtl Material"+str(index) +"\n") f.write("map_Kd " + entry[1] +"\n") f.write("\n") f.close() # Now lets build an obj file: with open(filename+".obj", 'w') as f: if(withTextures): f.write("mtllib " + filename+".mtl" + "\n") f.write("\n") # vertices for entry in REO.vertices: f.write("v "+" ".join(entry)+"\n") f.write("\n") # vertex texture coords if(withTextures): for en in REO.vtentries: f.write("vt "+" ".join(en)+"\n") f.write("\n") if(withTextures): # INFO: blocks with entries per material -> smaller obj file but changes order of the faces! # face definitions with vt ids # preprocess and build the blocks # dictionary holds lists of face definitions per materialid faceblocks={} # initialize list for index,entry in enumerate(REO.materials): faceblocks[str(index)] = [] for entry in REO.faces: withtexid = [] # associate the vertices with their corresponding vt entry id for index,vert in enumerate(entry[0]): withtexid.append(str(int(vert)+1)+'/'+ str(int(entry[1][index])+1) ) faceblocks[entry[2]].append("f "+" ".join(withtexid)+"\n") # write blocks of faces per material for entry in sorted(faceblocks): f.write("usemtl Material"+entry+"\n") for face in faceblocks[entry]: f.write(face) else: for entry in REO.faces: withtexid = [str(int(d)+1) for d in entry[0]] f.write("f "+" ".join(withtexid)+"\n") f.close()

Code: Select all

''' Created on 08.08.2016 @author: BuXXe ''' import time class REO(object): ''' classdocs ''' def __init__(self): ''' Constructor ''' self.version = "2.2" = "Unnamed Model" = "Anonymous" self.creationDate = time.strftime("%d.%m.%Y") = "1" self.transform = [[1,0,0,0],[0,1,0,0],[0,0,1,0]] # Format: List of Lists with each containing [Type, Filename] self.materials = [] self.vertices = [] self.faces = [] self.vtentries = [] # bounding data self.bspheres = [] self.bboxes = []
Last edited by unterbuchse on Wed Aug 10, 2016 11:13 pm, edited 4 times in total.

User avatar
Arokhs Twin
Site Admin
Posts: 1217
Joined: Wed Jul 18, 2001 9:36 pm
Location: United Kingdom

Re: Python REO. to OBJ. Converter

Postby Arokhs Twin » Mon Aug 08, 2016 5:23 pm

I noticed it has been updated again, will upload fixed version.
By fire and by blood I join with thee in the Order of the Flame!
Webmaster of Arokh's Lair

Posts: 17
Joined: Thu Feb 27, 2014 2:46 am

Re: Python REO. to OBJ. Converter

Postby unterbuchse » Mon Aug 08, 2016 6:06 pm

I already used the most updated version ( Aug 07, 2016) with the QT gui and I still got the wrong model.
In the following screens you see the generated output file contains much more vertices (180) than it should (32).
Theres something definately wrong in this one.
REOtoOBJConverter-Problem2.jpg (215.92 KiB) Viewed 3394 times
(249.05 KiB) Downloaded 1868 times

Posts: 17
Joined: Thu Feb 27, 2014 2:46 am

Re: Python REO. to OBJ. Converter

Postby unterbuchse » Tue Aug 09, 2016 5:11 am

Put some more effort into this and got a version with .mtl file generation and some bug fixes.
Right now theres still a problem for some files (if the vertices are in the file but never used in a polygon...) Will work this one out tomorrow.

Stay tuned.

Side Note: does anyone know what the "fl 2S" entry for a face / polygon means? I just got the theory that it would have to do with 2 sided faces but not sure yet. and one more thing: for bboxes there is a number after the transform martices block (0, or 1). What does this stand for?

** Update **
One of my Assumptions was wrong and backfired hard. Texture coords are broken for now but already got a solution. Hopefully done in some hours.

** Update **
So everythings fixed. Still the fl 2S statement in the polygon blocks will just be ignored. As for now, I could not find a problem in the obj, so if there are problems with the normals of certain faces, this should be checked again.

So the current version is capable of converting REO files to .obj + .mtl files. If you keep the texture bmps in the same directory as the mtl and the obj you can import it into blender for example and it should work out of the box. The current converter source can be found in the first post of this thread. I will put the stuff on github when Ive got some motivation. Next step: .obj to .reo

Problems so far:
There will never (!) be a 1:1 conversion of .obj <-> .reo as they have different approaches. Besides the additional information (metadata, bboxes/bspheres, lighting-model, transform) which could be handled as "comments" in the .obj file, the structure of a .obj file forces changes in the order of the faces compared to the reo file. The obj file uses blocks for each material entry, while the reo handles materials on a "per face" basis. Of course this could be done for the .obj as well but this would mean +1 row for each face and therefor create bigger files. One more thing is the fl 2S entry. If this entry is related to double sided faces, the creation of a reo file out of an obj file would never (or lets say only with much effort) identify such faces and write the fl 2S option. You would need to find faces with same vertex entries but different order (meaning: mirrored normal). I just do not think that this is worth the time. If there will ever be a case of normal problems or poly count problems due to 2 faces for 1 double sided face, we may discuss about it, but not now.

Some more eye-candy:
(179.6 KiB) Downloaded 1833 times
Church-TexturedCorrectly.jpg (192.4 KiB) Viewed 3359 times

Return to “Drakan Level Editing and Game Mods”

Who is online

Users browsing this forum: No registered users and 1 guest