OOGL format

From the GeomView manual.
Edited by Paul Bourke
October 1996

Support libraries: offlib.c, offlib.h, and simplex.h.

The objects that you can load into Geomview are called OOGL objects. OOGL stands for "Object Oriented Graphics Library"; it is the library upon which Geomview is built.

There are many different kinds of OOGL objects. This chapter gives syntactic descriptions of file formats for OOGL objects.


Syntax Common to All OOGL File Formats

Most OOGL object file formats are free-format ASCII -- any amount of white space (blanks, tabs, newlines) may appear between tokens (numbers, key words, etc.). Line breaks are almost always insignificant, with a couple of exceptions as noted. Comments begin with # and continue to the end of the line; they're allowed anywhere a newline is.

Binary formats are also defined for several objects; See section on binary format, and the individual object descriptions.

Typical OOGL objects begin with a key word designating object type, possibly with modifiers indicating presence of color information etc. In some formats the key word is optional, for compatibility with file formats defined elsewhere. Object type is then determined by guessing from the file suffix (if any) or from the data itself.

Key words are case sensitive. Some have optional prefix letters indicating presence of color or other data; in this case the order of prefixes is significant, e.g. CNMESH is meaningful but NCMESH is invalid.

File Names

When OOGL objects are read from disk files, the OOGL library uses the file suffix to guess at the file type.

If the suffix is unrecognized, or if no suffix is available (e.g. for an object being read from a pipe, or embedded in another OOGL object), all known types of objects are tried in turn until one accepts the data as valid.


Several objects share a common style of representing vertices with optional per-vertex surface-normal and color. All vertices within an object have the same format, specified by the header key word.

All data for a vertex is grouped together (as opposed to e.g. giving coordinates for all vertices, then colors for all vertices, and so on).

The syntax is

`x y z'
(3-D floating-point vertex coordinates) or
`x y z w'
(4-D floating-point vertex coordinates)

optionally followed by

`nx ny nz'
(normalized 3-D surface-normal if present)

optionally followed by

`r g b a'
(4-component floating-point color if present, each component in range 0..1. The a (alpha) component represents opacity: 0 transparent, 1 opaque.)

optionally followed by

`s t'
`s t u'

(two or three texture-coordinate values).

Values are separated by white space, and line breaks are immaterial.

Letters in the object's header key word must appear in a specific order; that's the reverse of the order in which the data is given for each vertex. So a `CN4OFF' object's vertices contain first the 4-component space position, then the 3-component normal, finally the 4-component color. You can't change the data order by changing the header key word; an `NCOFF' is just not recognized.

Surface normal directions

Geomview uses normal vectors to determine how an object is shaded. The direction of the normal is significant in this calculation.

When normals are supplied with an object, the direction of the normal is determined by the data given.

When normals are not supplied with the object, Geomview computes normal vectors automatically; in this case normals point toward the side from which the vertices appear in counterclockwise order.

On parametric surfaces (Bezier patches), the normal at point P(u,v) is in the direction dP/du cross dP/dv.

Transformation matrices

Some objects incorporate 4x4 real matrices for homogeneous object transformations. These matrices act by multiplication on the right of vectors. Thus, if p is a 4-element row vector representing homogeneous coordinates of a point in the OOGL object, and A is the 4x4 matrix, then the transformed point is p' = p A. This matrix convention is common in computer graphics; it's the transpose of that often used in mathematics, where points are column vectors multiplied on the right of matrices.

Thus for Euclidean transformations, the translation components appear in the fourth row (last four elements) of A. A's last column (4th, 8th, 12th and 16th elements) are typically 0, 0, 0, and 1 respectively.

Binary format

Many OOGL objects accept binary as well as ASCII file formats. These files begin with the usual ASCII token (e.g. CQUAD) followed by the word BINARY. Binary data begins at the byte following the first newline after BINARY. White space and a single comment may intervene, e.g.

OFF BINARY	# binary-format "OFF" data follows 

Binary data comprise 32-bit integers and 32-bit IEEE-format floats, both in big-endian format (i.e., with most significant byte first). This is the native format for 'int's and 'float's on Sun-3's, Sun-4's, and Irises, among others.

Binary data formats resemble the corresponding ASCII formats, with ints and floats in just the places you'd expect. There are some exceptions though, specifically in the QUAD, OFF and COMMENT file formats. Details are given in the individual file format descriptions. See section on QUAD: collection of quadrilaterals, See section on OFF Files, and See section on COMMENT Objects.

Binary OOGL objects may be freely mixed in ASCII object streams:

... binary data for mesh here ...
{ = QUAD
	1 0 0   0 0 1   0 1 0  0 1 0

Note that ASCII data resumes immediately following the last byte of binary data.

Naturally, it's impossible to embed comments inside a binary-format OOGL object, though comments may appear in the header before the beginning of binary data.

Embedded objects and external-object references

Some object types (LIST, INST) allow references to other OOGL objects, which may appear literally in the data stream, be loaded from named disk files, or be communicated from elsewhere via named objects. Gcl commands also accept geometry in these forms.

The general syntax is

 <oogl-object>  ::=
	[ "{" ]
	    [ "define" symbolname ]
	    [ "appearance" appearance ]
	    [ ["="] object-keyword ...
		 | "<" filename
		 | ":" symbolname ]
	[ "}" ]

where "quoted" items are literal strings (which appear without the quotes), [bracketed] items are optional, and | denotes alternatives. Curly braces, when present, must match; the outermost set of curly braces is generally required when the object is in a larger context, e.g. when it is part of a larger object or embedded in a Geomview command stream.

For example, each of the following three lines:

	{ define fred   QUAD 1 0 0  0 0 1  0 1 0  1 0 0 }

	{ appearance { +edge } LIST { < "file1" } { : fred } }

	VECT 1 2 0   2 0   0 0 0   1 1 2
is a valid OOGL object. The last example is only valid when it is delimited unambiguously by residing in its own disk file.

The "<" construct causes a disk file to be read. Note that this isn't a general textual "include" mechanism; a complete OOGL object must appear in the referenced file.

Files read using "<" are sought first in the directory of the file which referred to them, if any; failing that, the normal search path (set by Geomview's load-path command) is used. The default search looks first in the current directory, then in the Geomview data directories.

The ":" construct allows references to symbols, created with define. A symbol's initial value is a null object. When a symbol is (re)defined, all references to it are automatically changed; this is a crucial part of the support for interprocess communication. Some future version of the documentation should explain this better...

Again, white space and line breaks are insignificant, and "#" comments may appear anywhere.


Geometric objects can have associated "appearance" information, specifying shading, lighting, color, wireframe vs. shaded-surface display, and so on. Appearances are inherited through object hierarchies, e.g. attaching an appearance to a LIST means that the appearance is applied to all the LIST's members.

Some appearance-related properties are relegated to "material" and "lighting" substructures. Take care to note which properties belong to which structure.

Here's an example appearance structure including values for all attributes. Order of attributes is unimportant. As usual, white space is irrelevant. Boolean attributes may be preceded by "+" or "-" to turn them on or off; "+" is assumed if only the attribute name appears. Other attributes expect values.

A "*" prefix on any attribute, e.g. "*+edge" or "*linewidth 2" or "material { *diffuse 1 1 .25 }", selects "override" status for that attribute.

appearance {
  +face               # (Do) draw faces of polygons.  On by default.
  -edge               # (Don't) draw edges of polygons
  +vect               # (Do) draw VECTs.  On by default.
  -transparent        # (Disable) transparency. enabling transparency 
                      # does NOT result in a correct Geomview picture, 
                      # but alpha values are used in RenderMan snapshots.
  -normal             # (Do) draw surface-normal vectors
  normscale 1         # ... with length 1.0 in object coordinates

  +evert              # do evert polygon normals where needed so as
                      #   to always face the camera

  -texturing          # (Disable) texture mapping
  -backcull           # (Don't) discard clockwise-oriented faces
  -concave            # (Don't) expect and handle concave polygons
  -shadelines	      # (Don't) shade lines as if they were lighted cylinders
		      # These four are only effective where the graphics system
		      # supports them, namely on GL and Open GL.

  -keepcolor	      # Normally, when N-D positional coloring is enabled as
		      # with geomview's (ND-color ...) command, all
		      # objects' colors are affected.  But, objects with the
		      # "+keepcolor" attribute are immune to N-D coloring.

  shading smooth      # or "shading constant" or "shading flat" or
                      # or "shading csmooth".
                      # smooth = Gouraud shading, flat = faceted,
                      # csmooth = smoothly interpolated but unlighted.

  linewidth 1         # lines, points, and edges are 1 pixel wide.

  patchdice 10 10     # subdivide Bezier patches this finely in u and v

  material {         # Here's a material definition;
                      # it could also be read from a file as in
                      #  "material < file.mat"

      ka  1.0         # ambient reflection coefficient.
      ambient .3 .5 .3 # ambient color (red, green, blue components)
                      # The ambient contribution to the shading is
                      # the product of ka, the ambient color,
                      # and the color of the ambient light.

      kd  0.8         # diffuse-reflection coefficient.
      diffuse .9 1 .4 # diffuse color.
                        # (In "shading constant" mode, the surface
                        # is colored with the diffuse color.)

      ks 1.0          # specular reflection coefficient.
      specular 1 1 1  # specular (highlight) color.
      shininess  25   # specular exponent; larger values give
                      # sharper highlights.

      backdiffuse .7 .5 0 # back-face color for two-sided surfaces
                        # If defined, this field determines the diffuse
                        # color for the back side of a surface.
                        # It's implemented by the software shader, and
                        # by hardware shading on GL systems which support
                        # two-sided lighting, and under Open GL.

      alpha   1.0     # opacity; 0 = transparent (invisible), 1 = opaque.
                      # Ignored when transparency is disabled.

      edgecolor   1 1 0  # line & edge color

      normalcolor 0 0 0  # color for surface-normal vectors

  lighting {         # Lighting model

      ambient  .3 .3 .3  # ambient light

      replacelights   # "Use only the following lights to
                      # illuminate the objects under this
                      # appearance."
                      # Without "replacelights", any lights listed
                      # are added to those already in the scene.

                      # Now a collection of sample lights:
      light { 
          color  1 .7 .6      # light color
          position  1 0 .5 0  # light position [distant light]
                              # given in homogeneous coordinates.
                              # With fourth component = 0,
                              # this means a light coming from
                              # direction (1,0,.5).

      light {                        # Another light.
          color 1 1 1
          position  0 0 .5 1  # light at finite position ...
          location camera     # specified in camera coordinates.
                              # (Since the camera looks toward -Z,
                              # this example places the light
                              # .5 unit behind the eye.)
          # Possible "location" keywords:
          #  global    light position is in world coordinates
          #             This is the default if no location specified.
          #  camera   position is in the camera's coordinate system
          #  local    position is in the coordinate system where
          #                   the appearance was defined
  }                   # end lighting model
  texture {
        clamp st               # or "s" or "t" or "none"
        file lump.tiff         # file supplying texture-map image
        alphafile mask.pgm.Z   # file supplying transparency-mask image
        apply blend            # or "modulate" or "decal"
        transform  1 0 0 0     # surface (s,t,0,1) * tfm -> texture coords
                   0 1 0 0
                   0 0 1 0
                  .5 0 0 1

        background 1 0 0 1     # relevant for "apply blend"
}                     # end appearance

There are rules for inheritance of appearance attributes when several are imposed at different levels in the hierarchy.

For example, Geomview installs a backstop appearance which provides default values for most parameters; its control panels install other appearances which supply new values for a few attributes; user-supplied geometry may also contain appearances.

The general rule is that the child's appearance (the one closest to the geometric primitives) wins. Further, appearance controls with "override" status (e.g. *+face or material { *diffuse 1 1 0 }) win over those without it.

Geomview's appearance controls use the "override" feature so as to be effective even if user-supplied objects contain their own appearance settings. However, if a user-supplied object contains an appearance field with override status set, that property will be immune to Geomview's controls.

Texture Mapping

Some platforms support texture-mapped objects. (On those which don't, attempts to use texture mapping are silently ignored.) A texture is specified as part of an appearance structure, as in See section on appearances. Briefly, one provides a texture image, which is considered to lie in a square in (s,t) parameter space in the range 0 <= s <= 1, 0 <= t <= 1. Then one provides a geometric primitive, with each vertex tagged with (s,t) texture coordinates. If texturing is enabled, the appropriate portion of the texture image is pasted onto each face of the textured object.

There is (currently) no provision for inheritance of part of a texture structure; if the texture keyword is mentioned in an appearance, it supplants any other texture specification.

The appearance attribute texturing controls whether textures are used; there's no performance penalty for having texture { ... } fields defined when texturing is off.

The available fields are:

clamp	none  -or-  s  -or-  t  -or-  st
  Determines the meaning of texture coordinates outside the range 0..1.
  With clamp none, the default, coordinates are interpreted
  modulo 1, so (s,t) = (1.25,0), (.25,0), and (-.75,0) all refer to
  the same point in texture space.  With s or t or
  st, either or both of s- or t-coordinates less than 0 or
  greater than 1 are clamped to 1 or 0, respectively.

file	filename
alphafile	filename
  Specifies image file(s) containing the texture.
  The file file's image specifies color or lightness information;
  the alphafile if present, specifies a transparency ("alpha") mask;
  where the mask is zero, pixels are simply not drawn.
  Several image file formats are available; the file type must be
  indicated by the last few characters of the file name:
    .ppm or .ppm.Z or .ppm.gz  24-bit 3-color image in PPM format
    .pgm or .pgm.Z or .pgm.gz  8-bit grayscale image in PGM format
    .tiff 		       8-bit or 24-bit TIFF image
    .gif		       GIF image
    .sgi		       8-bit or 24-bit SGI image
  (Though 4-channel TIFF and SGI images are possible, and could
  represent both color and transparency information in one image,
  that's not supported in geomview at present.)
  For this feature to work, some programs must be available in
  geomview's search path:
    zcat  for .Z files
    gzip  for .gz files
    tifftopnm for .tiff files
    giftoppm for .gif files
    convert (from the ImageMagick package) for .sgi files

  If an alphafile image is supplied, it must be the same size
  as the file image.

apply	modulate  -or-  blend  -or-  decal
  Indicates how the texture image is applied to the surface.
  Here the "surface color" means the color that surface would have
  in the absence of texture mapping.

  With modulate, the default, the texture color (or lightness,
  if textured by a gray-scale image) is multiplied by the surface color.

  With blend, texture blends between the background color
  and the surface color.  The file parameter must specify a
  gray-scale image.  Where the texture image is 0, the surface color is
  unaffected; where it's 1, the surface is painted in the color given
  by background; and color is interpolated for intermediate values.

  With decal, the file parameter must specify a
  3-color image.  If an alphafile parameter is present,
  its value interpolates between the surface color (where alpha=0)
  and the texture color (where alpha=1).  Lighting does not affect the
  texture color in decal mode; effectively the texture is

background  R G B A
  Specifies a 4-component color, with R, G, B, and A floating-point
  numbers normally in the range 0..1, used when apply blend
  is selected.

transform transformation-matrix
  Expects a list of 16 numbers, or one of the other ways of representing
  a transformation (: handlename or < filename).
  The 4x4 transformation matrix is applied to texture coordinates,
  in the sense of a 4-component row vector (s,t,0,1) multiplied on
  the left of the matrix, to produce new coordinates (s',t')
  which actually index the texture.

Object File Formats

QUAD: collection of quadrilaterals

The conventional suffix for a QUAD file is `.quad'.

The file syntax is

   [C][N][4]QUAD  -or-  [C][N][4]POLY		   # Key word
   vertex  vertex  vertex  vertex  # 4*N vertices for some N
   vertex  vertex  vertex  vertex

The leading key word is [C][N][4]QUAD or [C][N][4]POLY, where the optional C and N prefixes indicate that each vertex includes colors and normals respectively. That is, these files begin with one of the words


(but not NCQUAD or NCPOLY). QUAD and POLY are synonymous; both forms are allowed just for compatibility with ChapReyes.

Following the key word is an arbitrary number of groups of four vertices, each group describing a quadrilateral. See the Vertex syntax above. The object ends at end-of-file, or with a closebrace if incorporated into an object reference (see above).

A QUAD BINARY file format is accepted; See section on the binary format. The first word of binary data must be a 32-bit integer giving the number of quads in the object; following that is a series of 32-bit floats, arranged just as in the ASCII format.

MESH: rectangularly-connected mesh

The conventional suffix for a MESH file is `.mesh'.

The file syntax is

[U][C][N][Z][4][u][v][n]MESH # Key word
[Ndim]                 # Space dimension, present only if nMESH
Nu Nv            # Mesh grid dimensions
                             # Nu*Nv vertices, in format specified
                             # by initial key word
vertex(u=0,v=0)  vertex(1,0)  ... vertex(Nu-1,0)
vertex(0,1) ...    vertex(Nu-1,1)
vertex(0,Nv-1) ... vertex(Nu-1,Nv-1)

The key word is [U][C][N][Z][4][u][v][n]MESH. The optional prefix characters mean:

Each vertex includes a 3-component texture space parameter. The first two components are the usual S and T texture parameters for that vertex; the third should be specified as zero.
Each vertex (see Vertices above) includes a 4-component color.
Each vertex includes a surface normal vector.
Of the 3 vertex position values, only the Z component is present; X and Y are omitted, and assumed to equal the mesh (u,v) coordinate so X ranges from 0 .. (Nu-1), Y from 0 .. (Nv-1) where Nu and Nv are the mesh dimensions -- see below.
Vertices are 4D, each consists of 4 floating values. Z and 4 cannot both be present.
The mesh is wrapped in the u-direction, so the (0,v)'th vertex is connected to the (Nu-1,v)'th for all v.
The mesh is wrapped in the v-direction, so the (u,0)'th vertex is connected to the (u,Nv-1)'th for all u. Thus a u-wrapped or v-wrapped mesh is topologically a cylinder, while a uv-wrapped mesh is a torus.
Specifies a mesh whose vertices live in a higher dimensional space. The dimension follows the "MESH" keyword. Each vertex then has Ndim components.

Note that the order of prefix characters is significant; a coloured, u-wrapped mesh is a CuMESH not a uCMESH.

Following the mesh header are integers Nu and Nv, the dimensions of the mesh.

Then follow Nu*Nv vertices, each in the form given by the header. They appear in v-major order, i.e. if we name each vertex by (u,v) then the vertices appear in the order

(0,0) (1,0) (2,0) (3,0) ...  (Nu-1,0)
(0,1) (1,1) (2,1) (3,1) ...  (Nu-1,1)
(0,Nv-1)		...  (Nu-1,Nv-1)

A MESH BINARY format is accepted; See section on the binary format. The values of Nu and Nv are 32-bit integers; all other values are 32-bit floats.

Bezier Surfaces

The conventional file suffixes for Bezier surface files are `.bbp' or `.bez'. A file with either suffix may contain either type of patch.


  [ST]BBP -or- [C]BEZ<Nu><Nv><Nd>[_ST]
			# Nu, Nv are u- and v-direction 
			# polynomial degrees in range 1..6
			# Nd = dimension: 3->3-D, 4->4-D (rational)
			# (The '<' and '>' do not appear in the input.)
			# Nu,Nv,Nd are each a single decimal digit.
			# BBP form implies Nu=Nv=Nd=3 so BBP = BEZ333.

		# Any number of patches follow the header
			# (Nu+1)*(Nv+1) patch control points
			# each 3 or 4 floats according to header
  vertex(u=0,v=0)  vertex(1,0) ... vertex(Nu,0)
  vertex(0,1)			   ... vertex(Nu,1)
  vertex(0,Nv)		   ... vertex(Nu,Nv)

			# ST texture coordinates if mentioned in header
  S(u=0,v=0)	T(0,0)	S(0,Nv) T(0,Nv)
  S(Nu,0)	T(Nu,0)	S(Nu,Nv) T(Nu,Nv)

			# 4-component float (0..1) R G B A colors
			# for each patch corner if mentioned in header
  RGBA(0,0)   RGBA(0,Nv)
  RGBA(Nu,0)  RGBA(Nu,Nv)

These formats represent collections of Bezier surface patches, of degrees up to 6, and with 3-D or 4-D (rational) vertices.

The header keyword has the forms [ST]BBP or [C]BEZ<Nu><Nv><Nd>[_ST] (the '<' and '>' are not part of the keyword.

The ST prefix on BBP, or _ST suffix on BEZuvn, indicates that each patch includes four pairs of floating-point texture-space coordinates, one for each corner of the patch.

The C prefix on BEZuvn indicates a coloured patch, including four sets of four-component floating-point colors (red, green, blue, and alpha) in the range 0..1, one color for each corner.

Nu and Nv, each a single digit in the range 1..6, are the patch's polynomial degree in the u and v direction respectively.

Nd is the number of components in each patch vertex, and must be either 3 for 3-D or 4 for homogeneous coordinates, that is, rational patches.

BBP patches are bicubic patches with 3-D vertices, so BBP = BEZ333 and STBBP = BEZ333_ST.

Any number of patches follow the header. Each patch comprises a series of patch vertices, followed by optional (s,t) texture coordinates, followed by optional (r,g,b,a) colors.

Each patch has (Nu+1)*(Nv+1) vertices in v-major order, so that if we designate a vertex by its control point indices (u,v) the order is

     (0,0) (1,0) (2,0) ...  (Nu,0)
     (0,1) (1,1) (2,1) ...  (Nu,1)
     (0,Nv)            ...  (Nu,Nv)
with each vertex containing either 3 or 4 floating-point numbers as specified by the header. If the header calls for ST coordinates, four pairs of floating-point numbers follow: the texture-space coordinates for the (0,0), (Nu,0), (0,Nv), and (Nu,Nv) corners of the patch, respectively.

If the header calls for colors, four four-component (red, green, blue, alpha) floating-point colors follow, one for each patch corner.

The series of patches ends at end-of-file, or with a closebrace if incorporated in an object reference.

OFF Files

The conventional suffix for OFF files is `.off'.


[ST][C][N][4][n]OFF	# Header keyword
[Ndim]		# Space dimension of vertices, present only if nOFF
NVertices  NFaces  NEdges   # NEdges not used or checked

x[0]  y[0]  z[0]	# Vertices, possibly with normals,
			# colors, and/or texture coordinates, in that order,
			# if the prefixes N, C, ST
			# are present.
			# If 4OFF, each vertex has 4 components,
			# including a final homogeneous component.
			# If nOFF, each vertex has Ndim components.
			# If 4nOFF, each vertex has Ndim+1 components.
x[NVertices-1]  y[NVertices-1]  z[NVertices-1]

    			# Faces
    			# Nv = # vertices on this face
    			# v[0] ... v[Nv-1]: vertex indices
    			#		in range 0..NVertices-1
Nv  v[0] v[1] ... v[Nv-1]  colorspec
    			# colorspec continues past v[Nv-1]
    			# to end-of-line; may be 0 to 4 numbers
    			# nothing: default
    			# integer: colormap index
    			# 3 or 4 integers: RGB[A] values 0..255
			# 3 or 4 floats: RGB[A] values 0..1

OFF files (name for "object file format") represent collections of planar polygons with possibly shared vertices, a convenient way to describe polyhedra. The polygons may be concave but there's no provision for polygons containing holes.

An OFF file may begin with the keyword OFF; it's recommended but optional, as many existing files lack this keyword.

Three ASCII integers follow: NVertices, NFaces, and NEdges. These are the number of vertices, faces, and edges, respectively. Current software does not use nor check NEdges; it needn't be correct but must be present.

The vertex coordinates follow: dimension * Nvertices floating-point values. They're implicitly numbered 0 through NVertices-1. dimension is either 3 (default) or 4 (specified by the key character 4 directly before OFF in the keyword).

Following these are the face descriptions, typically written with one line per face. Each has the form

N  Vert1 Vert2 ... VertN  [color]
Here N is the number of vertices on this face, and Vert1 through VertN are indices into the list of vertices (in the range 0..NVertices-1).

The optional color may take several forms. Line breaks are significant here: the color description begins after VertN and ends with the end of the line (or the next # comment). A color may be:

the default color
one integer
index into "the" colormap; see below
three or four integers
RGB and possibly alpha values in the range 0..255
three or four floating-point numbers
RGB and possibly alpha values in the range 0..1

For the one-integer case, the colormap is currently read from the file `cmap.fmap' in Geomview's `data' directory. Some better mechanism for supplying a colormap is likely someday.

The meaning of "default color" varies. If no face of the object has a color, all inherit the environment's default material color. If some but not all faces have colors, the default is gray (R,G,B,A=.666).

A [ST][C][N][n]OFF BINARY format is accepted; See section on the binary format. It resembles the ASCII format in almost the way you'd expect, with 32-bit integers for all counters and vertex indices and 32-bit floats for vertex positions (and texture coordinates or vertex colors or normals if COFF/NOFF/CNOFF/STCNOFF/etc. format).

Exception: each face's vertex indices are followed by an integer indicating how many color components accompany it. Face color components must be floats, not integer values. Thus a colourless triangular face might be represented as

int int int int int
3   17   5   9   0

while the same face coloured red might be

int int int int int float float float float
 3  17   5   9   4   1.0   0.0   0.0   1.0

VECT Files

The conventional suffix for VECT files is `.vect'.


NPolylines  NVertices  NColors

Nv[0] ... Nv[NPolylines-1]     # number of vertices
                                           # in each polyline

Nc[0] ... Nc[NPolylines-1]     # number of colors supplied
                                           # in each polyline

Vert[0] ... Vert[NVertices-1]  # All the vertices
                                           # (3*NVertices floats)

Color[0] ... Color[NColors-1]  # All the colors
                                           # (4*NColors floats, RGBA)

VECT objects represent lists of polylines (strings of connected line segments, possibly closed). A degenerate polyline can be used to represent a point.

A VECT file begins with the key word VECT or 4VECT and three integers: NLines, NVertices, and NColors. Here NLines is the number of polylines in the file, NVertices the total number of vertices, and NColors the number of colors as explained below.

Next come NLines integers

Nv[0] Nv[1] Nv[2] ... Nv[NLines-1]

giving the number of vertices in each polyline. A negative number indicates a closed polyline; 1 denotes a single-pixel point. The sum (of absolute values) of the Nv[i] must equal NVertices.

Next come NLines more integers Nc[i]: the number of colors in each polyline. Normally one of three values:

No color is specified for this polyline. It's drawn in the same color as the previous polyline.
A single color is specified. The entire polyline is drawn in that color.
Each vertex has a color. Either each segment is drawn in the corresponding color, or the colors are smoothly interpolated along the line segments, depending on the implementation.

The sum of the Nc[i] must equal NColors.

Next come NVertices groups of 3 or 4 floating-point numbers: the coordinates of all the vertices. If the keyword is 4VECT then there are 4 values per vertex. The first abs(Nv[0]) of them form the first polyline, the next abs(Nv[1]) form the second and so on.

Finally NColors groups of 4 floating-point numbers give red, green, blue and alpha (opacity) values. The first Nc[0] of them apply to the first polyline, and so on.

A VECT BINARY format is accepted; See section on the binary format. The binary format exactly follows the ASCII format, with 32-bit ints where integers appear, and 32-bit floats where real values appear.

SKEL Files

SKEL files represent collections of points and polylines, with shared vertices. The conventional suffix for SKEL files is `.skel'.


[NDim]                    # Vertex dimension, present only if nSKEL
NVertices  NPolylines

x[0]  y[0]  z[0]      # Vertices
				    # (if nSKEL, each vertex has NDim components)
x[NVertices-1]  y[NVertices-1]  z[NVertices-1]

                        # Polylines
                        # Nv = # vertices on this polyline (1 = point)
                        # v[0] ... v[Nv-1]: vertex indices                        #               in range 0..NVertices-1
Nv  v[0] v[1] ... v[Nv-1]  [colorspec]
                        # colorspec continues past v[Nv-1]
                        # to end-of-line; may be nothing, or 3 or 4 numbers.
                        # nothing: default color
			# 3 or 4 floats: RGB[A] values 0..1

The syntax resembles that of OFF files, with a table of vertices followed by a sequence of polyline descriptions, each referring to vertices by index in the table. Each polyline has an optional color.

For nSKEL objects, each vertex has NDim components. For 4nSKEL objects, each vertex has NDim+1 components; the final component is the homogeneous divisor.

No BINARY format is implemented as yet for SKEL objects.


The conventional suffix for SPHERE files is `.sph'.

Xcenter Ycenter Zcenter

Sphere objects are drawn using rational Bezier patches, which are diced into meshes; their smoothness, and the time taken to draw them, depends on the setting of the dicing level, 10x10 by default. From Geomview, the (dice N) GCL command or <N>ad keyboard command sets this; within the OOGL libraries, use GeomDice().

INST Files

The conventional suffix for a INST file is `.inst'.

There is no INST BINARY format.

An INST applies a 4x4 transformation to another OOGL object. It begins with INST followed by these sections which may appear in any order:

geom oogl-object
specifies the OOGL object to be instantiated. See section on embedded objects and external-object references, for the syntax of an oogl-object. The keyword unit is a synonym for geom.
transform   ["{"] 4x4 transform ["}"]
specifies a single transformation matrix. Either the matrix may appear literally as 16 numbers, or there may be a reference to a "transform" object, i.e.
"<" file-containing-4x4-matrix
":" symbol-representing-"transform"-object>

Another way to specify the transformation is


The oogl-object must be a TLIST object (list of transformations) object, or a LIST whose members are ultimately TLIST objects. In effect, the transforms keyword takes a collection of 4x4 matrices and replicates the geom object, making one copy for each 4x4 matrix.

If no transform nor transforms keyword appears, no transformation is applied (actually the identity is applied). You could use this for, e.g., wrapping an appearance around an externally-supplied object, though a single-membered LIST would do this more efficiently.

See section on transformation matrices, for the matrix format.

Two more INST fields are accepted: location and origin.

location [global or camera or ndc or screen or local]
Normally an INST specifies a position relative to its parent object; the location field allows putting an object elsewhere.
  • location global attaches the object to the global coordinate system -- the same as that in which geomview's World objects, alien geometry, and cameras are placed.
  • location camera places the object relative to the camera. (Thus if there are multiple views, it may appear in a different spatial position in each view.) The center of the camera's view is along its negative Z axis; positive X is rightward, positive Y upward. Normally the units of camera space are the same as global coordinates. When a camera is reset, the global origin is at (0,0,-3.0).
  • location ndc places the object relative to the normalized unit cube into which the camera's projection (perspective or orthographic) maps the visible world. X, Y, and Z are each in the range from -1 to +1, with Z = -1 the near and Z = +1 the far clipping plane, and X and Y increasing rightward and upward respectively. Thus something like
    INST  transform  1 0 0 0  0 1 0 0  0 0 1 0  -.9 -.9 -.999 1
          location ndc
          geom < label.vect
    pastes label.vect onto the lower left corner of each window, and in front of nearly everything else, assuming label.vect's contents lie in the positive quadrant of the X-Y plane. It's tempting to use -1 rather than -.999 as the Z component of the position, but that may put the object just nearer than the near clipping plane and make it (partially) invisible, due to floating-point error.
  • location screen places the object in screen coordinates. The range of Z is still -1 through +1 as for ndc coordinates; X and Y are measured in pixels, and range from (0,0) at the lower left corner of the window, increasing rightward and upward.

location local is the default; the object is positioned relative to its parent.

origin [global or camera or ndc or screen or local] x y z
The origin field translates the contents of the INST to place the origin at the specified point of the given coordinate system. Unlike location, it doesn't change the orientation, only the choice of origin. Both location and origin can be used together.

So for example

  location screen
  origin ndc 0 0 -.99
  geom { < xyz.vect }
  transform { 100 0 0 0  0 100 0 0  0 0 -.009 0   0 0 0 1 }

places xyz.vect's origin in the center of the window, just beyond the near clipping plane. The unit-length X and Y edges are scaled to be just 100 screen units -- pixels -- long, regardless of the size of the window.

INST Examples

Here are some examples of INST files

     unit < xyz.vect
     transform {
        1 0 0 0
        0 1 0 0
        0 0 1 0
        1 3 0 1

{ appearance { +edge  material { edgecolor 1 1 0 } }
    INST geom < mysurface.quad }

{INST transform {: T} geom {<dodec.off}}

         { LIST
     	{ < some-matrices.prj }
     	{ < others.prj }
     	{ TLIST <still more of them> }
         { # stuff replicated by all the above matrices

This one resembles the origin example in the section above, but makes the X and Y edges be 1/4 the size of the window (1/4, not 1/2, since the range of ndc X and Y coordinates is -1 to +1).

  location ndc
  geom { < xyz.vect }
  transform { .5 0 0 0  0 .5 0 0  0 0 -.009 0   0 0 -.99 1 }

LIST Files

The conventional suffix for a LIST file is `.list'.

A list of OOGL objects



Note that there's no explicit separation between the oogl-objects, so they should be enclosed in curly braces ({ }) for sanity. Likewise there's no explicit marker for the end of the list; unless appearing alone in a disk file, the whole construct should also be wrapped in braces, as in:

   { LIST { QUAD ... } { < xyz.quad } }

A LIST with no elements, i.e. { LIST }, is valid, and is the easiest way to create an empty object. For example, to remove a symbol's definition you might write

   { define somesymbol  { LIST } }


The conventional suffix for a TLIST file is `.grp' ("group") or or `.prj' ("projective" matrices).

Collection of 4x4 matrices, used in the transforms section of and INST object.


TLIST			# key word

<4x4 matrix (16 floats)>
...				# Any number of 4x4 matrices

TLISTs are used only within the transforms clause of an INST object. They cause the INSTs geom object to be instantiated once under each of the transforms in the TLIST. The effect is like that of a LIST of INSTs each with a single transform, and all referring to the same object, but is more efficient.

Be aware that a TLIST is a kind of geometry object, distinct from a transform object. Some contexts expect one type of object, some the other. For example in

INST transform { : myT } geom { ... }
myT must be a transform object, which might have been created with the gcl
(read transform { define myT 1 0 0 1 ... })
while in
    INST transforms { : myTs } geom { ... }
or  INST transforms { LIST {: myTs} {< more.prj} } geom { ... }
myTs must be a geometry object, defined e.g. with
    (read geometry { define myTs { TLIST 1 0 0 1 ... } })

A TLIST BINARY format is accepted. Binary data begins with a 32-bit integer giving the number of transformations, followed by that number of 4x4 matrices in 32-bit floating-point format. The order of matrix elements is the same as in the ASCII format.


This format is obsolete, but is still accepted. It combined the functions of INST and TLIST, taking a series of transformations and a single Geom (unit) object, and replicating the object under each transformation.

GROUP ... < matrices > ... unit { oogl-object }

is still accepted and effectively translated into

	transforms { TLIST ... <matrices> ... }
	unit { oogl-object }


This format is for discrete groups, such as appear in the theory of manifolds or in symmetry patterns. This format has its own man page. See discgrp(5).


The COMMENT object is a mechanism for encoding arbitrary data within an OOGL object. It can be used to keep track of data or pass data back and forth between external modules.


COMMENT                 # key word
name type   # individual name and type specifier
{ ... }             # arbitrary data

The data, which must be enclosed by curly braces, can include anything except unbalanced curly braces. The type field can be used to identify data of interest to a particular program through naming conventions.

COMMENT objects are intended to be associated with other objects through inclusion in a LIST object. (See section on LIST Files) The "#" OOGL comment syntax does not suffice for data exchange since these comments are stripped when an OOGL object is read in to Geomview. The COMMENT object is preserved when loaded into Geomview and is written out intact.

Here is an example associating a WorldWide Web URL with a piece of geometry:

 { < Tetrahedron} 
 {COMMENT GCHomepage HREF { http://www.geom.umn.edu/ }}

A binary COMMENT format is accepted. Its format is not consistent with the other OOGL binary formats. See section on the binary format. The name and type are followed by

N Byte1 Byte2 ... ByteN
instead of data enclosed in curly braces.

Non-geometric objects

The syntax of these objects is given in the form used in See section on Embedded objects and external-object references, where "quoted" items should appear literally but without quotes, square bracketed ([ ]) items are optional, and | separates alternative choices.

Transform Objects

Where a single 4x4 matrix is expected -- as in the INST transform field, the camera's camtoworld transform and the Geomview xform* commands -- use a transform object.

Note that a transform is distinct from a TLIST, which is a type of geometry. TLISTs can contain one or more 4x4 transformations; "transform" objects must have exactly one.

Why have both? In many places -- e.g. camera positioning -- it's only meaningful to have a single transform. Using a separate object type enforces this.

Syntax for a transform object is

<transform> ::= 
  [ "{" ]             (curly brace, generally needed to make
                       the end of the object unambiguous.)

   [ "transform" ]    (optional keyword; unnecessary if the type
                       is determined by the context, which it
                       usually is.)
   [ "define" <name> ]
                      (defines a transform named <name>, setting
                       its value from the stuff which follows)

      <sixteen floating-point numbers>
                      (interpreted as a 4x4 homogeneous transform
		       given row by row, intended to apply to a
                       row vector multiplied on its LEFT, so that e.g.
                       Euclidean translations appear in the bottom row)
      "<" <filename>  (meaning: read transform from that file)
      ":" <name>      (meaning: use variable <name>,
                        defined elsewhere; if undefined the initial
                        value is the identity transform)

 [ "}" ]              (matching curly brace)

The whole should be enclosed in { braces }. Braces are not essential if exactly one of the above items is present, so e.g. a 4x4 array of floats standing alone may but needn't have braces.

Some examples, in contexts where they might be used:

# Example 1: A gcl command to define a transform
# called "fred"

(read transform { transform  define fred
         1 0 0 0
         0 1 0 0
         0 0 1 0
        -3 0 1 1

# Example 2:  A camera object using transform
# "fred" for camera positioning
# Given the definition above, this puts the camera at
# (-3, 0, 1), looking toward -Z.

{ camera
        halfyfield 1
        aspect 1.33
        camtoworld { : fred }


A camera object specifies the following properties of a camera:

position and orientation
specified by either a camera-to-world or world-to-camera transformation; this transformation does not include the projection, so it's typically just a combination of translation and rotation. Specified as a transform object, typically a 4x4 matrix.
"focus" distance
Intended to suggest a typical distance from the camera to the object of interest; used for default camera positioning (the camera is placed at (X,Y,Z) = (0,0,focus) when reset) and for adjusting field-of-view when switching between perspective and orthographic views.
window aspect ratio
True aspect ratio in the sense <Xsize>/<Ysize>. This normally should agree with the aspect ratio of the camera's window. Geomview normally adjusts the aspect ratio of its cameras to match their associated windows.
near and far clipping plane distances
Note that both must be strictly greater than zero. Very large <far>/<near> distance ratios cause Z-buffering to behave badly; part of an object may be visible even if somewhat more distant than another.
field of view
Specified in either of two forms.
@item fov is the field of view -- in degrees if perspective, or linear distance if orthographic -- in the shorter direction. @item halfyfield is half the projected Y-axis field, in world coordinates (not angle!), at unit distance from the camera. For a perspective camera, halfyfield is related to angular field:

halfyfield = tan( Y_axis_angular_field / 2 )

while for an orthographic one it's simply:

halfyfield = Y_axis_linear_field / 2

This odd-seeming definition is (a) easy to calculate with and (b) well-defined in both orthographic and perspective views.

The syntax for a camera is:

<camera> ::=

   [ "camera" ]			(optional keyword)
    [ "{" ]			(opening brace, generally required)
	[ "define" <name> ]

	"<" <filename>
	":" <name>
				(or any number of the following,
				 in any order...)

	"perspective"  {"0" | "1"}		(default 1)
					(otherwise orthographic)

	"stereo"       {"0" | "1"}		(default 0)
					(otherwise mono)

	"worldtocam" <transform>	(see transform syntax above)

	"camtoworld" <transform>
				(no point in specifying both
				 camtoworld and worldtocam; one is
				 constrained to be the inverse of						 the other)

	"halfyfield" <half-linear-Y-field-at-unit-distance>
				(default tan 40/2 degrees)

	"fov"		(angular field-of-view if perspective,
			 linear field-of-view otherwise.
			 Measured in whichever direction is smaller,
			 given the aspect ratio.  When aspect ratio
			 changes -- e.g. when a window is reshaped --
			 "fov" is preserved.)

	"frameaspect" <aspect-ratio>	(X/Y) (default 1.333)

	"near"  <near-clipping-distance>	(default 0.1)
	"far"	<far-clipping-distance>		(default 10.0)

	"focus" <focus-distance>		(default 3.0)

     [ "}" ]				(matching closebrace)


A window object specifies size, position, and other window-system related information about a window in a device-independent way.

The syntax for a window object is:

window ::=

	[ "window" ]			(optional keyword)
	  [ "{" ]			(curly brace, often required)

	    			(any of the following, in any order)

		"size"  <xsize> <ysize>
				(size of the window)

		"position"  <xmin> <xmax> <ymin> <ymax>
				(position & size)

				(specifies the window should
				 have no window border)

		"pixelaspect"  <aspect>
			    (specifies the true visual aspect ratio
			     of a pixel in this window in the sense
			     xsize/ysize, normally 1.0.
			     For stereo hardware which stretches the
			     display vertically by a factor of 2,
			     "pixelaspect 0.5" might do.
			     The value is used when computing the
			     projection of a camera associated with
			     this window.)

	  [ "}" ]			(matching closebrace)

Window objects are used in the Geomview window and ui-panel commands to set default properties for future windows or to change those of an existing window.