diff --git a/README.md b/README.md
index fdaab98..3dd46e8 100644
--- a/README.md
+++ b/README.md
@@ -1,3 +1,24 @@
 # geopy
 
-Python tool for manipulating .geo files.
\ No newline at end of file
+Python tools for manipulating .geo files. The long term plan is to be able to export .geos direct from Blender.
+
+## geo.py
+Contains the Geo class, which represents the contents of .geo files. Can be run to test the reading and writing functionality.
+
+geo.py <infile.geo> [<outfile.geo>]
+
+If only an input file is specified, it will read the input and dump the contents of the .geo to the console.
+
+If an output file is specified, it will read the input file, and write the contents to the output as a new .geo file.
+
+## stl_dump
+Dumps the meshes of a .geo file to .stl files. Used for testing and validation, as .stl isn't useful for games.
+
+stl_dump.py <file.geo>
+
+Dumps all the meshes contained in <file.geo> to <geo_name>/<model_name.stl>. <geo_name> and <model_name> are read from the .geo.
+
+
+##Known issues:
+ - Not all structures are read from the .geo file. (AltPivotInfo)
+ - Not all structures are regenerated when writing a .geo file. (PolyGrid and Reductions)
diff --git a/bones.py b/bones.py
new file mode 100644
index 0000000..68ea8f3
--- /dev/null
+++ b/bones.py
@@ -0,0 +1,120 @@
+
+
+#List of bone ids. Should be kept strictly ordered.
+BONES_LIST = [
+    "BONEID_HIPS",
+    "BONEID_WAIST",
+    "BONEID_CHEST",
+    "BONEID_NECK",
+    "BONEID_HEAD",
+    "BONEID_COL_R",
+    "BONEID_COL_L",
+    "BONEID_UARMR",
+    "BONEID_UARML",
+    "BONEID_LARMR",
+    "BONEID_LARML",
+    "BONEID_HANDR",
+    "BONEID_HANDL",
+    "BONEID_F1_R",
+    "BONEID_F1_L",
+    "BONEID_F2_R",
+    "BONEID_F2_L",
+    "BONEID_T1_R",
+    "BONEID_T1_L",
+    "BONEID_T2_R",
+    "BONEID_T2_L",
+    "BONEID_T3_R",
+    "BONEID_T3_L",
+    "BONEID_ULEGR",
+    "BONEID_ULEGL",
+    "BONEID_LLEGR",
+    "BONEID_LLEGL",
+    "BONEID_FOOTR",
+    "BONEID_FOOTL",
+    "BONEID_TOER",
+    "BONEID_TOEL",
+
+    "BONEID_FACE",
+    "BONEID_DUMMY",
+    "BONEID_BREAST",
+    "BONEID_BELT",
+    "BONEID_GLOVEL",
+    "BONEID_GLOVER",
+    "BONEID_BOOTL",
+    "BONEID_BOOTR",
+    "BONEID_RINGL",
+    "BONEID_RINGR",
+    "BONEID_WEPL",
+    "BONEID_WEPR",
+    "BONEID_HAIR",
+    "BONEID_EYES",
+    "BONEID_EMBLEM",
+    "BONEID_SPADL",
+    "BONEID_SPADR",
+    "BONEID_BACK",
+    "BONEID_NECKLINE",
+    "BONEID_CLAWL",
+    "BONEID_CLAWR",
+    "BONEID_GUN",
+
+    "BONEID_RWING1",
+    "BONEID_RWING2",
+    "BONEID_RWING3",
+    "BONEID_RWING4",
+
+    "BONEID_LWING1",
+    "BONEID_LWING2",
+    "BONEID_LWING3",
+    "BONEID_LWING4",
+
+    "BONEID_MYSTIC",
+
+    "BONEID_SLEEVEL",
+    "BONEID_SLEEVER",
+    "BONEID_ROBE",
+    "BONEID_BENDMYSTIC",
+
+    "BONEID_COLLAR",
+    "BONEID_BROACH",
+
+    "BONEID_BOSOMR",
+    "BONEID_BOSOML",
+
+    "BONEID_TOP",
+    "BONEID_SKIRT",
+    "BONEID_SLEEVES",
+
+    "BONEID_BROW",
+    "BONEID_CHEEKS",
+    "BONEID_CHIN",
+    "BONEID_CRANIUM",
+    "BONEID_JAW",
+    "BONEID_NOSE",
+
+    "BONEID_HIND_ULEGL",
+    "BONEID_HIND_LLEGL",
+    "BONEID_HIND_FOOTL",
+    "BONEID_HIND_TOEL",
+    "BONEID_HIND_ULEGR",
+    "BONEID_HIND_LLEGR",
+    "BONEID_HIND_FOOTR",
+    "BONEID_HIND_TOER",
+    "BONEID_FORE_ULEGL",
+    "BONEID_FORE_LLEGL",
+    "BONEID_FORE_FOOTL",
+    "BONEID_FORE_TOEL",
+    "BONEID_FORE_ULEGR",
+    "BONEID_FORE_LLEGR",
+    "BONEID_FORE_FOOTR",
+    "BONEID_FORE_TOER",
+
+    "BONEID_LEG_L_JET1",
+    "BONEID_LEG_L_JET2",
+    "BONEID_LEG_R_JET1",
+    "BONEID_LEG_R_JET2",
+]
+
+BONES_LOOKUP = {}
+
+for i in range(len(BONES_LIST)):
+    BONES_LOOKUP[BONES_LIST[i]] = i
diff --git a/geo.py b/geo.py
new file mode 100755
index 0000000..aaa3acd
--- /dev/null
+++ b/geo.py
@@ -0,0 +1,1137 @@
+#! /bin/python3
+
+import struct
+import zlib
+import sys
+import traceback
+import math
+from bones import *
+
+#Ver 0 .geo pre-header:
+#Offset	Size	Description
+#0	4	'ziplen' ? Compressed header data size + 4
+#4	4	Header size (Must be non-zero for old formats.)
+#8	variable	zlib Compressed header data
+#variable	4	Dummy data?
+
+#Version 2+ .geo pre-header:
+#Offset	Size	Description
+#0	4	'ziplen' ? Compressed header data size + 12
+#4	4	Legacy header size (Must be 0 for new file formats.)
+#8	4	Version number 2 to 5, and 7 to 8
+#12	4	Header size
+#16	variable	zlib Compressed header data
+
+
+
+#.geo header
+#Offset	Size	Description
+#0	4	i32: gld->datasize
+#4	4	i32: texname_blocksize
+#8	4	i32: objname_blocksize
+#12	4	i32: texidx_blocksize
+#16	4	i32: lodinfo_blocksize (only present in version 2 to 6)
+#?	texname_blocksize	PackNames block of texture names
+#?	objname_blocksize	objnames
+#?	texidx_blocksize	texidx
+#?	lodinfo_blocksize	lodinfo (only present in version 2 to 6)
+#?	ModelHeader	???
+
+
+#PackNames block:
+#Size	Description
+#4	name_count
+#4*name_count	offset of the name, relative to the start of name_block
+#variable	name_block (ASCIIZ?)
+
+#ModelHeader:
+#Size	Description
+#124	ASCIIZ name.
+#4	Model pointer (0 in file?)
+#4	Float track length (technically a fallback if not specified elsewhere?)
+#4	Pointer to an array of Model pointers (-1 in file?)
+#*note: pointers aren't really needed.
+
+#Model:
+#version 0 to 2: (struct ModelFormatOnDisk_v2: "Common/seq/anim.c" )
+#Offset	Size	Description
+#0	4	u32: Flags
+#4	4	f32: radius
+#8	4	ptr: VBO?
+#12	4	i32: tex_count
+#16	2	i16: id  ("I am this bone" ?)
+#18	1	u8: blend_mode
+#19	1	u8: loadstate
+#20	4	ptr(BoneInfo)
+#24	4	ptr(TrickNode)
+#28	4	i32: vert_count
+#32	4	i32: tri_count
+#36	4	ptr: TexID
+#40	32	PolyGrid: polygrid
+#72	4	ptr(CTri): ctris ?
+#76	4	ptr(i32): tags
+#80	4	ptr(char): name
+#84	4	ptr(AltPivotInfo): api
+#88	4	ptr(ModelExtra): extra
+#92	12	Vec3: scale
+#104	12	Vec3: min
+#116	12	Vec3: max
+#120	4	ptr(GeoLoadData): gld
+#124	84	PackBlockOnDisk
+#208
+
+#version 3+: "Common/seq/anim.c"
+#	Offset	Size	Description
+#	0	4	i32: size
+#	4	4	f32: radius
+#	8	4	i32: tex_count
+#	12	4	ptr(BoneInfo): boneinfo ?
+#	16	4	i32: vert_count
+#	20	4	i32: tri_count
+#ver 8+	24	4	i32: reflection_quad_count
+#	+4	4	i32: tex_idx
+#	+4	32	PolyGrid: grid
+#	+32	4	ptr(char): name
+#	+4	4	ptr(AltPivotInfo):
+#	+4	12	Vec3: scale
+#	+12	12	Vec3: min
+#	+12	12	Vec3: max
+#	+12	12	PackData: pack.tris
+#	+12	12	PackData: pack.verts
+#	+12	12	PackData: pack.norms
+#	+12	12	PackData: pack.sts
+#	+12	12	PackData: pack.sts3
+#	+12	12	PackData: pack.weights
+#	+12	12	PackData: pack.matidxs
+#	+12	12	PackData: pack.grid
+#Ver 4	+12	12	PackData: pack.lmap_utransforms
+#Ver 4	+12	12	PackData: pack.lmap_vtransforms
+#Ver 7+	+12	12	PackData: pack.reductions
+#Ver 8+	+12	12	PackData: pack.reflection_quads
+#Ver 7+	+12	12	f32[3]: autolod_dists
+#	+12	2	i16: id
+
+#PackBlockOnDisk: "Common/seq/anim.c"
+#Offset	Size	Description
+#0	12	PackData: tris
+#12	12	PackData: verts
+#24	12	PackData: norms
+#36	12	PackData: sts
+#48	12	PackData: weights
+#60	12	PackData: matidxs
+#72	12	PackData: grid
+#84
+
+#PackData:
+#Offset	Size	Description
+#0	4	i32: packsize, The compressed size of this data block. 0 if this is uncompressed.
+#4	4	u32: unpacksize, The size of this data block when uncompressed.
+#8	4	ptr(u8): data, The offset of this block of data inside the .geo's main data block. 
+#12
+
+
+#struct PolyGrid: "libs/UtilitiesLib/components/gridpoly.h"
+#Offset	Size	Description
+#0	4	ptr(PolyCell): cell
+#4	12	Vec3: pos
+#16	4	f32: size
+#20	4	f32: inv_size
+#24	4	i32: tag
+#28	4	i32: num_bits
+#32
+
+
+#PolyCell:
+#Offset	Size	Description
+#0	4	ptr(ptr(PolyCell)): children
+#4	4	ptr(u16): tri_idxs, Triangle indexes.
+#8	4	i32: tri_count
+
+#BoneInfo
+#Offset Size	Description
+#0	4	i32: numbones
+#4	4*15	i32[15]: bone_ID, Bones used by this geometry.
+#64
+
+#Reductions
+#Size	Description
+#4	int: num_reductions
+#num_reductions*4	int[]:num_tris_left
+#num_reductions*4	int[]:error_values
+#num_reductions*4	int[]:remap_counts
+#num_reductions*4	int[]:changes_counts
+#4	int: total_remaps
+#total_remaps*4 	int[]:remaps
+#4	int: total_remap_tris
+#total_remap_tris*4	int[]:remap_tris
+#4	int: total_changes
+#total_changes*4	int[]:changes
+#4	int: positions_delta_length
+#positions_delta_length	byte[]: compressDelta(positions, 3, total_changes, "f", 0x8000)
+#4	int: total_changes_delta_length
+#total_changes_delta_length	byte[]: compressDelta(changes, 2, total_changes, "f", 0x1000)
+
+ZERO_BYTE = struct.pack("B", 0)
+#def unbyte(v):
+#    return struct.unpack("B", v)[0]
+if sys.version_info[0] < 3:
+    byte = chr
+    unbyte = ord
+else:
+    def byte(v):
+        #return struct.pack("B", v)
+        return bytes((v,))
+    def unbyte(v):
+        return v
+
+def unpackNames(data):
+    """Extract strings from data, as layed out as a PackNames structure."""
+    (length, ) = struct.unpack("<i", data[0:4])
+    strings = []
+    data_strings = data[length * 4 + 4:]
+    for i in range(length):
+        #todo: scan string for zero
+        (offset, ) = struct.unpack("<i", data[i * 4 + 4: i * 4 + 4 + 4])
+        for j in range(offset, len(data_strings)):
+            if unbyte(data_strings[j]) == 0:
+                strings.append(data_strings[offset:j])
+                break
+            pass
+        pass
+    return strings
+def packNames(strings):
+    """Convert a sequence of strings into a PackNames structure."""
+    data_strings = b""
+    data = struct.pack("<i", len(strings))
+    for s in strings:
+        data += struct.pack("<i", len(data_strings))
+        data_strings += s + ZERO_BYTE
+    return data + data_strings
+
+def unpackStrings(data):
+    """Convert a block of ASCIIZ strings into list of strings."""
+    strings = []
+    start = 0
+    for i in range(len(data)):
+        if unbyte(data[i]) == 0:
+            strings.append(data[start:i])
+            start = i + 1
+            pass
+        pass
+    return strings
+def packStrings(strings):
+    """Convert a list of strings into a block of ASCIIZ strings"""
+    data = b""
+    for s in strings:
+        data += s + ZERO_BYTE
+    return data
+
+def extractString(data, offset = 0):
+    """Extract an ASCIIZ string from a fixed length buffer."""
+    for i in range(offset, len(data)):
+        if unbyte(data[i]) == 0:
+            return data[offset : i]
+        pass
+    return data
+def storeString(string, length):
+    """Store an ASCIIZ string into a fixed length buffer. If the string is longer than the buffer the output will be truncated, with \\x00 character guarenteed at the end."""
+    if len(string) >= length:
+        return string[0 : length - 1] + ZERO_BYTE
+    else:
+        return string + ZERO_BYTE * (length - len(string))
+
+def uncompressDeltas(src, stride, count, pack_type):
+    """Expand a list a delta compressed with deltas
+
+src: Source data containg 3 parts:
+  - delta codes: Contain 'count' * 'stride' 2 bit size code fields. The end of the block is padded to the byte boundary.
+  - float scale: A byte with the exponent for scaling floating point deltas. This part is present with integer types, but unused.
+  - byte data: This is read and processed according to the codes specified in the delta codes block.
+stride: the number of channels in the data (1 to 3)
+count: the number off elements in the data
+pack_type: Is a character indicating the type of packing. Valid values are "f" (float32), "H" (unsigned16), "I" (unsigned32) (these match the struct modules types).
+The returned data is an array of arrays, with the inner most arrays being 'stride' long, and the outer array being 'count' in length.
+"""
+    if len(src) <= 0:
+        return None
+    #Compute offset to the byte data after all the bit fields.
+    byte_offset = int((2 * count * stride + 7) / 8)
+    #print("stride: %d  count: %d  pack_type: %s" % (stride, count, pack_type))
+    #print("src:%s" % ([src],))
+    float_scale = float(1 << unbyte(src[byte_offset]))
+    float_scale_inv = 1.0 / float_scale
+    byte_offset += 1
+    current_bit = 0
+    if pack_type in "f":
+        fLast = [0.0] * stride
+    else:
+        iLast = [0] * stride
+    out_data = []
+    for i in range(count):
+        row_data = []
+        for j in range(stride):
+            code = (unbyte(src[current_bit >> 3]) >> (current_bit & 0x7)) & 0x3
+            current_bit += 2
+            if code == 0:
+                iDelta = 0
+            elif code == 1:
+                iDelta = unbyte(src[byte_offset]) - 0x7f
+                byte_offset += 1
+            elif code == 2:
+                iDelta = (
+                    unbyte(src[byte_offset]) |
+                    (unbyte(src[byte_offset + 1]) << 8)
+                ) - 0x7fff
+                byte_offset += 2
+            elif code == 3:
+                (iDelta, ) = struct.unpack("<i", src[byte_offset : byte_offset + 4])
+                byte_offset += 4
+            if pack_type == "f":
+                if code == 3:
+                    (fDelta, ) = struct.unpack("<f", src[byte_offset - 4 : byte_offset])
+                else:
+                    fDelta = iDelta * float_scale_inv
+                fLast[j] += fDelta
+                row_data.append(fLast[j])
+            elif pack_type == "I":
+                iLast[j] += iDelta + 1
+                iLast[j] &= 0xffffffff
+                row_data.append(iLast[j])
+            elif pack_type == "H":
+                iLast[j] += iDelta + 1
+                iLast[j] &= 0xffff
+                row_data.append(iLast[j])
+        out_data.append(row_data)
+    return out_data
+
+def quantF32(v, float_scale, float_scale_inv):
+    i = int(v * float_scale)
+    i &= ~1
+    return i * float_scale_inv
+def compressDeltas(src, stride, count, pack_type, float_scale):
+    if float_scale != 0:
+        float_scale_inv = 1.0 / float_scale
+    else:
+        float_scale_inv = 1.0
+    codes = [0] * (stride * count + 3) #over fill it by 3 for easier conversion into byte data
+    float_scale_byte = byte(int(math.log(float_scale, 2)))
+    bytes_data = b""
+    if pack_type in "f":
+        fLast = [0.0] * stride
+    else:
+        iLast = [0] * stride
+        if pack_type == "I":
+            iMask = 0xffffffff
+        else:
+            iMask = 0xffff
+
+    k = 0
+    for i in range(count):
+        for j in range(stride):
+            if pack_type == "f":
+                fDelta = quantF32(src[i][j], float_scale, float_scale_inv) - fLast[j]
+                val8 = int(fDelta * float_scale + 0x7f)
+                val16 = int(fDelta * float_scale + 0x7fff)
+            else:
+                t = src[i][j]
+                iDelta = t - iLast[j] - 1
+                iLast[j] = t
+                val8 = iDelta + 0x7f
+                val16 = iDelta + 0x7fff
+                val32 = iDelta;
+            if val8 == 0x7f:
+                codes[k] = 0
+            elif val8 & ~0xff == 0:
+                codes[k] = 1
+                bytes_data += byte(val8)
+                if pack_type == "f":
+                    fLast[j] = (val8 - 0x7f) * float_scale_inv + fLast[j];
+            elif pack_type == "H" or val16 & ~0xffff == 0:
+                codes[k] = 2
+                bytes_data += struct.pack("<H", val16)
+                if pack_type == "f":
+                    fLast[j] = (val16 - 0x7fff) * float_scale_inv + fLast[j];
+            elif pack_type == "I":
+                codes[k] = 3
+                bytes_data += struct.pack("<I", val32 & iMask)
+            else:
+                codes[k] = 3
+                bytes_data += struct.pack("<f", fDelta)
+                fLast[j] = fDelta + fLast[j]
+            k += 1
+                
+        
+    code_data = b""
+    for i in range(0, stride * count, 4):
+        v = ((codes[i + 0] << 0) |
+             (codes[i + 1] << 2) |
+             (codes[i + 2] << 4) |
+             (codes[i + 3] << 6))
+        code_data += byte(v)
+    return code_data + float_scale_byte + bytes_data
+
+def inferSizes(size, pack_list, other_list):
+    """Infer the size of the blocks in 'other_list' as infered by the total size and items in 'pack_list'."""
+    starts = []
+    for p in pack_list:
+        s = p[2]
+        if s in starts:
+            continue
+        starts.append(s)
+    for o in other_list:
+        if o in starts:
+            continue
+        starts.append(o)
+    starts.sort()
+    output = []
+    for o in other_list:
+        if o == 0:
+            output.append( (0, 0, 0) )
+            continue
+        i = starts.index(o) + 1
+        if i < 0 or i >= len(starts):
+            output.append( (0, 0, o) )
+            continue
+        output.append( (0, starts[i] - o, o) )
+        
+    return tuple(output)
+
+class PolyCell:
+    def __init__(self, model):
+        self.model = model
+        self.children = None
+        self.tri_idxs = []
+        self.tri_count = 0
+    def decode(self, data, offset):
+        (children_offset, tri_idxs_offset, self.tri_count) = struct.unpack("<iii", data[offset : offset + 12])
+        if children_offset == 0:
+            self.children = None
+        else:
+            self.children = [None] * 8
+            children_offsets = struct.unpack("<iiiiiiii", data[children_offset : children_offset + 4 * 8])
+            for i in range(8):
+                if children_offsets[i] == 0:
+                    continue
+                self.children[i] = PolyCell(self.model)
+                self.children[i].decode(data, children_offsets[i])
+        if tri_idxs_offset == 0:
+            self.tri_idxs = []
+        else:
+            self.tri_idxs = struct.unpack("<" + "H" * self.tri_count, data[tri_idxs_offset : tri_idxs_offset + self.tri_count * 2])
+    def dump(self, indent):
+        print(indent + "tri_idxs: %s" % ([self.tri_idxs], ))
+        print(indent + "children? %s" % (self.children is not None, ))
+        if self.children is not None:
+            for c in self.children:
+                if c is not None:
+                    c.dump(indent + "    ")
+
+class PolyGrid:
+    def __init__(self, model):
+        self.model = model
+        self.cell = None
+        pass
+    def parsePolyGridData(self, data):
+        self.cell = PolyCell(self.model)
+        self.cell.decode(data, 0)
+        pass
+    def dump(self):
+        if self.cell is not None:
+            self.cell.dump("        ")
+
+class Data:
+    def __init__(self, rawdata = b"", off = 0):
+        self.data = rawdata
+        self.offset = off
+    def setData(self, rawdata, off = 0):
+        self.data = rawdata
+        self.offset = off
+    def seek(self, off):
+        self.offset = off
+    def seekEnd(self, off = 0):
+        self.offset = len(self.data) + off
+    def seekRel(self, off = 0):
+        self.offset += off
+    def decode(self, fmt):
+        size = struct.calcsize(fmt)
+        val = struct.unpack(fmt, self.data[self.offset : self.offset + size])
+        self.offset += size
+        return val
+    def encode(self, fmt, *args):
+        data = struct.pack(fmt, *args)
+        self.write(data)
+    def read(self, length):
+        val = self.data[self.offset : self.offset + length]
+        self.offset += length
+        return val
+    def write(self, data):
+        #todo: handle
+        if self.offset == len(self.data):
+            self.data += data
+            self.offset = len(self.data)
+        elif self.offset > len(self.data):
+            self.data += ZERO_BYTE * (self.offset - len(self.data))
+            self.data += data
+            self.offset = len(self.data)
+        elif self.offset + len(data) >= len(self.data):
+            self.data = self.data[0 : self.offset] + data
+        else:
+            self.data = self.data[0 : self.offset] + data + self.data [self.offset + len(data) : ]
+    def truncate(self, offset = None):
+        if offset is None:
+            offset = self.offset
+        self.data[0 : offset]
+
+            
+class Reductions:
+    def __init__(self, model):
+        self.model = model
+    def decode(self, data):
+        (self.num_reductions, ) = data.decode("<i")
+        self.num_tris_left = data.decode("<" + "i" * self.num_reductions)
+        self.error_values = data.decode("<" + "f" * self.num_reductions)
+        self.remap_counts = data.decode("<" + "i" * self.num_reductions)
+        self.changes_counts = data.decode("<" + "i" * self.num_reductions)
+        (self.total_remaps, ) = data.decode("<i")
+        self.remaps = data.decode("<" + "i" * (self.total_remaps * 3))
+        (self.total_remap_tris, ) = data.decode("<i")
+        self.remap_tris = data.decode("<" + "i" * self.total_remap_tris)
+        (self.total_changes, ) = data.decode("<i")
+        self.changes = data.decode("<" + "i" * self.total_changes)
+        (positions_delta_length, ) = data.decode("<i")
+        #print("positions_delta_length: %s" % (positions_delta_length, ))
+        #print("data remaining: %s" %([data.data[data.offset : ]], ))
+        #self.dump()
+        self.positions = uncompressDeltas(data.read(positions_delta_length), 3, self.total_changes, "f")
+        (tex1s_delta_length, ) = data.decode("<i")
+        self.tex1s = uncompressDeltas(data.read(tex1s_delta_length), 2, self.total_changes, "f")
+        #print("data remaining: %s" %([data.data[data.offset : ]], ))
+    def encode(self):
+        data = b""
+        data += struct.pack("<i", self.num_reductions)
+        data += struct.pack("<" + "i" * self.num_reductions, *self.num_tris_left)
+        data += struct.pack("<" + "f" * self.num_reductions, *self.error_values)
+        data += struct.pack("<" + "i" * self.num_reductions, *self.remap_counts)
+        data += struct.pack("<" + "i" * self.num_reductions, *self.changes_counts)
+        data += struct.pack("<i", self.total_remaps)
+        data += struct.pack("<" + "i" * (self.total_remaps * 3), *self.remaps)
+        data += struct.pack("<i", self.total_remap_tris)
+        data += struct.pack("<" + "i" * self.total_remap_tris, *self.remap_tris)
+        data += struct.pack("<i", self.total_changes)
+        data += struct.pack("<" + "i" * self.total_changes, *self.changes)
+        d = compressDeltas(self.positions, 3, self.total_changes, "f", 0x8000)
+        data += struct.pack("<i", len(d))
+        data += d
+        d = compressDeltas(self.tex1s, 2, self.total_changes, "f", 0x1000)
+        data += struct.pack("<i", len(d))
+        data += d
+        return data
+    def dump(self):
+        print("    reductions:")
+        print("        num_reductions: %s" % (self.num_reductions, ))
+        print("        num_tris_left: %s" % (self.num_tris_left, ))
+        print("        error_values: %s" % (self.error_values, ))
+        print("        remap_counts: %s" % (self.remap_counts, ))
+        print("        changes_counts: %s" % (self.changes_counts, ))
+        print("        total_remaps: %s" % (self.total_remaps, ))
+        print("        remaps: %s" % (self.remaps, ))
+        print("        total_remap_tris: %s" % (self.total_remap_tris, ))
+        print("        remap_tris: %s" % (self.remap_tris, ))
+        print("        total_changes: %s" % (self.total_changes, ))
+        print("        changes: %s" % (self.changes, ))
+        print("        positions: %s" % (self.positions, ))
+        print("        tex1s: %s" % (self.tex1s, ))
+        
+class Model:
+    def __init__(self, geo):
+        self.geo = geo
+        self.radius = None
+        self.tex_count = None
+        self.vert_count = None
+        self.tri_count = None
+        self.reflection_quad_count = None
+        self.tex_idx = None
+        pass
+    def parseHeaderDataV2(self):
+        (self.flags, ) = self.geo.getHeaderElement("<I")
+        (self.radius, ) = self.geo.getHeaderElement("<f")
+        self.geo.header_offset += 4 #ptr(VBO)
+        #print("VBO: %s" % (self.geo.getHeaderElement("<i"), ))
+        (self.tex_count, ) = self.geo.getHeaderElement("<i")
+        (self.id, blend_mode, loadstate) = self.geo.getHeaderElement("<hBB")
+        #print("blend_mode, loadstate: %d, %d" % (blend_mode, loadstate))
+        (self.boneinfo_ptr, ) = self.geo.getHeaderElement("<i")
+        self.geo.header_offset += 4 #ptr(TrickNode)
+        #print("TrickNode: %s" % (self.geo.getHeaderElement("<i"), ))
+        (self.vert_count, self.tri_count) = self.geo.getHeaderElement("<ii")
+        (self.tex_idx_ptr, ) = self.geo.getHeaderElement("<i")
+        self.grid_header = self.geo.getHeaderElement("<ifffffii")
+        self.geo.header_offset += 4 #ptr(CTri)
+        self.geo.header_offset += 4 #ptr(tags)
+        #print("CTri: %s" % (self.geo.getHeaderElement("<i"), ))
+        #print("Tags: %s" % (self.geo.getHeaderElement("<i"), ))
+        (self.name_ptr, self.api_ptr) = self.geo.getHeaderElement("<ii")
+        #print("???: %s" % (self.geo.getHeaderElement("<f"), ))
+        self.geo.header_offset += 4 #ptr(ModelExtra)
+        self.scale = list(self.geo.getHeaderElement("<fff"))
+        self.min = list(self.geo.getHeaderElement("<fff"))
+        self.max = list(self.geo.getHeaderElement("<fff"))
+        self.geo.header_offset += 4 #ptr(GeoLoadData)
+        #print("GeoLoadData: %s" % (self.geo.getHeaderElement("<i"), ))
+        self.pack_tris = self.geo.getHeaderElement("<iii")
+        self.pack_verts = self.geo.getHeaderElement("<iii")
+        self.pack_norms = self.geo.getHeaderElement("<iii")
+        self.pack_sts = self.geo.getHeaderElement("<iii")
+        self.pack_sts3 = (0, 0, 0)
+        self.pack_weights = self.geo.getHeaderElement("<iii")
+        self.pack_matidxs = self.geo.getHeaderElement("<iii")
+        self.pack_grid = self.geo.getHeaderElement("<iii")
+        self.pack_reductions = (0, 0, 0)
+        self.pack_reflection_quads = (0, 0, 0)
+
+        self.name = extractString(self.geo.header_objname_data, self.name_ptr)
+        self.autolod_dists = [-1, -1, -1]
+        self.skipped_data = None
+        #self.dump()
+    def parseHeaderData(self):
+        if self.geo.version <= 2:
+            self.parseHeaderDataV2()
+            return
+        (size, ) = struct.unpack("<i", self.geo.header_data[self.geo.header_offset: self.geo.header_offset + 4])
+        final_offset = self.geo.header_offset + size
+        self.geo.header_offset += 4
+        (self.radius, ) = self.geo.getHeaderElement("<f")
+        (self.tex_count, self.boneinfo_ptr) = self.geo.getHeaderElement("<ii")
+        (self.vert_count, self.tri_count) = self.geo.getHeaderElement("<ii")
+        if self.geo.version >= 8:
+            (self.reflection_quad_count, ) = self.geo.getHeaderElement("<i")
+        (self.tex_idx_ptr, ) = self.geo.getHeaderElement("<i")
+        self.grid_header = self.geo.getHeaderElement("<ifffffii")
+        (self.name_ptr, self.api_ptr) = self.geo.getHeaderElement("<ii")
+        self.scale = list(self.geo.getHeaderElement("<fff"))
+        self.min = list(self.geo.getHeaderElement("<fff"))
+        self.max = list(self.geo.getHeaderElement("<fff"))
+        self.pack_tris = self.geo.getHeaderElement("<iii")
+        self.pack_verts = self.geo.getHeaderElement("<iii")
+        self.pack_norms = self.geo.getHeaderElement("<iii")
+        self.pack_sts = self.geo.getHeaderElement("<iii")
+        self.pack_sts3 = self.geo.getHeaderElement("<iii")
+        self.pack_weights = self.geo.getHeaderElement("<iii")
+        self.pack_matidxs = self.geo.getHeaderElement("<iii")
+        self.pack_grid = self.geo.getHeaderElement("<iii")
+        if self.geo.version == 4:
+            #pack.lmap_utransforms, pack.lmap_vtransforms
+            self.geo.header_offset += 12 * 2
+        if self.geo.version >= 7:
+            self.pack_reductions = self.geo.getHeaderElement("<iii")
+        else:
+            self.pack_reductions = (0, 0, 0)
+        if self.geo.version >= 8:
+            self.pack_reflection_quads = self.geo.getHeaderElement("<iii")
+        else:
+            self.pack_reflection_quads = (0, 0, 0)
+        pack_list = [
+            self.pack_tris,
+            self.pack_verts,
+            self.pack_norms,
+            self.pack_sts,
+            self.pack_sts3,
+            self.pack_weights,
+            self.pack_matidxs,
+            self.pack_grid,
+            #pack.lmap_utransforms, pack.lmap_vtransforms
+            self.pack_reductions,
+            self.pack_reflection_quads,
+        ]
+        other_list = [
+            self.boneinfo_ptr,
+            self.name_ptr,
+            self.api_ptr,
+        ]
+        (self.pack_boneinfo, self.pack_name, self.pack_api) = inferSizes(self.geo.main_data_size, pack_list, other_list)
+        self.name = extractString(self.geo.header_objname_data, self.name_ptr)
+
+        #self.geo.header_offset = final_offset - 12 - 2
+        if self.geo.version >= 7:
+            self.autolod_dists = list(self.geo.getHeaderElement("<fff"))
+        else:
+            self.autolod_dists = [-1, -1, -1]
+        #(self.mystery, ) = self.geo.getHeaderElement("<f")
+        (self.id, ) = self.geo.getHeaderElement("<h")
+        #self.mystery = self.geo.getHeaderElement("<fh")
+        #print("automatic offset: %d  final offset: %d" % (self.geo.header_offset, final_offset))
+        self.skipped_data = self.geo.header_data[self.geo.header_offset : final_offset]
+        self.geo.header_offset = final_offset
+    def parseData(self):
+        self.tris_data = self.geo.getDataBlock(self.pack_tris)
+        self.verts_data = self.geo.getDataBlock(self.pack_verts)
+        self.norms_data = self.geo.getDataBlock(self.pack_norms)
+        self.sts_data = self.geo.getDataBlock(self.pack_sts)
+        self.sts3_data = self.geo.getDataBlock(self.pack_sts3)
+        self.weights_data = self.geo.getDataBlock(self.pack_weights)
+        self.matidxs_data = self.geo.getDataBlock(self.pack_matidxs)
+        self.grid_data = self.geo.getDataBlock(self.pack_grid)
+        self.reductions_data = self.geo.getDataBlock(self.pack_reductions)
+        self.reflection_quads_data = self.geo.getDataBlock(self.pack_reflection_quads)
+
+        self.tris = uncompressDeltas(self.tris_data, 3, self.tri_count, "I")
+        self.verts = uncompressDeltas(self.verts_data, 3, self.vert_count, "f")
+        self.norms = uncompressDeltas(self.norms_data, 3, self.vert_count, "f")
+        self.sts = uncompressDeltas(self.sts_data, 2, self.vert_count, "f")
+        self.sts3 = uncompressDeltas(self.sts3_data, 2, self.vert_count, "f")
+
+        if self.boneinfo_ptr != 0:
+            self.geo.seekMainData(self.boneinfo_ptr)
+            (self.bone_count, ) = self.geo.getMainElement("<i")
+            self.bone_ids = self.geo.getMainElement("<" + ("i" * 15))
+        else:
+            self.bone_count = 0
+            self.bone_ids = [0] * 15
+        if len(self.weights_data) == 0:
+            self.weights = None
+            self.weight_bones = None
+        else:
+            self.weights = []
+            self.weight_bones = []
+            for i in range(self.vert_count):
+                w = unbyte(self.weights_data[i]) / 255.0
+                b = self.matidxs_data[i * 2 : i * 2 + 2]
+                b1 = self.bone_ids[unbyte(b[0]) // 3]
+                b2 = self.bone_ids[unbyte(b[1]) // 3]
+                self.weights.append([w, 1 - w])
+                self.weight_bones.append([b1, b2])
+
+        if len(self.grid_data) == 0:
+            self.polygrid = None
+        else:
+            self.polygrid = PolyGrid(self)
+            self.polygrid.parsePolyGridData(self.grid_data)
+
+        if len(self.reductions_data) == 0:
+            self.reductions = None
+        else:
+            #self.reductions = None
+            #self.dump()
+            self.reductions = Reductions(self)
+            self.reductions.decode(Data(self.reductions_data))
+        if len(self.reflection_quads_data) == 0:
+            self.reflection_quads = None
+        else:
+            #todo:
+            self.reflection_quads = None
+        #if self.geo.version <= 2:
+        #    self.dump()
+
+    def packDeltas(self, data, stride, count, pack_type):
+        deltas = compressDeltas(data, stride, count, pack_type)
+        pack = self.geo.encodeMainDataPacked(deltas)
+    def rebuildWeightsAndBones(self):
+        self.weights_data = b""
+        self.matidxs_data = b""
+        if self.weights is None or len(self.weights) == 0:
+            self.bone_count = 0
+            self.bone_ids = [0] * 15
+            return
+        self.bone_ids = []
+        bone_lookup = {}
+        for wb in self.weight_bones:
+            for wbx in wb:
+                if wbx not in bone_lookup:
+                    bone_lookup[wbx] = len(self.bone_ids)
+                    self.bone_ids.append(wbx)
+        #print(self.weights)
+        for i in range(len(self.weights)):
+            #Copy values to avoid corruption.
+            w = self.weights[i][:]
+            wb = self.weight_bones[i][:]
+            if len(w) > 2:
+                #.geos only support 2 weights per vertex.
+                #todo: get only largest weights?
+                w = w[0 : 2]
+                wb = wb[0 : 2]
+            if len(w) == 1 or wb[0] == wb[1]:
+                self.weights_data += byte(255)
+                self.matidxs_data += byte(bone_lookup[wb[0]] * 3) + ZERO_BYTE
+                continue
+            if w[0] + w[1] == 0:
+                w[0] = 0.5
+                w[1] = 0.5
+            w[0] = w[0] / float(w[0] + w[1])
+            if w[0] < 0:
+                w[0] = 0.0
+            elif w[0] > 1:
+                w[0] = 1.0
+            #print("weights: %s -> %s" % (self.weights[i], w))
+            self.weights_data += byte(int(math.floor(255 * w[0] + 0.5)))
+            self.matidxs_data += byte(bone_lookup[wb[0]] * 3) +byte(bone_lookup[wb[1]] * 3)
+        self.bone_count = len(self.bone_ids)
+        self.bone_ids += [0] * (15 - self.bone_count)
+    def encode(self):
+        #Regenerate dynamic data
+        #todo: build PolyGrid
+        if self.geo.version >= 7:
+            #todo: build reductions
+            if self.reductions is not None:
+                self.reductions_data = self.reductions.encode()
+            else:
+                self.reductions_data = b""
+            pass
+        self.rebuildWeightsAndBones()
+
+        #Encode data into the main block.
+        #note: PackData should go first, otherwise other ptr types might point to the start of the data block, which would result in a 0 point, which is treated as not present.
+        self.pack_tris = self.geo.encodeMainDataPackedDeltas(self.tris, 3, len(self.tris), "I", 1)
+        self.pack_verts = self.geo.encodeMainDataPackedDeltas(self.verts, 3, len(self.verts), "f", 0x8000)
+        self.pack_norms = self.geo.encodeMainDataPackedDeltas(self.norms, 3, len(self.norms), "f", 0x100)
+        self.pack_sts = self.geo.encodeMainDataPackedDeltas(self.sts, 2, len(self.verts), "f", 0x1000)
+        self.pack_sts3 = self.geo.encodeMainDataPackedDeltas(self.sts3, 2, len(self.verts), "f", 0x8000)
+        self.pack_weights = self.geo.encodeMainDataPacked(self.weights_data)
+        self.pack_matidxs = self.geo.encodeMainDataPacked(self.matidxs_data)
+        self.pack_grid = self.geo.encodeMainDataPacked(self.grid_data)
+        self.pack_reductions = self.geo.encodeMainDataPacked(self.reductions_data)
+        self.pack_reflection_quads = self.geo.encodeMainDataPacked(self.reflection_quads_data)
+        bone_data = struct.pack("<" + "i" * (1 + 15), self.bone_count, *self.bone_ids)
+        self.boneinfo_ptr = self.geo.encodeMainData(bone_data)
+
+        #Encode shared header data
+        print("unhandled: model.texidx_ptr !")
+        self.texidx_ptr = 0
+        self.name_ptr = len(self.geo.header_objname_data)
+        self.geo.header_objname_data += self.name + ZERO_BYTE
+        self.geo.header_objnames.append(self.name)
+        print("unhandled: model.api_ptr !")
+        self.api_ptr = 0
+        
+        #Encode the header
+        self.header_data = b""
+        self.header_data += struct.pack("<f", self.radius)
+        self.header_data += struct.pack("<i", self.tex_count)
+        self.header_data += struct.pack("<i", self.boneinfo_ptr)
+        self.header_data += struct.pack("<i", self.vert_count)
+        self.header_data += struct.pack("<i", self.tri_count)
+        if self.geo.version >= 8:
+            if self.reflection_quads is None or len(self.reflection_quads):
+                self.reflection_quads_count = 0
+            else:
+                self.reflection_quads_count = len(self.reflection_quads)
+            self.header_data += struct.pack("<i", self.reflection_quads_count)
+            pass
+        self.header_data += struct.pack("<i", self.texidx_ptr)
+        self.header_data += struct.pack("<ifffffii", *self.grid_header)
+        self.header_data += struct.pack("<i", self.name_ptr)
+        self.header_data += struct.pack("<i", self.api_ptr)
+        self.header_data += struct.pack("<fff", *self.scale)
+        self.header_data += struct.pack("<fff", *self.min)
+        self.header_data += struct.pack("<fff", *self.max)
+        self.header_data += struct.pack("<iii", *self.pack_tris)
+        self.header_data += struct.pack("<iii", *self.pack_verts)
+        self.header_data += struct.pack("<iii", *self.pack_norms)
+        self.header_data += struct.pack("<iii", *self.pack_sts)
+        self.header_data += struct.pack("<iii", *self.pack_sts3)
+        self.header_data += struct.pack("<iii", *self.pack_weights)
+        self.header_data += struct.pack("<iii", *self.pack_matidxs)
+        self.header_data += struct.pack("<iii", *self.pack_grid)
+        if self.geo.version == 4:
+            #pack.lmap_utransforms, pack.lmap_vtransforms
+            self.header_data += struct.pack("<iiiiii", 0, 0, 0, 0, 0, 0)
+        if self.geo.version >= 7:
+            self.header_data += struct.pack("<iii", *self.pack_reductions)
+            pass
+        if self.geo.version >= 8:
+            self.header_data += struct.pack("<iii", *self.pack_reflection_quads)
+            pass
+        self.header_data += struct.pack("<fff", *self.autolod_dists)
+        self.header_data += struct.pack("<h", self.id)
+
+        self.header_data = struct.pack("<i", len(self.header_data) + 4) + self.header_data
+        pass
+    def encodeHeader(self):
+        self.geo.header_data += self.header_data
+
+    def dump(self):
+        print("    radius: %s" % self.radius)
+        print("    tex_count: %s" % self.tex_count)
+        print("    boneinfo_ptr: %s" % self.boneinfo_ptr)
+        print("    vert_count: %s" % self.vert_count)
+        print("    tri_count: %s" % self.tri_count)
+        print("    name_ptr: %s" % self.name_ptr)
+        print("    api_ptr: %s" % self.api_ptr)
+        print("    grid: %s" % (self.grid_header, ))
+        print("    scale: %s" % self.scale)
+        print("    min: %s" % self.min)
+        print("    max: %s" % self.max)
+        print("    pack_tris: %s" % (self.pack_tris, ))
+        print("    pack_verts: %s" % (self.pack_verts, ))
+        print("    pack_norms: %s" % (self.pack_norms, ))
+        print("    pack_sts: %s" % (self.pack_sts, ))
+        print("    pack_sts3: %s" % (self.pack_sts3, ))
+        print("    pack_weights: %s" % (self.pack_weights, ))
+        print("    pack_matidxs: %s" % (self.pack_matidxs, ))
+        print("    pack_grid: %s" % (self.pack_grid, ))
+        #print("    pack_boneinfo: %s" % (self.pack_boneinfo, ))
+        #print("    pack_name: %s" % (self.pack_name, ))
+        #print("    pack_api: %s" % (self.pack_api, ))
+        print("    autolod_dists: %s" % (self.autolod_dists, ))
+        print("    id: %s" % (self.id, ))
+        #print("    mystery: %s" % (self.mystery, ))
+        print("    skipped_data: %s" % ([self.skipped_data], ))
+
+        print("    name: %s" % self.name)
+        #print("    tris_data: %s" % ([self.tris_data], ))
+        #print("    verts_data: %s" % ([self.verts_data], ))
+        #print("    sts_data: %s" % ([self.sts_data], ))
+        #print("    sts3_data: %s" % ([self.sts3_data], ))
+        #print("    weights_data: %s" % ([self.weights_data], ))
+        #print("    matidxs_data: %s" % ([self.matidxs_data], ))
+        #print("    grid_data: %s" % ([self.grid_data], ))
+        #print("    reductions_data: %s" % ([self.reductions_data], ))
+        print("    reflection_quads_data: %s" % ([self.reflection_quads_data], ))
+
+        print("    tris: %s" % ([self.tris], ))
+        print("    verts: %s" % ([self.verts], ))
+        print("    norms: %s" % ([self.norms], ))
+        print("    weights: %s" % ([self.weights], ))
+        print("    weight_bones (matidxs): %s" % ([self.weight_bones], ))
+        if self.polygrid == None:
+            print("    grid: None")
+        else:
+            print("    grid:")
+            self.polygrid.dump()
+        if self.reductions == None:
+            print("    reductions: None")
+        else:
+            self.reductions.dump()
+            
+        print("    sts: %s" % ([self.sts], ))
+        print("    sts3: %s" % ([self.sts3], ))
+
+        print("    bone_count: %d" % self.bone_count)
+        print("    bone_ids: %s" % (self.bone_ids, ))
+        for i in range(self.bone_count):
+            print("        : %s" % (BONES_LIST[self.bone_ids[i]], ))
+            
+class Geo:
+    def __init__(self):
+        self.data = b""
+        self.offset = 0
+        self.version = -1
+        pass
+    def loadFromData(self, data):
+        self.data = data
+        self.offset = 0
+        self.parseData()
+        pass
+    def loadFromFile(self, filein):
+        self.data = filein.read()
+        self.offset = 0
+        self.parseData()
+        pass
+    def saveToData(self):
+        self.encode()
+        return self.data
+    def saveToFile(self, fileout):
+        self.encode()
+        fileout.write(self.data)
+    def parseData(self):
+        (ziplen, self.header_size) = struct.unpack("<ii", self.data[self.offset : self.offset + 8])
+        if self.header_size != 0:
+            self.version = 0
+            self.offset += 8
+            ziplen -= 4
+        else:
+            (self.version,
+             self.header_size) = struct.unpack("<ii", self.data[self.offset + 8: self.offset + 8 + 8])
+            ziplen -= 12
+            self.offset += 16
+        #print("%s" % [self.data[self.offset : self.offset + ziplen]])
+        self.header_data = zlib.decompress(self.data[self.offset : self.offset + ziplen])#, bufsize = self.header_size)
+        self.offset += ziplen
+        if self.version == 0:
+            self.offset += 4
+
+        self.header_offset = 0
+        (self.main_data_size,
+         self.texname_blocksize,
+         self.objname_blocksize,
+         self.texidx_blocksize) = struct.unpack("<iiii", self.header_data[self.header_offset : self.header_offset + 16])
+        self.header_offset += 16
+        if self.version >= 2 and self.version <= 6:
+            (self.lodinfo_blocksize, ) = struct.unpack("<i", self.header_data[self.header_offset : self.header_offset + 4])
+            self.header_offset += 4
+        else:
+            self.lodinfo_blocksize = 0
+        self.header_texname_data = self.header_data[self.header_offset : self.header_offset + self.texname_blocksize]
+        self.header_offset += self.texname_blocksize
+        self.header_objname_data = self.header_data[self.header_offset : self.header_offset + self.objname_blocksize]
+        self.header_offset += self.objname_blocksize
+        self.header_texidx_data = self.header_data[self.header_offset : self.header_offset + self.texidx_blocksize]
+        self.header_offset += self.texidx_blocksize
+        self.header_lodinfo_data = self.header_data[self.header_offset : self.header_offset + self.lodinfo_blocksize]
+        self.header_offset += self.lodinfo_blocksize
+
+        self.header_modelheader_data = self.header_data[self.header_offset : self.header_offset + 124 + 4 + 4 + 4 + 4]
+        self.header_offset += 124 + 4 + 4 + 4 + 4
+
+        self.header_texnames = unpackNames(self.header_texname_data)
+        self.header_objnames = unpackStrings(self.header_objname_data)
+        self.header_modelheader_name = extractString(self.header_modelheader_data[0:124])
+        (self.header_modelheader_tracklength, ) = struct.unpack("<f", self.header_modelheader_data[124 + 4: 124 + 4 + 4])
+        (self.header_modelheader_modelcount, ) = struct.unpack("<i", self.header_modelheader_data[124 + 4 + 4 + 4: 124 + 4 + 4 + 4 + 4])
+        #print("main offset: %d 0x%x" % (self.offset, self.offset))
+        self.main_data = self.data[self.offset : self.offset + self.main_data_size]
+        #print(" %s" % [self.main_data])
+        self.models = []
+        #self.dump()
+        for i in range(self.header_modelheader_modelcount):
+            self.models.append(Model(self))
+            self.models[-1].parseHeaderData()
+            self.models[-1].parseData()
+            #if i < len(self.header_objnames):
+            #    self.models[-1].name = self.header_objnames[i]
+
+    def getHeaderElement(self, fmt):
+        size = struct.calcsize(fmt)
+        data = struct.unpack(fmt, self.header_data[self.header_offset : self.header_offset + size])
+        self.header_offset += size
+        return data
+    def getDataBlock(self, tup):
+        #print("%s" % (tup, ))
+        (packsize, unpacksize, offset) = tup
+        if packsize == 0:
+            return self.main_data[offset : offset + unpacksize]
+        rawdata = self.main_data[offset : offset + packsize]
+        #print("%s" % [rawdata])
+        try:
+            data = zlib.decompress(rawdata)
+        except Exception as e:
+            print("%s" % (tup, ))
+            print("%s" % [rawdata])
+            traceback.print_exc()
+            #data = rawdata
+            raise e
+        #todo: sanity check size?
+        return data
+
+    def encode(self):
+        self.data = None
+        self.version = 8
+        self.main_data = b""
+        self.header_data = b""
+        self.header_objnames = []
+        self.header_objname_data = b""
+        self.lodinfo_data = b""
+        #Encode models into main data
+        for m in self.models:
+            m.encode()
+        #todo: convert data
+        print("texname unhandled!")
+        print("texidx unhandled!")
+        #Encode information into header data.
+        self.header_data += struct.pack("<iiii", len(self.main_data), len(self.header_texname_data), len(self.header_objname_data), len(self.header_texidx_data))
+        if self.version >= 2 and self.version <= 6:
+            self.header_data += struct.pack("<i", len(self.lodinfo_data))
+        self.header_data += self.header_texname_data + self.header_objname_data + self.header_texidx_data
+        if self.version >= 2 and self.version <= 6:
+            self.header_data += self.lodinfo_data
+        #Encode the main model header.
+        self.header_data += storeString(self.header_modelheader_name, 124)
+        self.header_data += struct.pack("<ifii", 0, self.header_modelheader_tracklength, 0, len(self.models))
+        #Encode model headers into header data.
+        for m in self.models:
+            m.encodeHeader()
+        zheader_data = zlib.compress(self.header_data)
+        if self.version == 0:
+            preheader = struct.pack("<ii", len(zheader_data) + 4, len(self.header_data))
+        else:
+            preheader = struct.pack("<iiii", len(zheader_data) + 12, 0, self.version, len(self.header_data))
+        if self.version == 0:
+            zheader += struct.pack("<i", 0)
+        self.data = preheader + zheader_data + self.main_data
+    def encodeObjName(self, name):
+        o = len(self.header_objname_data)
+        self.header_objname_data += name + ZERO_BYTE
+        self.header_objnames.append(name)
+        return o
+    def encodeMainData(self, data):
+        o = len(self.main_data)
+        self.main_data += data
+        return o
+    def encodeMainDataPacked(self, data):
+        #todo: minimum compression?
+        o= len(self.main_data)
+        d = zlib.compress(data)
+        if len(d) >= len(data):
+            self.main_data += data
+            return (0, len(data), o)
+        else:
+            self.main_data += d
+            return (len(d), len(data), o)
+    def encodeMainDataPackedDeltas(self, data, stride, count, pack_type, float_scale):
+        if data is None:
+            return (0, 0, 0)
+        deltas = compressDeltas(data, stride, count, pack_type, float_scale)
+        pack = self.encodeMainDataPacked(deltas)
+        return pack
+
+    def dump(self):
+        print("version: %d" % self.version)
+        print("header_size: expected: %d  actual: %d" % (self.header_size, len(self.header_data)))
+        #print("header_data: %s" % [self.header_data])
+        print("main_data_size: %d" % self.main_data_size)
+        print("header_data sizes: texname: %d  objname: %d  texidx: %d  lodinfo: %d" % (self.texname_blocksize, self.objname_blocksize, self.texidx_blocksize, self.lodinfo_blocksize))
+        print("header_texname_data: %s" % [self.header_texname_data])
+        print("header_texnames: %s" % [self.header_texnames])
+        print("header_objname_data: %s" % [self.header_objname_data])
+        print("header_objnames: %s" % [self.header_objnames])
+        print("header_texidx_data: %s" % [self.header_texidx_data])
+        print("header_lodinfo_data: %s" % [self.header_lodinfo_data])
+        print("header_modelheader_name: %s" % [self.header_modelheader_name])
+        print("header_modelheader_tracklength: %s" % self.header_modelheader_tracklength)
+        print("header_modelheader_modelcount: %s" % self.header_modelheader_modelcount)
+        print("header_modelheader_data: %s" % [self.header_modelheader_data[124:]])
+        print("header remaining: %d" % ((len(self.header_data) - self.header_offset), ))
+        print("header remaining: %s" %  ([self.header_data[self.header_offset:]], ))
+
+        #%d  objname: %d  texidx: %d  lodinfo: %d" % (self.texname_blocksize, self.objname_blocksize, self.texidx_blocksize, self.lodinfo_blocksize)
+        for i in range(len(self.models)):
+            print("Model %d:" % (i, ))
+            self.models[i].dump()
+
+    def seekMainData(self, offset):
+        self.main_data_offset = offset
+    def skipMainData(self, skip):
+        if type(skip) is int:
+            self.main_data_offset += skip
+        else:
+            self.main_data_offset += struct.calcsize(skip)
+    def getMainElement(self, fmt):
+        size = struct.calcsize(fmt)
+        data = struct.unpack(fmt, self.main_data[self.main_data_offset : self.main_data_offset + size])
+        self.main_data_offset += size
+        return data
+    def getElement(self, offset, fmt):
+        size = struct.calcsize(fmt)
+        data = struct.unpack(fmt, self.main_data[offset : offset + size])
+        return data
+        
+
+if __name__ == "__main__":
+    if len(sys.argv) <= 1:
+        print("Usage:")
+        print("   %s <file_in> [<file_out>]")
+        print("Test loads a .geo file, dumps its content, and optionally writes its content out.")
+        exit(0)
+    fh = open(sys.argv[1], "rb")
+    print(sys.argv)
+    if len(sys.argv) <= 2:
+        fho = None
+    else:
+        fho = open(sys.argv[2], "wb")
+    geo = Geo()
+    geo.loadFromFile(fh)
+    if fho is not None:
+        #geo.dump()
+        geo.saveToFile(fho)
+    else:
+        geo.dump()
+    #print("%s" % [geo.header_data])
diff --git a/stl_dump.py b/stl_dump.py
new file mode 100755
index 0000000..784aeb7
--- /dev/null
+++ b/stl_dump.py
@@ -0,0 +1,42 @@
+import struct
+import geo
+import sys
+import os
+import os.path
+
+if len(sys.argv) < 2:
+    print("Usage:")
+    print("    %s <file.geo>")
+    print("Dumps all the meshes contained in <file.geo> to <geo_name>/<model_name.stl>.")
+    print("<geo_name> and <model_name> are read from the .geo.")
+
+fh = open(sys.argv[1], "rb")
+g = geo.Geo()
+g.loadFromFile(fh)
+
+geo_name = g.header_modelheader_name.decode("utf-8")
+if not os.path.exists(geo_name):
+    os.mkdir(geo_name)
+
+for i in range(len(g.models)):
+    model = g.models[i]
+    model_name = model.name.decode("utf-8")
+    filename = geo_name + "/" + model_name + ".stl"
+    print(filename)
+    ofh = open(filename, "wb")
+    ofh.write(b"\x00" * 80)
+    ofh.write(struct.pack("<i", model.tri_count))
+    for i in range(model.tri_count):
+        tri_index = model.tris[i]
+        tri_verts = (model.verts[tri_index[0]],
+             model.verts[tri_index[1]],
+             model.verts[tri_index[2]])
+        t = b""
+        #print "%5d: %s" % (i, tri_verts)
+        t += struct.pack("<fff", 0, 0, 0)
+        for v in tri_verts:
+            t += struct.pack("<fff", v[0], v[1], v[2])
+        t += b"\x00\x00"
+        ofh.write(t)
+    ofh.close()
+