cv/mgflib/mgfdoc.tr

.\" SCCSid "$SunId$ LBL"
.nr PS 11
.ps 11
.nr VS 12
.vs 12
.nr PD .5v
.ds LF MGF
.ds RF Version 1.0
.DA May 1995
.TL
The Materials and Geometry Format
.AU
Greg Ward
.br
Lawrence Berkeley Laboratory
.NH
Introduction
.LP
The Materials and Geometry Format (referred to henceforth as MGF)
is a description language for 3-dimensional environments expressly
suited to visible light simulation and rendering.
The materials are physically-based and rely on standard and
well-accepted definitions of color, reflectance and transmittance
for good accuracy and reproducibility.
The geometry is based on boundary representation using simple
geometric primitives such as polygons, spheres and cones.
The file format itself is terse but human-readable ASCII text.
.NH 2
What makes MGF special?
.LP
There are three principal reasons to use MGF as an input language for
lighting simulation and physically-based rendering:
.RS
.IP 1.
It's the only existing format that describes materials physically.
.IP 2.
It is endorsed by the Illuminating Engineering Society of North
America (IESNA) as part of their LM-63-1995 standard for luminaire data.
.IP 3.
It's easy and fun to support since it comes with a standard parser
and sample scenes and objects at the web site,
"http://radsite.lbl.gov/mgf/HOME.html".
.RE
.LP
The standard parser provides both immediate and a long-term
benefits, since it presents a programming interface that is more
stable even than the language itself.
Unlike AutoCAD DXF and other de facto standards, a change to the
language will not break existing programs.
This is because the parser gives the calling software only those
entities it can handle.
If the translator understands only polygons, it will be given only
polygons.
If a new geometric primitive is included in a later version of the
standard, the new parser that comes with it will still be able to
express this entity as polygons.
Thus, the urgency of modifying code to support a changing standard
is removed, and long-term stability is assured.
.LP
This notion of
.I extensibility
is a cornerstone of the format, and it goes well beyond the
extensibility of other languages because is guarantees that new
versions of the standard will not break existing programs, and the
new information will be used as much as possible.
Other languages either require that all translators stay up to date
with the latest standard, or allow forward compatibility by simply
.I ignoring
new entities.
In MGF, if NURBS are added at some point and the translator or
loader does not handle them directly, the new version of the parser
will automatically convert them to smoothed polygons without
changing a single line of the calling program.
It is merely necessary to link to the new library, and all the new
entities are supported\(dg.
.FS
\(dgIf an old version of the parser encounters new entities it does
not recognize, the default action is to ignore them, printing a warning
message.
This may be overridden to support custom entities, but such
practice is discouraged because it weakens the standard.
.FE
.NH 2
What does MGF look like?
.LP
MGF has a simple entity-per-line structure, with a similar
appearance to Wavefront's .OBJ format.
Each entity is specified by a short keyword, and
arguments are separated by white space (tabs and/or spaces).
A newline may be escaped with a backslash ('\\'), in which case it
counts as a space.
Lines and continued lines may have up to 4096 characters, including
newlines, tabs and spaces.
A comment is an ignored entity whose keyword is the pound sign ('#').
.LP
Here is an MGF file that describes a simple two-drawer file cabinet:
.DS
# Conversion from inches to meters
xf -s .0254
# Surface material
m burgundy_formica =
        c
                cxy .362 .283
        rd .0402
        c
        rs .0284 .05
        sides 1
# Cabinet vertices
v fc.xy =
        p .05 0 0
v fc.xY =
        p .05 18 0
v fc.XY =
        p 35.95 18 0
v fc.Xy =
        p 35.95 0 0
# Cabinet
prism fc.xy fc.xY fc.XY fc.Xy 24
# Drawer vertices
v fcd.Xz =
        p 34 0 0
v fcd.XZ =
        p 34 0 10
v fcd.xZ =
        p 0 0 10
v fcd.xz =
        p 0 0 0
# Two drawers
o drawer
        xf -t 1 18.1 2 -a 2 -t 0 0 11
        prism fcd.xz fcd.Xz fcd.XZ fcd.xZ .9
        xf
o
# End of units conversion
xf
.DE
.NH 2
MGF's place in the world of standards
.LP
MGF was developed initially to support detailed geometric
description of light fixtures for the IESNA luminaire data standard,
publication LM-63\(dg.
.FS
\(dgTo obtain the latest version of this standard, write to:
Illuminating Engineering Society of North America,
345 East 47th St.,
New York, NY  10017.
.FE
Existing standards for geometric description were either too
cumbersome (e.g.
.I Radiance)
or did not include physical materials (e.g. IGES).
It was noted early on that a standard able to fully describe
luminaires would necessarily be
capable of describing other objects as well; indeed whole
environments could be defined this way.
Since the descriptions would be physical, they could serve as input
to both lighting simulation and rendering software.
A standard language for describing the appearance of physical
objects has been lacking for some time, and current efforts in this
direction (i.e. STEP) seem several years away from fruition.
(There are other languages for describing realistic scenes
that deserve mention here, such as VRML and the Manchester Scene
Description Language, but none give specific attention to physical
material properties and are thus unsuitable for lighting
simulation.)\0
.LP
In short, we saw this as an opportunity to offer the lighting and
rendering community a simple and easy-to-support standard for
describing environments in a physically valid way.
Our hope is that this will promote sharing color, material and object
libraries as well as complete scene descriptions.
Sharing libraries is of obvious benefit to users and software
developers alike.
Sharing scenes should also permit
comparisons between rendering systems and
intervalidation of lighting calculations.
As anyone who works in this field knows, modeling is the most
difficult step in creating any simulation or rendering, and there is
no excuse for this data being held prisoner by a proprietary data
format.
.NH
MGF Basics
.LP
The default coordinate system in MGF is right-handed with
distances given in meters, though this can be effectively changed
by specifying a global transformation.
The transformation context is affected by the
.UL xf
entity, and the whole of MGF can be understood in terms of entities
and contexts.
.NH 2
Entities and Contexts
.LP
An
.I entity
in MGF is any non-blank line, which must be one of a finite set of
command keywords followed by zero or more arguments.
(As mentioned previously, an entity may continue over multiple lines
by escaping the newline with a backslash.)\0
Table 1 gives a list of entities and their expected arguments.
Section 3 gives more detailed information on each entity.
.KF
.TS
expand, box;
l l l.
Keyword Arguments       Interpretation
=       =       =
#       [anything ...]  a comment
o       [name]  begin/end object context
xf      [xform] begin/end transformation context
i       pathname [xform]        include file (with transformation)
ies     pathname [-m f][xform]  include IES luminaire (with transformation)
_       _       _
c       [id [= [template]]]     get/set color context
cxy     x y     set CIE (x,y) chromaticity for current color
cspec   l_min l_max v1 v2 ...   set relative spectrum for current color
cct     temperature     set spectrum based on black body temperature
cmix    w1 c1 w2 c2 ... mix named colors to make current color
_       _       _
m       [id [= [template]]]     get/set material context
sides   {1|2}   set number of sides for current material
rd      rho_d   set diffuse reflectance for current material
td      tau_d   set diffuse transmittance for current material
ed      epsilon_d       set diffuse emittance for current material
rs      rho_s alpha_r   set specular reflectance for current material
ts      tau_s alpha_t   set specular transmittance for current material
ir      n_real n_imag   set index of refraction for current material
_       _       _
v       [id [= [template]]]     get/set vertex context
p       x y z   set point position for current vertex
n       dx dy dz        set surface normal for current vertex
_       _       _
f       v1 v2 v3 ...    polygon using current material, spec. vertices
sph     vc radius       sphere
cyl     v1 radius v2    truncated right cylinder (open-ended)
cone    v1 rad1 v2 rad2 truncated right cone (open-ended)
prism   v1 v2 v3 ... length     truncated right prism (closed solid)
ring    vc rmin rmax    circular ring with inner and outer radii
torus   vc rmin rmax    circular torus with inner and outer radii
.TE
.QP
.B "Table 1".
MGF entities and their arguments.
Arguments in brackets are optional.
Arguments in curly braces mean one of the given choices must
appear.
Ellipsis (...) mean that any number of arguments may be given.
.sp
.KE
.LP
A
.I context
describes the current state of the interpreter, and affects or is
affected by certain entities as they are read in.
MGF contexts can be divided into two types,
.I "hierarchical contexts"
and
.I "named contexts".
.LP
Hierarchical contexts are manipulated by a single entity and
have an associated "stack" onto which new
contexts are "pushed" using the entity.
The last context may be "popped" by giving the entity again with no
arguments.
The two hierarchical contexts in MGF are the current transformation,
manipulated with the
.UL xf
entity, and the current object, manipulated with the
.UL o
entity.
.KF
.TS
expand, allbox;
l c l l l.
Context Cntl. Entity    Default Value   Field Entities  Affects
=       =       =       =       =
Object  o       -       -       -
Transform       xf      -       -       T{
f, sph, cyl, cone,
ring, torus, prism
T}
Material        m       2-sided black   T{
sides, rd, td,
ed, rs, ts
T}      T{
f, sph, cyl, cone,
ring, torus, prism
T}
Color   c       neutral grey    T{
cxy, cspec, cct, cmix
T}      T{
rd, td, ed, rs, ts
T}
Vertex  v       T{
(0,0,0),
no normal
T}      p, n    T{
f, sph, cyl, cone,
ring, torus, prism
T}
.TE
.QP
.B "Table 2".
MGF contexts and their related entities and default values.
.sp
.KE
.LP
Named contexts in contrast hold sets of values that are swapped
in and out one at a time.
There are three named contexts in MGF, the current material, the
current color and the current vertex.
Each one may be associated with an identifier (any non-white
sequence of printing ASCII characters beginning with a letter),
and one of each is in effect at any given time.
Initially, these contexts are unnamed, and invoking an unnamed
context always returns to the original (default) values.
(See Table 2 for a list of contexts, their related
entities and defaults.)\0
.LP
It is easiest to think of a context as a "scratch space" where
values are written by some entities and read by others.
Naming a context allows us to reestablish the same scratch space
later, usually for reference but it can be altered as well.
Let us say we wanted to create a smooth blue plastic material with a
diffuse reflectance of 20% and a specular reflectance of 4%:
.DS
# Establish a new material context called "blue_plastic"
m blue_plastic =
        # Reestablish a previous color context called "blue"
        c blue
        # Set the diffuse reflectance, which uses the above color
        rd .20
        # Get the unnamed color context (always starts out grey)
        c
        # Set the specular reflectance, which is uncolored
        rs .04 0
# We're done, the current material context is now "blue_plastic"
.DE
Note that the above assumes that we have previously defined a color
context named "blue".
If we forgot to do that, the above description would generate an
"undefined" error.
The color context affects the material context indirectly because it
is read by the specular and diffuse reflectance entities, which are
in turn written to the current material.
It is not necessary to indent the entities that affect the material
definition, but it improves readability.
Note also that there is no explicit end to the material definition.
As long as a context remains in effect, its contents may be altered
by its field entities.
This will not affect previous uses of the context, however.
For example, a surface entity following the above definition will 
have the specified color and reflectance, and later changes to the
material "blue_plastic" will have no effect on it.
.LP
Each of the three named contexts has an associated entity that
controls it.
The material context is controlled by the
.UL m
entity, the color context is controlled by the
.UL c
entity, and the vertex context is controlled by the
.UL v
entity.
There are exactly four forms for each entity.
The first form is the keyword by itself, which establishes
an unnamed context with predetermined default values.
This is a useful way to set values without worrying about saving
them for recall later.
The second form is to give the keyword with a previously defined
name.
This reestablishes a prior context for reuse.
The third form is to give the keyword with a name followed by an
equals sign.
(There must be a space between the name and the equals sign, since
it is a separate argument.)\0
This establishes a new context and assigns it the same default
values as the unnamed context.
The fourth and final form gives the keyword followed by a name then
an equals then the name of a previous context definition.
This establishes a new context for the first name, assigning the
values from the second named context rather than the usual defaults.
This is a convenient way create an alias or
to modify a context under a new name (i.e. "save as").
.NH 2
Hierarchical Contexts and Transformations
.LP
As mentioned in the last subsection, there are two hierarchical
contexts in MGF, the current object and the current transformation.
We will start by discussing the current object, since it is
the simpler of the two.
.NH 3
Objects
.LP
There is no particular need in lighting simulation or rendering to
name objects, but it may help the user
to know what object a particular surface is associated with.
The
.UL o
entity provides a convenient mechanism for associating names with
surfaces.
The basic use of this entity is as follows:
.DS
o object_name
        [object entities...]
        o subobject_name
                [subobject entities...]
        o
        [more object entities and subobjects...]
o
.DE
The
.UL o
keyword by itself marks the end of an object context.
Any number of hierarchical context levels are supported, and there are no
rules governing the choice of object names except that they begin
with a letter and be made up of printing, non-white ASCII characters.
Indentation is not necessary of course, but it does improve
readability.
.NH 3
Transformations
.LP
MGF supports only rigid-body (i.e. non-distorting) transformations
with uniform scaling.
Unlike the other contexts, transformations have no associated
name, only arguments.
Thus, there is no way to reestablish a previous transformation other
than to give the same arguments over again.
Since the arguments are concise and self-explanatory, this was thought
sufficient.
The following transformation flags and
parameters are defined:
.TS
center;
l l.
-t dx dy dz     translate objects along the given vector
-rx degrees     rotate objects about the X-axis
-ry degrees     rotate objects about the Y-axis
-rz degrees     rotate objects about the Z-axis
-s scalefactor  scale objects by the given factor
-mx     mirror objects about the Y-Z plane
-my     mirror objects about the X-Z plane
-mz     mirror objects about the X-Y plane
-i N    repeat the following arguments N times
-a N    make an array of N geometric instances
.TE
Transform arguments have a cumulative effect.
That is, a rotation
about X of 20 degrees followed by a rotation about X of -50 degrees
results in a total rotation of -30 degrees.
However, if the two
rotations are separated by some translation vector, the cumulative
effect is quite different.
It is best to think of each argument as
acting on the included geometric objects, and each subsequent transformation
argument affects the objects relative to their new position/orientation.
.LP
For example, rotating an object about its center is most easily done
by translating
the object back to the origin, applying the desired rotation, and translating
it again back to its original position, like so:
.DS
# rotate an included object 20 degrees clockwise looking down
# an axis parallel to Y and passing through the point (15,0,-35)
xf -t -15 0 35 -ry -20 -t 15 0 -35
i object.mgf
xf
.DE
Note that the include entity,
.UL i,
permits a transformation to be given with it, so the above could
have been written more compactly as:
.DS
i object.mgf -t -15 0 35 -ry -20 -t 15 0 -35
.DE
.LP
Rotations are given in degrees counter-clockwise about a principal axis.
That is, with the thumb of the right hand pointing in the direction
of the axis, rotation follows the curl of the fingers.
.LP
The transform entity itself is cumulative, but in the reverse
order to its arguments.
That is, later transformations (i.e. enclosed transformations)
are prepended to existing (i.e. enclosing) ones.
A transform command
with no arguments is used to return to the previous condition.
It is
necessary that transforms and their end statements ("xf" by itself) be
balanced in a file, so that later or enclosing files are not affected.
.LP
Transformations apply only to geometric types, e.g. polygons, spheres, etc.
Vertices and the components that go into geometry are not directly affected.
This is to avoid confusion and the inadvertent multiple application of a
given transformation.
For example:
.DS
xf -t 5 0 0
v v1 =
        p 0 10 0
        n 0 0 1
xf -rx 180
# Transform now in effect is "-rx 180 -t 5 0 0"
ring v1 0 2
xf
xf
.DE
The final ring center is (5,-10,0) -- note that the vertex itself is
not affected by the transformation, only the geometry that calls on
it.
The normal orientation is (0,0,-1) due to the rotation about X,
which also reversed the sign of the central Y coordinate.
.NH 3
Arrays
.LP
The -a N transform specification causes the following transform
arguments to be repeated along with the contents of the included
objects N times.
The first instance of the geometry will be in its
initial location; the second instance will be repositioned according
to the named transformation; the third instance will be repositioned by
applying this transformation twice, and so on up to N-1 applications.
.LP
Multi-dimensional arrays may be specified with a single include
entity by giving multiple array commands separated by their
corresponding transforms.
A final transformation may be given
by preceding it with a -i 1 specification.
In other words, the
scope of an array command continues until the next -i or -a option.
.LP
The following MGF description places 60 spheres at a one unit spacing
in a 3x4x5 array, then moves the whole thing to an origin of
(15,30,45):
.DS
v v0 =
        p 0 0 0
xf -a 3 -t 1 0 0 -a 4 -t 0 1 0 -a 5 -t 0 0 1 -i 1 -t 15 30 45
sph v0 0.1
xf
.DE
Note the "-i 1" in the specification, which marks the end of the
third array arguments before the final translation.
.NH 2
Detailed MGF Example
.LP
The following example of a simple room with a single door
and six file cabinets shows MGF in action, with copious comments to
help explain what's going on.
.LP
.DS
# "ceiling_tile" is a diffuse white surface with 75% reflectance:
# Create new named material context and clear it
m ceiling_tile =
        # Specify one-sided material so we can see through from above
        sides 1
        # Set neutral color
        c
        # Set diffuse reflectance
        rd .75
# "stainless_steel" is a mostly specular surface with 70% reflectance:
m stainless_steel =
        sides 1
        c
        # Set specular reflectance to 50%, .08 roughness
        rs .5 .08
        # Other 20% reflectance is diffuse
        rd .2

# The following materials were measured with a spectrophotometer:
m beige_paint =
        sides 1
        # Set diffuse spectral reflectance
        c
                # Spectrum measured in 10 nm increments from 400 to 700 nm
                cspec 400 700 35.29 44.87 47.25 47.03 46.87 47.00 47.09 \\\\
                         47.15 46.80 46.17 46.26 48.74 51.08 51.31 51.10 \\\\
                         51.11 50.52 50.36 51.72 53.61 53.95 52.08 49.49 \\\\
                         48.30 48.75 49.99 51.35 52.75 54.44 56.34 58.00
        rd 0.5078
        # Neutral (grey) specular component
        c
        rs 0.0099 0.08000
m mottled_carpet =
        sides 1
        c
                cspec 400 700 11.23 11.28 11.39 11.49 11.61 11.73 11.88 \\\\
                         12.02 12.12 12.19 12.30 12.37 12.37 12.36 12.34 \\\\
                         12.28 12.22 12.29 12.45 12.59 12.70 12.77 12.82 \\\\
                         12.88 12.98 13.24 13.67 14.31 15.55 17.46 19.75
        rd 0.1245
m reddish_cloth =
        # 2-sided so we can observe it from behind
        sides 2
        c
                cspec 400 700 28.62 27.96 27.86 28.28 29.28 30.49 31.61 \\\\
                         32.27 32.26 31.83 31.13 30.07 29.14 29.03 29.69 \\\\
                         30.79 32.30 33.90 34.56 34.32 33.85 33.51 33.30 \\\\
                         33.43 34.06 35.26 37.04 39.41 42.55 46.46 51.00
        rd 0.3210
m burgundy_formica =
        sides 1
        c
                cspec 400 700  3.86  3.74  3.63  3.51  3.34  3.21  3.14 \\\\
                          3.09  3.08  3.14  3.13  2.91  2.72  2.74  2.72 \\\\
                          2.60  2.68  3.40  4.76  6.05  6.65  6.75  6.68 \\\\
                          6.63  6.56  6.51  6.46  6.41  6.36  6.34  6.34
        rd 0.0402
        c
        rs 0.0284 0.05000
m speckled_grey_formica =
        sides 1
        c
                cspec 400 700 30.95 44.77 51.15 52.60 53.00 53.37 53.68 \\\\
                         54.07 54.33 54.57 54.85 55.20 55.42 55.51 55.54 \\\\
                         55.46 55.33 55.30 55.52 55.81 55.91 55.92 56.00 \\\\
                         56.22 56.45 56.66 56.72 56.58 56.44 56.39 56.39
        rd 0.5550
        c
        rs 0.0149 0.15000

# 40' x 22' x 9' office space with no windows and one door

# All measurements are in inches, so enclose with a metric conversion:
xf -s .0254

# The room corner vertices:
v rc.xyz =
        p 0 0 0
v rc.Xyz =
        p 480 0 0
v rc.xYz =
        p 0 264 0
v rc.xyZ =
        p 0 0 108
v rc.XYz =
        p 480 264 0
v rc.xYZ =
        p 0 264 108
v rc.XyZ =
        p 480 0 108
v rc.XYZ =
        p 480 264 108

# The floor:
# Push object name
o floor
        # Get previously defined carpet material
        m mottled_carpet
        # Polygonal face using defined vertices
        f rc.xyz rc.Xyz rc.XYz rc.xYz
# Pop object name
o

# The ceiling:
o ceiling
        m ceiling_tile
        f rc.xyZ rc.xYZ rc.XYZ rc.XyZ
o

# The door outline vertices:
v do.xz =
        p 216 0 0
v do.Xz =
        p 264 0 0
v do.xZ =
        p 216 0 84
v do.XZ =
        p 264 0 84

# The walls:
o wall
        m beige_paint
        o x
                f rc.xyz rc.xYz rc.xYZ rc.xyZ
        o
        o X
                f rc.Xyz rc.XyZ rc.XYZ rc.XYz
        o
        o y
                f rc.xyz rc.xyZ rc.XyZ rc.Xyz do.Xz do.XZ do.xZ do.xz
        o
        o Y
                f rc.xYz rc.XYz rc.XYZ rc.xYZ
        o
o

# The door and jam vertices:
v djo.xz =
        p 216 .5 0
v djo.xZ =
        p 216 .5 84
v djo.XZ =
        p 264 .5 84
v djo.Xz =
        p 264 .5 0
v dji.Xz =
        p 262 .5 0
v dji.XZ =
        p 262 .5 82
v dji.xZ =
        p 218 .5 82
v dji.xz =
        p 218 .5 0
v door.xz =
        p 218 0 0
v door.xZ =
        p 218 0 82
v door.XZ =
        p 262 0 82
v door.Xz =
        p 262 0 0

# The door, jam and knob
o door
        m burgundy_formica
        f door.xz door.xZ door.XZ door.Xz
        o jam
                m beige_paint
                f djo.xz djo.xZ djo.XZ djo.Xz dji.Xz dji.XZ dji.xZ dji.xz
                f djo.xz do.xz do.xZ djo.xZ
                f djo.xZ do.xZ do.XZ djo.XZ
                f djo.Xz djo.XZ do.XZ do.Xz
                f dji.xz dji.xZ door.xZ door.xz
                f dji.xZ dji.XZ door.XZ door.xZ
                f dji.Xz door.Xz door.XZ dji.XZ
        o
        o knob
                m stainless_steel
                # Define vertices needed for curved geometry
                v kb1 =
                        p 257 0 36
                v kb2 =
                        p 257 .25 36
                        n 0 1 0
                v kb3 =
                        p 257 2 36
                # 1" diameter cylindrical base from kb1 to kb2
                cyl kb1 1 kb2
                # Ring at base of knob stem
                ring kb2 .4 1
                # Knob stem
                cyl kb2 .4 kb3
                # Spherical knob
                sph kb3 .85
        o
o

# Six file cabinets (36" wide each)
# ("filecab.inc" was given as an earlier example in Section 1.2)
o filecab.x
        # include a file as an array of three 36" apart
        i filecab.inc -t -36 0 0 -rz -90 -t 1 54 0 -a 3 -t 0 36 0
o
o filecab.X
        # the other three cabinets
        i filecab.inc -rz 90 -t 479 54 0 -a 3 -t 0 36 0
o

# End of transform from inches to meters:
xf

# The 10 recessed fluorescent ceiling fixtures
ies hlrs2gna.ies -t 1.2192 2.1336 2.74 -a 5 -t 2.4384 0 0 -a 2 -t 0 2.4384 0
.DE
.bp
.NH
MGF Entity Reference
.LP
There are currently 28 entities in the MGF specification.
For ease of reference we have broken these into five categories:
.IP 1.
General
.TS
lw(.75i) lw(1.75i) lw(3i).
#       [anything ...]  a comment
o       [name]  begin/end object context
xf      [xform] begin/end transformation context
i       pathname [xform]        include file (with transformation)
ies     pathname [-m f][xform]  include IES luminaire (with transformation)
.TE
.IP 2.
Color
.TS
lw(.75i) lw(1.75i) lw(3i).
c       [id [= [template]]]     get/set color context
cxy     x y     set CIE (x,y) chromaticity for current color
cspec   l_min l_max v1 v2 ...   set relative spectrum for current color
cct     temperature     set spectrum based on black body temperature
cmix    w1 c1 w2 c2 ... mix named colors to make current color
.TE
.IP 3.
Material
.TS
lw(.75i) lw(1.75i) lw(3i).
m       [id [= [template]]]     get/set material context
sides   {1|2}   set number of sides for current material
rd      rho_d   set diffuse reflectance for current material
td      tau_d   set diffuse transmittance for current material
ed      epsilon_d       set diffuse emittance for current material
rs      rho_s alpha_r   set specular reflectance for current material
ts      tau_s alpha_t   set specular transmittance for current material
ir      n_real n_imag   set index of refraction for current material
.TE
.IP 4.
Vertex
.TS
lw(.75i) lw(1.75i) lw(3i).
v       [id [= [template]]]     get/set vertex context
p       x y z   set point position for current vertex
n       dx dy dz        set surface normal for current vertex
.TE
.IP 5.
Geometry
.TS
lw(.75i) lw(1.75i) lw(3i).
f       v1 v2 v3 ...    polygon using current material, spec. vertices
sph     vc radius       sphere
cyl     v1 radius v2    truncated right cylinder (open-ended)
cone    v1 rad1 v2 rad2 truncated right cone (open-ended)
prism   v1 v2 v3 ... length     truncated right prism (closed solid)
ring    vc rmin rmax    circular ring with inner and outer radii
torus   vc rmin rmax    circular torus with inner and outer radii
.TE
.ds LH General Entities
.ds RH #
.bp
.SH
NAME
.LP
# - a comment
.SH
SYNOPSIS
.LP
.B #
[
.I anything
]
.SH
DESCRIPTION
.LP
A comment is a bit of text explanation.
Since it is an entity like any other (except that it has no effect),
there must be at least one space between the keyword (which is a
pound sign) and the "arguments," and the end of line may be escaped
as usual with the backslash character ('\\').
.LP
A comment may actually be used to hold auxiliary information such as
view parameters, which may be interpreted by some destination program.
Care should be taken under such circumstances that the user does not
inadvertently mung or mimic this information in other comments, and
it is therefore advisable to use an additional set of identifying
characters to distinguish such data.
.SH
EXAMPLE
.DS
# The following include file is in inches, so convert to meters
i cubgeom.inc -s .0254
# Stuff we don't want to see at the moment:
# i person.mgf -t 3 2 0
# ies hlrs3gna.ies -rz 90 -t 1.524 1.8288 2.74 \\\\
                -a 6 -t 1.8288 0 0 -a 2 -t 0 3.048 0
.DE
.ds RH O
.bp
.SH
NAME
.LP
o - begin or end object context
.SH
SYNOPSIS
.LP
.B o
[
.I name
]
.SH
DESCRIPTION
.LP
If
.I name
is given, we push a new object context onto the stack, which is to
say that we begin a new subobject by this name\(dg.
.FS
\(dgA name is any sequence of printing, non-white ASCII characters
beginning with a letter.
.FE
If the
.UL o
keyword is given by itself, then we pop the last object context off
the stack, which means that we leave the current subobject.
.LP
All geometry between the start of an object context and its matching
end statement is associated with the given name.
This may be used in modeling software to help identify objects and
subobjects, or it may be ignored altogether.
.LP
Object begin and end statements should be balanced in a file, and
care should be taken not to overlap transform
.UL (xf)
contexts with object contexts, especially when arrays are involved.
This is because the standard parser will assign object contexts to
instanced geometry, which can get confused with other object
contexts if a clear enclosure is not maintained.
.SH
EXAMPLE
.DS
o body
        o torso
                i torso.mgf
        o
        o arm
                o left
                        i leftarm.mgf
                o
                o right
                        i leftarm.mgf -mx
                o
        o
o
.DE
.SH
SEE ALSO
.LP
.UL xf
.ds RH XF
.bp
.SH
NAME
.LP
xf - begin or end transformation context
.SH
SYNOPSIS
.LP
.B xf
[
.I transform
]
.SH
DESCRIPTION
.LP
If a set of
.I transform
arguments are given, we push a new transformation context onto the
stack.
If the
.UL xf
keyword is given by itself, then we pop the last transformation
context off the stack.
The total transformation in effect at any given time is
computed by prepending each set subcontext arguments onto those of
its enclosing context.
This and other details about transformation specifications
are explained in some detail in section 2.2.2.
.LP
The following transformation flags and
parameters are defined:
.TS
center;
l l.
-t dx dy dz     translate objects along the given vector
-rx degrees     rotate objects about the X-axis
-ry degrees     rotate objects about the Y-axis
-rz degrees     rotate objects about the Z-axis
-s scalefactor  scale objects by the given factor
-mx     mirror objects about the Y-Z plane
-my     mirror objects about the X-Z plane
-mz     mirror objects about the X-Y plane
-i N    repeat the following arguments N times
-a N    make an array of N geometric instances
.TE
.SH
EXAMPLE
.DS
# Create 3x5 array of evenly-spaced spheres (grid size = 3)
v vc =
        p 0 0 0
xf -t 1 1 10 -a 3 -t 3 0 0 -a 5 -t 0 3 0
        sph vc .5
xf
.DE
.SH
SEE ALSO
.LP
.UL i,
.UL ies,
.UL o
.ds RH I
.bp
.SH
NAME
.LP
i - include MGF data file
.SH
SYNOPSIS
.LP
.B i
.I pathname
[
.I transform
]
.SH
DESCRIPTION
.LP
Include the information contained in the file
.I pathname.
If a
.I transform
specification is given, then it will be applied as though the
include statement were enclosed by beginning and ending
.UL xf
entities with this transformation.
.LP
The
.I pathname
will be interpreted relative to the enclosing MGF file.
That is, if the file containing the include statement is in some
parent or subdirectory, then the given pathname is appended to this
directory.
It is illegal to specify a
.I pathname
relative to the root directory, and the MGF standard requires that
all filenames adhere to the ISO-9660 8.3 name format for maximum
portability between systems.
The directory separator is defined to be slash ('/'), and drive
specifications (such as "c:") are not allowed.
All pathnames should be given in lower case, and will be converted to
upper case on systems that require it.
(That way, there are no accidental name collisions.)\0
.LP
The suggested suffix for MGF-adherent files is ".mgf".
Files that are not in metric units but are in MGF may be given any
suffix, but we suggest using ".inc" as a convention.
.SH
EXAMPLE
.DS
# Define vertices for 62x30" partition
i pv62x30.inc
# Insert 2 62x30" partitions
o cpart1
        i partn.inc -t 75 130.5 0
o
o cpart3
        i partn.inc -t 186 130.5 0
o
# Define vertices for 62x36" partition
i pv62x36.inc
# Insert 62x36" partition
o cpart2
        i partn.inc -t 105 130.5 0
o
.DE
.SH
SEE ALSO
.LP
.UL ies,
.UL o,
.UL xf
.ds RH IES
.bp
.SH
NAME
.LP
ies - include IESNA luminaire file
.SH
SYNOPSIS
.LP
.B ies
.I pathname
[
.B \-m
.I multiplier
]
[
.I transform
]
.SH
DESCRIPTION
.LP
Load the IES standard luminaire information contained in the file
.I pathname.
If a
.I multiplier
is given, all candela values will be multiplied by this factor.
(This option must appear first if present.)\0
If a
.I transform
specification is given, then it will be applied as though the
statement were enclosed by beginning and ending
.UL xf
entities with this transformation.
.LP
The
.I pathname
will be interpreted relative to the enclosing MGF file, and all
restrictions discussed under the
.UL i
entity also apply to the IES file name.
The suggested suffix is ".ies", but this has not been followed
consistently by lighting manufacturers.
.SH
EXAMPLE
.DS
# Insert 10 2x4' fluorescent troffers in two groups
ies cf9pr240.ies -t 3.6576 2.1336 2.74 -a 3 -t 2.4384 0 0 -a 2 -t 0 2.4384 0
ies cf9pr240.ies -rz 90 -t 1.2192 1.8288 2.74 \\\\
                -a 2 -t 9.7536 0 0 -a 2 -t 0 3.048 0
.DE
.SH
SEE ALSO
.LP
.UL i,
.UL o,
.UL xf
.ds LH Color Entities
.ds RH C
.bp
.SH
NAME
.LP
c - get or set the current color context
.SH
SYNOPSIS
.LP
.B c
[
.I id
[
.B =
[
.I template
]
]
]
.SH
DESCRIPTION
.LP
If the
.UL c
keyword is given by itself, then it establishes the unnamed color
context, which is neutral (i.e. equal-energy) grey.
This context may be modified, but the changes will not be saved.
.LP
If the keyword is followed by an identifier
.I id,
then it reestablishes a previous context.
If the specified context was never defined, an error will result.
.LP
If the entity is given with an identifier
followed by an equals sign ('='), then a new context is established,
and cleared to the default neutral grey.
(Note that the equals sign must be separated from other
arguments by white space to be properly recognized.)\0
If the equals sign is followed by a second identifier
.I template,
then this previously defined color will be used as a source of
default values rather than grey.
This is most useful for establishing a color alias.
.SH
EXAMPLE
.DS
# Define the color "red32"
c red32 =
        cxy .42 .15
# Make "cabinet_color" an alias for "red32"
c cabinet_color = red32

# Later in another part of the description...

# Get our cabinet color
c cabinet_color
# Get the geometry
i cabgeom.mgf
.DE
.SH
SEE ALSO
.LP
.UL cct,
.UL cmix,
.UL cspec,
.UL cxy,
.UL m
.ds RH CXY
.bp
.SH
NAME
.LP
cxy - set the CIE (x,y) chromaticity for the current color
.SH
SYNOPSIS
.LP
.B cxy
.I "x y"
.SH
DESCRIPTION
.LP
This entity sets the current color using (x,y) chromaticity
coordinates for the 1931 CIE standard 2 degree observer.
Legal values for
.I x
and
.I y
are greater than zero and sum to less than one, and more
specifically they must fit within the curve of the visible spectrum.
The
.I x
coordinate roughly corresponds to the red part of the spectrum and
the
.I y
coordinate corresponds to the green.
The CIE z coordinate is implicit, since it is equal to (1-x-y).
.LP
All colors in MGF are absolute, thus colorimeter measurements should
be conducted the same for surfaces as for light sources.
Applying a standard illuminant calculation is redundant and
introduces inaccuracies, and should therefore be avoided if
possible.
.LP
Conversion between CIE colors and those more commonly used in
computer graphics are described in the application notes section
6.1.1.
.SH
EXAMPLE
.DS
# Set unnamed color context
c
# Set CIE chromaticity to a bluish hue
cxy .15 .2
# Apply color to diffuse reflectance of 15%
rd .15
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL cct,
.UL cmix,
.UL cspec
.ds RH CSPEC
.bp
.SH
NAME
.LP
cspec - set the relative spectrum for the current color
.SH
SYNOPSIS
.LP
.B cspec
.I "l_min l_max o1 o2 ... oN"
.SH
DESCRIPTION
.LP
Assign a relative spectrum measured between
.I l_min
and
.I l_max
nanometers at evenly spaced intervals.
The first value,
.I o1
corresponds to the measurement at
.I l_min,
and the last value,
.I oN
corresponds to the measurement at
.I l_max.
Values in between are separated by
.I "(l_max-l_min)/(N-1)"
nanometers.
All values should be non-negative unless defining a component for
complementary color mixing, and the spectrum outside of the
specified range is assumed to be zero.
(The visible range is 380 to 780 nm.)\0
The actual units and scale of the measurements do not matter,
since the total will be
normalized according to whatever the color is modifying
(e.g. photometric reflectance or emittance).
.SH
EXAMPLE
.DS
# Color measured at 10 nm increments from 400 to 700
m reddish_cloth =
        c
                cspec 400 700 28.62 27.96 27.86 28.28 29.28 30.49 31.61 \\\\
                         32.27 32.26 31.83 31.13 30.07 29.14 29.03 29.69 \\\\
                         30.79 32.30 33.90 34.56 34.32 33.85 33.51 33.30 \\\\
                         33.43 34.06 35.26 37.04 39.41 42.55 46.46 51.00
        rd 0.3210
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL cct,
.UL cmix,
.UL cxy
.ds RH CCT
.bp
.SH
NAME
.LP
cct - set the current color to a black body spectrum
.SH
SYNOPSIS
.LP
.B cct
.I temperature
.SH
DESCRIPTION
.LP
The
.UL cct
entity sets the current color to the spectrum of an ideal
black body radiating at
.I temperature
degrees Kelvin.
This is often the most convenient way to set the color of an
incandescent light source, but it is not recommended for
fluorescent lamps or other materials that do not fit a
black body spectrum.
.SH
EXAMPLE
.DS
# Define an incandescent source material at 3000 degrees K
m incand3000k =
        c
                cct 3000
        ed 1500
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL cmix,
.UL cspec,
.UL cxy
.ds RH CMIX
.bp
.SH
NAME
.LP
cmix - mix two or more named colors to make the current color
.SH
SYNOPSIS
.LP
.B cmix
.I "w1 c1 w2 c2 ..."
.SH
DESCRIPTION
.LP
The
.UL cmix
entity sums together two or more named colors using specified
weighting coefficients, which correspond to the relative
photometric brightness of each.
As in all color specifications, the result is normalized so the
absolute scale of the weights does not matter, only their relative 
values.
.LP
If any of the colors is a spectral quantity (i.e. from a
.UL cspec
or
.UL cct
entity), then all the colors are first converted to spectral
quantities.
This is done with an approximation for CIE (x,y) chromaticities,
which may be problematic depending on their values.
In general, it is safest to add together colors that are either
all spectral quantities or all CIE quantities.
.SH
EXAMPLE
.DS
# Define RGB primaries for a standard color monitor
c R =
        cxy 0.640 0.330
c G =
        cxy 0.290 0.600
c B =
        cxy 0.150 0.060
# Mix them together in appropriate amounts for white
c white =
        cmix 0.265 R 0.670 G 0.065 B
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL cct,
.UL cspec,
.UL cxy
.ds LH Material Entities
.ds RH M
.bp
.SH
NAME
.LP
m - get or set the current material context
.SH
SYNOPSIS
.LP
.B m
[
.I id
[
.B =
[
.I template
]
]
]
.SH
DESCRIPTION
.LP
If the
.UL m
keyword is given by itself, then it establishes
the unnamed material context, which is a perfect two-sided black absorber.
This context may be modified, but the changes will not be saved.
.LP
If the keyword is followed by an identifier
.I id,
then it reestablishes a previous context.
If the specified context was never defined, an error will result.
.LP
If the entity is given with an identifier
followed by an equals sign ('='), then a new context is established,
and cleared to the default material.
(Note that the equals sign must be separated from other
arguments by white space to be properly recognized.)\0
If the equals sign is followed by a second identifier
.I template,
then this previously defined material will be used as a source of
default values instead.
This may be used to establish a material alias, or to modify an
existing material and give it a new name.
.LP
The sum of the diffuse and specular reflectances and transmittances
must not be greater than one (with no negative values, obviously).
These values are assumed to be measured at normal incidence.
If an index of refraction is given, this may modify the balance between
diffuse and specular reflectance at other incident angles.
If the
material is one-sided (see 
.UL sides
entity), then it may be a dielectric interface.
In this case, the specular transmittance given is that which would be
measured at normal incidence for a pane of the material 5 mm thick.
This is important for figuring the actual transmittance for non-planar
geometries assuming a uniformly absorbing medium.
(Diffuse transmittance will not be affected by thickness.)\0
If the index of
refraction has an imaginary part, then the surface is a metal and this
implies other properties as well.
The default index of refraction is that of a vacuum, i.e. (1,0).
.SH
EXAMPLE
.DS
# Define a blue enamel paint
m blue_enamel =
        c
                cxy 0.2771 0.2975
        rd 0.5011
        c
        rs 0.0100 0.0350
# Assign blue_enamel to be the color of the south wall
m swall_mat = blue_enamel
# ...
# South wall face
m swall_mat
f sv1 sv2 sv3 sv4
.DE
.SH
SEE ALSO
.LP
.UL ed,
.UL ir,
.UL rd,
.UL rs,
.UL sides,
.UL td,
.UL ts
.ds RH SIDES
.bp
.SH
NAME
.LP
sides - set the number of sides for the current material
.SH
SYNOPSIS
.LP
.B sides
{
.B 1
|
.B 2
}
.SH
DESCRIPTION
.LP
The
.UL sides
entity is used to set the number of sides for the current material.
If a surface is two-sided, then it will appear
identical when viewed from either the front or the back.
If a surface is one-sided,
then it appears invisible when viewed from the back side.
This means
that a transmitting object will affect the light coming in through the
front surface and ignore the characteristics of the back surface,
unless the index of refraction is set.
If the index of refraction is set, then the object will act as a
solid piece of dielectric material.
In either case, the transmission properties of the exiting surface
should be the same as the incident surface for the model to be
physically valid.
.LP
The default number of sides is two.
.SH
EXAMPLE
.DS
# Describe a blue crystal ball
m blue_crystal =
        ir 1.650000 0
        # Solid dielectrics must use one-sided materials
        sides 1
        c
        rs 0.0602 0
        c
                cxy 0.3127 0.2881
        ts 0.6425 0
v sc =
        p 10 15 1.5
sph sc .02
.DE
.SH
SEE ALSO
.LP
.UL ed,
.UL ir,
.UL m,
.UL rd,
.UL rs,
.UL td,
.UL ts
.ds RH RD
.bp
.SH
NAME
.LP
rd - set the diffuse reflectance for the current material
.SH
SYNOPSIS
.LP
.B rd
.I rho_d
.SH
DESCRIPTION
.LP
Set the diffuse reflectance for the current material to
.I rho_d
using the current color to determine the spectral characteristics.
This is the fraction of visible light that is reflected from a
surface equally in all directions according to Lambert's law, and is
often called the "Lambertian component."
Photometric reflectance is measured according to v(lambda)
response function of the 1931 CIE standard 2
degree observer, and assumes an equal-energy white light source.
The value must be between zero and one, and may be further
restricted by the luminosity of the selected color.
(I.e. it is impossible to have a violet material with a photometric
reflectance close to one since the eye is less sensitive in this part
of the spectrum.)\0
.LP
The default diffuse reflectance is zero.
.SH
EXAMPLE
.DS
# An off-white paint with 70% reflectance
m flat_white70 =
        c
                cxy .3632 .3420
        rd .70
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL ed,
.UL ir,
.UL m,
.UL rs,
.UL sides,
.UL td,
.UL ts
.ds RH TD
.bp
.SH
NAME
.LP
td - set the diffuse transmittance for the current material
.SH
SYNOPSIS
.LP
.B td
.I tau_d
.SH
DESCRIPTION
.LP
Set the diffuse transmittance for the current material to
.I tau_d
using the current color to determine the spectral characteristics.
This is the fraction of visible light that is transmitted through a
surface equally in all (transmitted) directions.
Like reflectance, transmittance is measured according to the
standard v(lambda) curve, and assumes an equal-energy white light source.
It is probably not possible to create a material with a diffuse
transmittance above 50%, since well-diffused light will be reflected
as well.
.LP
The default diffuse transmittance is zero.
.SH
EXAMPLE
.DS
# Model a perfect spherical diffuser, i.e. light hitting \
        either side will be scattered equally in all directions
m wonderland_diffuser =
        c
        td .5
        rd .5
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL ed,
.UL ir,
.UL m,
.UL rd,
.UL rs,
.UL sides,
.UL ts
.ds RH ED
.bp
.SH
NAME
.LP
ed - set the diffuse emittance for the current material
.SH
SYNOPSIS
.LP
.B ed
.I epsilon_d
.SH
DESCRIPTION
.LP
Set the diffuse emittance for the current material to
.I epsilon_d
lumens per square meter using the current color to determine the
spectral characteristics.
Note that this is emittance rather than exitance, and therefore
does not include reflected or transmitted light, which is a function
of the other material settings and the illuminated environment.
.LP
The total lumen output of a convex emitting object
is the radiating area of that object multiplied by its emittance.
Therefore, one can compute the appropriate
.I epsilon_d
value for an emitter by dividing the total lumen output by the
radiating area (in square meters).
.LP
The default emittance is zero.
.SH
EXAMPLE
.DS
# A 100-watt incandescent bulb (1600 lumens) modeled as a sphere
m
        c
        cct 3000
        ed 87712
v cent =
        p 0 0 0
sph cent .0381
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL ir,
.UL m,
.UL rd,
.UL rs,
.UL sides,
.UL td,
.UL ts
.ds RH RS
.bp
.SH
NAME
.LP
rs - set the specular reflectance for the current material
.SH
SYNOPSIS
.LP
.B rs
.I "rho_s alpha_r"
.SH
DESCRIPTION
.LP
Set the specular reflectance for the current material to
.I rho_s
using the current color to determine the spectral characteristics.
The surface roughness parameter is set to
.I alpha_r,
which is the RMS height of surface variations over the
autocorrelation distance (equivalent to RMS facet slope).
A roughness value of zero means a perfectly smooth surface, and
values greater than 0.2 are unusual.
(See application notes section 6.1.2 for a comparison between the
roughness parameter and Phong specular power.)\0
.LP
The default specular reflectance is zero.
.SH
EXAMPLE
.DS
# Define a slightly rough brass metallic surface
m rough_brass =
        c
                cxy .3820 .4035
        # 30% specular, 9% diffuse
        rs .30 .08
        rd .09
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL ed,
.UL ir,
.UL m,
.UL rd,
.UL sides,
.UL td,
.UL ts
.ds RH TS
.bp
.SH
NAME
.LP
ts - set the specular transmittance for the current material
.SH
SYNOPSIS
.LP
.B ts
.I "tau_s alpha_t"
.SH
DESCRIPTION
.LP
Set the specular transmittance for the current material to
.I tau_s
using the current color to determine the spectral characteristics.
The effective surface roughness is set to
.I alpha_t.
Rays will be transmitted with the same distribution as they would
have been reflected with if this roughness value were given to the
.UL rs
entity.
.LP
The default specular transmittance is zero.
.SH
EXAMPLE
.DS
# Define a green glass material (58% transmittance)
m glass =
        sides 2
        ir 1.52 0
        c
        rs 0.0725 0
        c
                cxy .23 .38
        ts 0.5815 0
# Define an uncolored translucent plastic (40% transmittance)
m translucent =
        sides 2
        ir 1.4 0
        c
        rs .045 0
        ts .40 .05
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL ed,
.UL ir,
.UL m,
.UL rd,
.UL rs,
.UL sides,
.UL td
.ds RH IR
.bp
.SH
NAME
.LP
ir - set the complex index of refraction for the current material
.SH
SYNOPSIS
.LP
.B ir
.I "n_real n_imag"
.SH
DESCRIPTION
.LP
Set the index of refraction for the current material to
.I (n_real,n_imag).
If the material is a dielectric (as opposed to metallic), then
.I n_imag
should be zero.
For solid dielectric objects, the material should be made one-sided.
If it is being used for thin objects, then a two-sided
material is appropriate.
(See the
.UL sides
entity.)\0
.LP
The default index of refraction is that of a vacuum, (1,0).
.SH
EXAMPLE
.DS
# Define polished aluminum material
m polished_aluminum =
        # Complex index of refraction (from physics table)
        ir .770058 6.08351
        c
        rs .75 0
.DE
.SH
SEE ALSO
.LP
.UL c,
.UL ed,
.UL m,
.UL rd,
.UL rs,
.UL sides,
.UL td,
.UL ts
.ds LH Vertex Entities
.ds RH V
.bp
.SH
NAME
.LP
v - get or set the current vertex context
.SH
SYNOPSIS
.LP
.B v
[
.I id
[
.B =
[
.I template
]
]
]
.SH
DESCRIPTION
.LP
If the
.UL v
keyword is given by itself, then it establishes
the unnamed vertex context, which is the origin with no normal.
This context may be modified, but the changes will not be saved.
(The unnamed vertex is never used except as a source of default
values since all geometric entities call their vertices by name.)\0
.LP
If the keyword is followed by an identifier
.I id,
then it reestablishes a previous context.
If the specified context was never defined, an error will result.
.LP
If the entity is given with an identifier
followed by an equals sign ('='), then a new context is established,
and cleared to the default vertex (the origin).
(Note that the equals sign must be separated from other
arguments by white space to be properly recognized.)\0
If the equals sign is followed by a second identifier
.I template,
then this previously defined vertex will be used as a source of
default values instead.
This may be used to establish a vertex alias, or to modify an
existing vertex and give it a new name.
.LP
A non-zero vertex normal must be given for
certain entities, specifically
.UL ring
and
.UL torus
require a normal direction.
An
.UL f
entity will interpolate vertex normals if given, and
use the polygon plane normal otherwise.
See the
.UL prism
entry for an explanation of how it interprets and uses vertex
normals.
The other entities ignore vertex normals if present.
.LP
The actual position and normal direction for a vertex is determined
at the time of use by a geometric entity.
Specifically, the transformation in effect at the time the vertex is
defined is irrelevant.
The only transformation that matters is the one that is applied to
the geometry itself.
This prevents double-transformation of vertices and allows one set
of vertices to be used for multiple purposes, e.g. the front and
back sides of a drawer.
.SH
EXAMPLE
.DS
# Make a capped cylinder
v end1 =
        p 0 0 0
        n 0 0 -1
v end2 =
        p 0 0 1
cyl end1 1.2 end2
# Forgot normal for end2
v end2
        n 0 0 1
ring end1 0 1.2
ring end2 0 1.2
.DE
.SH
SEE ALSO
.LP
.UL cone,
.UL cyl,
.UL f,
.UL n,
.UL p,
.UL prism,
.UL ring,
.UL sph,
.UL torus
.ds RH P
.bp
.SH
NAME
.LP
p - set the point location for the current vertex
.SH
SYNOPSIS
.LP
.B p
.I "px py pz"
.SH
DESCRIPTION
.LP
Set the 3-dimensional position for the current vertex to
.I (px,py,pz).
The actual position of the vertex will be determined by the
transformation in effect at the time the vertex is applied to a
geometric surface entity.
The transform current when the position is set is irrelevant.
.LP
The default vertex position is the origin, (0,0,0).
.SH
EXAMPLE
.DS
# Make a small circle of 6 spheres
v scent =
        p 1 0 0
xf -a 6 -rz 60
        sph scent .05
xf
.DE
.SH
SEE ALSO
.LP
.UL cone,
.UL cyl,
.UL f,
.UL n,
.UL prism,
.UL ring,
.UL sph,
.UL torus,
.UL v
.ds RH N
.bp
.SH
NAME
.LP
n - set the surface normal direction for the current vertex
.SH
SYNOPSIS
.LP
.B n
.I "dx dy dz"
.SH
DESCRIPTION
.LP
Set the 3-dimensional surface normal for the current vertex to the
normalized vector along
.I (dx,dy,dz).
If this vector is zero, then the surface normal is effectively
unset.
The actual surface normal orientation of the vertex will be determined
by the transformation in effect at the time the vertex is applied to a
geometric surface entity.
The current transform when the normal is set is irrelevant.
.LP
The default vertex normal is the zero vector (i.e. no normal).
.SH
EXAMPLE
.DS
# Make a chain of 10 interlocking doughnuts
v tcent =
        p 0 0 0
        n 0 1 0
xf -a 10 -rx 90 -t .2 0 0
        torus tcent .1 .2
xf
.DE
.SH
SEE ALSO
.LP
.UL f,
.UL p,
.UL prism,
.UL ring,
.UL torus,
.UL v
.ds LH Geometric Entities
.ds RH F
.bp
.SH
NAME
.LP
f - create an N-sided polygonal face
.SH
SYNOPSIS
.LP
.B f
.I "v1 v2 ... vN"
.SH
DESCRIPTION
.LP
Create a polygonal face made of the current material
by connecting the named vertices in order, and connecting the last
vertex to the first.
There must be at least three vertices, and if any vertex is undefined,
an error will result.
.LP
The surface orientation is determined by the right-hand rule; when
the curl of the fingers follows the given order of the vertices, the
surface normal points in the thumb direction.
Face vertices should be coplanar, though this is difficult to guarantee
in a 3-dimensional specification.
.LP
If any vertices have associated surface normals, they will be used
instead of the average plane normal, though it is safest to specify
either all normals or no normals, and to stick with triangles
when normals are used.
Also, specified normals should point in the general direction of the
surface for best results.
.LP
There is no explicit representation of holes in MGF.  A hole must be
represented implicitly by connecting vertices to form "seams."  For
example, a wall with a window in it might look as shown in Figure 1.
In many systems, the wall itself would be represented with the first
list of vertices, (v1,v2,v3,v4) and the hole associated with that
wall as a second set of vertices (v5,v6,v7,v8).  In MGF, we must
give the whole thing as a single polygon, connecting the vertices so
as to create a "seam," as shown in Figure 2.
This could be written in MGF as "f v1 v2 v3 v4 v5 v6 v7 v8 v5 v4".
.LP
It is very important that the order of the hole be opposite to the
order of the outer perimeter, otherwise the polygon will be
"twisted" on top of itself.  Note also that the seam was traversed
in both directions, once going from v4 to v5, and again returning
from v5 to v4.  This is a necessary condition for a proper seam.
.LP
The choice of vertices to make into a seam is somewhat arbitrary, but
some rendering systems may not give sane results if you cross over a
hole with part of your seam.  If we had chosen to create the seam
between v2 and v5 in the above example instead of v4 and v5, the seam
would cross our hole and may not render correctly\(dg.
.FS
\(dgFor systems that
are sensitive to this, it is probably safest for their MGF
loader/translator to re-expresses seams in terms of holes again, which can
be done easily so long as vertices are shared in the fashion shown.
.FE
.bp
Replace this page with the first page from "figures.ps".
.bp
.SH
EXAMPLE
.DS
# Make a pyramid
v apex =
        p 1 1 1
v base0 =
        p 0 0 0
v base1 =
        p 0 2 0
v base2 =
        p 2 2 0
v base3 =
        p 2 0 0
# Bottom
f base0 base1 base2 base3
# Sides
f base0 apex base1
f base1 apex base2
f base2 apex base3
f base3 apex base0
.DE
.SH
SEE ALSO
.LP
.UL cone,
.UL cyl,
.UL m,
.UL prism,
.UL ring,
.UL sph,
.UL torus,
.UL v
.ds RH SPH
.bp
.SH
NAME
.LP
sph - create a sphere
.SH
SYNOPSIS
.LP
.B sph
.I "vc rad"
.SH
DESCRIPTION
.LP
Create a sphere made of the current material with its center at the
named vertex
.I vc
and a radius of
.I rad.
If the vertex is undefined an error will result.
.LP
The surface normal is usually directed outward, but will be directed
inward if the given radius is negative.
(This typically matters only for one-sided materials.)\0
A zero radius is illegal.
.SH
EXAMPLE
.DS
# Create a thick glass sphere with a hollow inside
m glass =
        sides 1
        ir 1.52 0
        c
        rs .06 0
        ts .88 0
v cent =
        p 0 0 1.1
# The outer shell
sph cent .1
# The inner bubble
sph cent -.08
.DE
.SH
SEE ALSO
.LP
.UL cone,
.UL cyl,
.UL f,
.UL m,
.UL prism,
.UL ring,
.UL torus,
.UL v
.ds RH CYL
.bp
.SH
NAME
.LP
cyl - create an open-ended, truncated right cylinder
.SH
SYNOPSIS
.LP
.B cyl
.I "v1 rad v2"
.SH
DESCRIPTION
.LP
Create a truncated right cylinder of radius
.I rad
using the current material, starting at the named vertex
.I v1
and continuing to
.I v2.
The ends will be open, but may be capped using the
.UL ring
entity if desired.
.LP
The surface normal will usually be directed outward, but may be
directed inward by giving a negative value for
.I rad.
A zero radius is illegal, and
.I v1
cannot equal
.I v2.
.SH
EXAMPLE
.DS
# A stylus with one rounded and one pointed end
o stylus
        v vtip0 =
                p 0 0 0
        v vtip1 =
                p 0 0 .005
        v vend =
                p 0 0 .05
        cyl vtip1 .0015 vend
        sph vend .0015
        cone vtip0 0 vtip1 .0015
o
.DE
.SH
SEE ALSO
.LP
.UL cone,
.UL f,
.UL m,
.UL prism,
.UL ring,
.UL sph,
.UL torus,
.UL v
.ds RH CONE
.bp
.SH
NAME
.LP
cone - create an open-ended, truncated right cone
.SH
SYNOPSIS
.LP
.B cone
.I "v1 rad1 v2 rad2"
.SH
DESCRIPTION
.LP
Create a truncated right cone using the current material.
The starting radius is
.I rad1
at
.I v1
and the ending radius is
is
.I rad2
at
.I v2.
The ends will be open, but may be capped using the
.UL ring
entity if desired.
.LP
The surface normal will usually be directed outward, but may be
directed inward by giving negative values for both radii.
(It is illegal for the signs of the two radii to disagree.)\0
One but not both radii may be zero, indicating that the cone comes
to a point.
.LP
Although it is not strictly forbidden to have equal cone radii, the
.UL cyl
entity should be used in such cases.
Likewise, the
.UL ring
entity must be used if
.I v1
and
.I v2
are equal.
.SH
EXAMPLE
.DS
# A parasol
o parasol
        v v1 =
                p 0 0 0
        v v2 =
                p 0 0 .75
        v v3 =
                p 0 0 .7
        m handle_mat
        cyl v1 .002 v2
        m parasol_paper
        cyl v2 0 v3 .33
o
.DE
.SH
SEE ALSO
.LP
.UL cyl,
.UL f,
.UL m,
.UL prism,
.UL ring,
.UL sph,
.UL torus,
.UL v
.ds RH PRISM
.bp
.SH
NAME
.LP
prism - create a closed right prism
.SH
SYNOPSIS
.LP
.B prism
.I "v1 v2 ... vN length"
.SH
DESCRIPTION
.LP
Create a closed right prism using the current material.
One end face will be enclosed by the named vertices, and the
opposite end face will be a mirror image at a distance
.I length
from the original.
The edges will be extruded into N quadrilaterals connecting
the two end faces.
.LP
The order of vertices determines the original face orientation
according to the right-hand rule as explained for the
.UL f
entity.
Normally, the prism is extruded in the direction opposite to the
original surface normal, resulting in faces that all point outward.
If the specified
.I length
is negative, the prism will be extruded above the original face
and all surface normals will point inward.
.LP
If the vertices have associated normals, they are applied to the
side faces only, and should generally point in the appropriate
direction (i.e. in or out depending on whether
.I length
is negative or positive).
.SH
EXAMPLE
.DS
# Make a unit cube starting at the origin and \\\\
        extending to the positive octant
v cv0 =
        p 0 0 0
v cv1 =
        p 0 1 0
v cv2 =
        p 1 1 0
v cv3 =
        p 1 0 0
# Right hand rule has original face looking in -Z direction
prism cv0 cv1 cv2 cv3 1
.DE
.SH
SEE ALSO
.LP
.UL cyl,
.UL cone,
.UL f,
.UL m,
.UL ring,
.UL sph,
.UL torus,
.UL v
.ds RH RING
.bp
.SH
NAME
.LP
ring - create a circular ring with inner and outer radii
.SH
SYNOPSIS
.LP
.B ring
.I "vc rmin rmax"
.SH
DESCRIPTION
.LP
Create a circular face of the current material centered on the named
vertex
.I vc
with an inner radius of
.I rmin
and an outer radius of
.I rmax.
The surface orientation is determined by the normal vector
associated with
.I vc.
If this vertex is undefined or has no normal, an error will result.
The minimum radius may be equal to but not less than zero, and the
maximum radius must be strictly greater than the minimum.
.SH
EXAMPLE
.DS
# The proverbial brass ring
o brass_ring
        m brass
        v end1 =
                p 0 -.005 0
                n 0 -1 0
        v end2 =
                p 0 .005 0
                n 0 1 0
        ring end1 .02 .03
        cyl end1 .03 end2
        ring end2 .02 .03
        cyl end2 -.02 end1
o
.DE
.SH
SEE ALSO
.LP
.UL cyl,
.UL cone,
.UL f,
.UL m,
.UL prism,
.UL sph,
.UL torus,
.UL v
.ds RH TORUS
.bp
.SH
NAME
.LP
torus - create a regular torus
.SH
SYNOPSIS
.LP
.B torus
.I "vc rmin rmax"
.SH
DESCRIPTION
.LP
Create a torus of the current material centered on the named vertex
.I vc
with an inner radius of
.I rmin
and an outer radius of
.I rmax.
The plane of the torus will be perpendicular to the normal vector
associated with
.I vc.
If this vertex is undefined or has no normal, an error will result.
.LP
If a torus with an inward facing surface normal is desired,
.I rmin
and
.I rmax
may be negative.
The minimum radius may be zero, but may not be negative when
.I rmax
is positive or vice versa.
The magnitude or
.I rmax
must always be strictly greater than that of
.I rmin.
.SH
EXAMPLE
.DS
# The proverbial brass ring -- easy grip version
o brass_ring
        m brass
        v center =
                p 0 0 0
                n 0 1 0
        torus center .02 .03
o
.DE
.SH
SEE ALSO
.LP
.UL cyl,
.UL cone,
.UL f,
.UL m,
.UL prism,
.UL ring,
.UL sph,
.UL v
.ds RH
.ds LH
.bp
.NH
MGF Translators
.LP
Initially, there are four translators for MGF data, but only
one of these is distributed with the MGF parser itself,
.I mgfilt.
Two of the other translators,
.I mgf2rad
and
.I rad2mgf
convert between MGF and the Radiance scene description language,
and are distributed for free with the rest of the Radiance
package\(dg.
.FS
\(dgRadiance is available by anonymous ftp from hobbes.lbl.gov and
nestor.epfl.ch, or by WWW from
"http://radsite.lbl.gov/radiance/HOME.html"
.FE
A third translator,
.I mgf2meta,
converts to a 2-dimensional line plot, and is also
distributed with Radiance.
.LP
Mgfilt is a simple but useful utility that takes MGF on its input
and produces MGF on its output.
It uses the parser to convert entities that are not wanted or
understood, and produces only the requested ones.
This is useful for seeing what exactly a program must understand
when it supports a given set of entities, and may serve as a
substitute for linking to the parser library for programmers who
wish to interpret the ASCII input directly but without all the
unwanted entities.
In future releases of MGF, this utility will also be handy for
taking new entities and producing older versions of MGF for
translators that have not yet been updated properly.
.ds LH Translators
.ds RH MGFILT
.bp
.SH
NAME
.LP
mgfilt - get usable MGF entities from input
.SH
SYNOPSIS
.LP
.B mgfilt
.B version
[
.B input ..
]
.br
or
.br
.B mgfilt
.B "e1,e2,.."
[
.B input ..
]
.SH
DESCRIPTION
.LP
.I Mgfilt
takes one or more MGF input files and converts all the entities to
the types listed.
In the first form, a single integer is given for the
.I version
of MGF that is to be produced.
Since MGF is in its first major release, this is not yet a useful
form, but it will be when the second major release comes out.
This has the necessary side-effect of expanding all included files.
(See the
.UL i
entity.)\0
.LP
In the second form,
.I mgfilt
produces only the entities listed in the first argument, which must
be comma-separated.
The listed entity order is not important, but all entities given
must be defined in the current version of MGF.
Unknown entities will be summarily discarded on the input, and a
warning message will be printed to the standard error.
.SH
EXAMPLES
.LP
To take an MGF version 3 file and send it to a version 2
translator:
.IP
mgfilt 2 input.mgf | mgf2rad > input.rad
.LP
To take an MGF file and produce only flat polygonal faces
with no materials:
.IP
mgfilt f,v,p,xf input.mgf > flatpoly.mgf
.SH
SEE ALSO
.LP
i, mgf2rad, rad2mgf
.ds RH MGF2RAD
.bp
.SH
NAME
.LP
mgf2rad - convert Materials and Geometry Format file to RADIANCE description
.SH
SYNOPSIS
.LP
.B mgf2rad
[
.B "\-m matfile"
][
.B "\-e mult"
][
.B "\-g dist"
]
[
.B input ..
]
.SH
DESCRIPTION
.LP
.I Mgf2rad
converts one or more Materials and Geometry Format (MGF)
files to a RADIANCE scene description.
By definition, all output dimensions are in meters.
The material names and properties
for the surfaces will be those assigned in MGF.
Any materials not defined in MGF will result in an error during
translation.
Light sources are described inline as IES luminaire files, and
.I mgf2rad
calls the program
.I ies2rad(1)
to translate these files.
If an IES file in turn contains an MGF description of the local
fixture geometry, this may result in a recursive call to
.I mgf2rad,
which is normal and should be transparent.
The only side-effect of this additional translation is the
appearance of other RADIANCE scene and data files produced
automatically by
.I ies2rad.
.LP
The
.I \-m
option may be used to put all the translated materials into a separate
RADIANCE file.
This is not always advisable, as any given material name may be
reused at different points in the MGF description, and writing them
to a separate file loses the contextual association between
materials and surfaces.
As long as unique material names are used throughout the MGF
description and material properties are not redefined, there
will be no problem.
Note that this is the only way to get all the translated materials
into a single file, since no output is produced for unreferenced
materials; i.e. translating just the MGF materials does not work.
.LP
The
.I \-e
option may be used to multiply all the emission values by the
given
.I mult
factor.
The
.I \-g
option may be used to establish a glow distance (in meters)
for all emitting surfaces.
These two options are employed principally by
.I ies2rad,
and are not generally useful to most users.
.SH
EXAMPLES
.LP
To translate two MGF files into one RADIANCE materials file and
one geometry file:
.IP
mgf2rad -m materials.rad building1.mgf building2.mgf > building1+2.rad
.LP
To create an octree directly from two MGF files and one RADIANCE
file:
.IP
oconv '\\!mgf2rad materials.mgf scene.mgf' source.rad > scene.oct
.SH
FILES
.LP
tmesh.cal       Used to smooth polygonal geometry
.br
*.rad           RADIANCE source descriptions created by ies2rad
.br
*.dat           RADIANCE source data created by ies2rad
.br
source.cal      Used for IES source coordinates
.SH
AUTHOR
.LP
Greg Ward
.SH
SEE ALSO
.LP
ies2rad(1), mgf2meta(1), obj2rad(1), oconv(1), rad2mgf(1), xform(1)
.ds RH RAD2MGF
.bp
.SH
NAME
.LP
rad2mgf - convert RADIANCE scene description to Materials and Geometry Format
.SH
SYNOPSIS
.LP
.B rad2mgf
[
.B \-dU
]
[
.B input ..
]
.SH
DESCRIPTION
.LP
.I Rad2mgf
converts one or more RADIANCE scene files
to the Materials and Geometry Format (MGF).
Input units are specified with the
.I \-mU
option, where
.I U
is one of 'm' (meters), 'c' (centimeters), 'f' (feet) or 'i'
(inches).
The assumed unit is meters, which is the required output unit for
MGF (thus the need to know).
If the input dimensions are in none of these units, then the user
should apply
.I xform(1)
with the
.I \-s
option to bring the units into line prior to translation.
.LP
The MGF material names and properties
for the surfaces will be those assigned in RADIANCE.
If a referenced material has not been defined, then its name will
be invoked in the MGF output without definition, and the description
will be incomplete.
.SH
LIMITATIONS
.LP
Although MGF supports all of the geometric types and the most
common material types used in RADIANCE, there is currently no
support for advanced BRDF materials, patterns, textures or mixtures.
Also, the special types "source" and "antimatter" are not supported,
and all light source materials are converted to simple diffuse emitters
(except "illum" materials, which are converted to their alternates).
These primitives are reproduced as comments in the output and
must be replaced manually if necessary.
.LP
The RADIANCE "instance" type is treated specially.
.I Rad2mgf
converts each instance to an MGF include statement, using the corresponding
transformation and a file name derived from the octree name.
(The original octree suffix is replaced by ".mgf".)\0
For this to work, the user must separately create the referenced
MGF files from the original RADIANCE descriptions.
The description file names can usually be determined using the
.I getinfo(1)
command run on the octrees in question.
.SH
EXAMPLES
.LP
To convert three RADIANCE files (in feet) to one MGF file:
.IP
mgf2rad -df file1.rad file2.rad file3.rad > scene.mgf
.LP
To translate a RADIANCE materials file to MGF:
.IP
mgf2rad materials.rad > materials.mgf
.SH
AUTHOR
.LP
Greg Ward
.SH
SEE ALSO
.LP
getinfo(1), ies2rad(1), mgf2meta(1), mgf2rad(1), obj2rad(1), oconv(1), xform(1)
.ds RH MGF2META
.bp
.SH
NAME
.LP
mgf2meta - convert Materials and Geometry Format file to Metafile graphics
.SH
SYNOPSIS
.LP
.B mgf2meta
[
.B "-t threshold"
]
.B "{x|y|z} xmin xmax ymin ymax zmin zmax"
[
.B input ..
]
.SH
DESCRIPTION
.LP
.I Mgf2meta
converts one or more Materials and Geometry Format (MGF)
files to a 2-D orthographic projection along the selected axis in the
.I metafile(1)
graphics format.
All geometry is clipped to the specified bounding box, and the
resulting orientation is as follows:
.sp .5
.nf
        Projection      Orientation
        ======= ========
        x               Y-axis right, Z-axis up
        y               Z-axis right, X-axis up
        z               X-axis right, Z-axis up
.fi
.LP
If multiple input files are given, the first file prints in black,
the second prints in red, the third in green and the fourth in blue.
If more than four input files are given, they cycle through the
colors again in three other line types:  dashed, dotted and
dot-dashed.
.LP
The
.I \-t
option may be used to randomly throw out line segments that are
shorter than the given
.I threshold
(given as a fraction of the plot width).
Segments are included with a
probability equal to the square of the line length over the square
of the threshold.
This can greatly reduce the number of lines in the drawing (and
therefore improve the drawing speed) with only a modest loss in
quality.
A typical value for this parameter is 0.005.
.LP
All MGF material information is ignored on the input.
.SH
EXAMPLES
.LP
To project two MGF files along the Z-axis and display them under
X11:
.IP
mgf2meta z 0 10 0 15 0 9 building1.mgf building2.mgf | x11meta
.LP
To convert a RADIANCE scene to a line drawing in RADIANCE picture
format:
.IP
rad2mgf scene.rad | mgf2meta x `getbbox -h scene.rad` | meta2tga |
ra_t8 -r > scene.pic
.SH
AUTHOR
.LP
Greg Ward
.SH
SEE ALSO
.LP
getbbox(1), meta2tga(1), metafile(5), mgf2rad(1), pflip(1),
protate(1), psmeta(1), ra_t8(1), rad2mgf(1), t4014(1), x11meta(1)
.ds RH
.ds LH
.bp
.NH
MGF Parser Library
.LP
The principal motivation for creating a standard parser library for
MGF is to make it easy for software developers to offer some base
level of compliance.
The key to making MGF easy to support in fact is the parser, which
has the ability to express higher order entities in terms of
lower order ones.
For example, tori are part of the MGF specification, but if a given
program or translator does not support them, the parser will convert
them to cones.
If cones are not supported either, it will convert them further into
smoothed polygons.
If smoothing (vertex normal information) is not supported, it will
be ignored and the program will just get flat polygons.
This is done in such a way that future versions of the standard may
include new entities that old software does not even have to know
about, and they will be converted appropriately.
Forward compatibility is thus built right into the parser loading
mechanism itself -- the programmer simply links to the new code and
the new standard is supported without any further changes.
.SH
Language
.LP
The provided MGF parser is written in ANSI-C.
This language was chosen for reasons of portability and efficiency.
Almost all systems support some form of ANSI-compatible C, and many
languages can cross-link to C libraries without modification.
Backward compatibility to Kernighan and Ritchie C is achieved by
compiling with the -DNOPROTO flag.
.LP
All of the data structures and prototypes needed for the library
are in the header file "parser.h".
This file is the best resource for the parser and is updated with
each MGF release.
.SH
Mechanism
.LP
The parser works by a simple callback mechanism to routines that
actually interpret the individual entities.
Some of these routines will belong to the calling program, and some
will be entity support routines included in the library itself.
There is a global array of function pointers, called
.I mg_ehand.
It is defined thus:
.DS
extern int      (*mg_ehand[MG_NENTITIES])(int argc, char **argv);
.DE
Before parsing begins, this dispatch table is initialized to point to the
routines that will handle each supported entity.
Every entity handler has the same basic prototype, which is the
same as the
.I main
function, i.e:
.DS
extern int      \f2handler\f1(int argc, char **argv);
.DE
The first argument is the number of words in the MGF entity
(counting the entity itself) and the second argument is an array of
nul-terminated strings with the entity and its arguments.
The function should return zero or one of the error
codes defined in "parser.h".
A non-zero return value causes the parser to abort, returning the
error up through its call stack to the entry function, usually
.I mg_load.
.LP
A special function pointer for undefined entities is
defined as follows:
.DS
extern int      (*mg_uhand)(int argc, char **argv);
.DE
By default, this points to the library function
.I mg_defuhand,
which prints an error message on the first unknown entity and keeps a
count from then on, which is stored in the global unsigned integer
.I mg_nunknown.
If the
.I mg_uhand
pointer is assigned a value of NULL instead, parsing will abort at the
first unrecognized entity.
The reason this is not the default action is that ignoring unknown entities
offers a certain base level of forward compatibility.
Ignoring things one does not understand is not the best approach, but it
is usually better than quitting with an error message if the input is
in fact valid, but is a later version of the standard.
The real solution is to update the interpreter by linking to a new version
of the parser, or use a new version of the
.I mgfilt
command to convert the new MGF input to an older standard.
.LP
The
.I mg_uhand
pointer may also be used to customize the language for a particular
application by adding entities, though this is discouraged because it
tends to weaken the standard.
.LP
The skeletal framework for an MGF loader or translator is to assign
function pointers to the
.I mg_ehand
array, call the parser initialization function
.I mg_init,
then call the file loader function
.I mg_load
once for each input file.
This will in turn make calls back to the functions assigned to
.I mg_ehand.
To give a simple example, let us look at a
translator that understands only flat polygonal faces, putting out
vertex locations immediately after each "face" keyword:
.DS
#include <stdio.h>
#include "parser.h"

int
myfaceh(ac, av)                 /* face handling routine */
int     ac;
char    **av;
{
        C_VERTEX        *vp;    /* vertex structure pointer */
        FVECT   vert;           /* vertex point location */
        int     i;

        if (ac < 4)                     /* check # arguments */
                return(MG_EARGC);
        printf("face\\\\n");            /* begin face output */
        for (i = 1; i < ac; i++) {
                if ((vp = c_getvert(av[i])) == NULL)    /* vertex from name */
                        return(MG_EUNDEF);
                xf_xfmpoint(vert, vp->p);                       /* apply transform */
                printf("%15.9f %15.9f %15.9f\\\\n",
                        vert[0], vert[1], vert[2]);                     /* output vertex */
        }
        printf(";\\\\n");                       /* end of face output */
        return(MG_OK);                  /* normal exit */
}

main(argc, argv)                /* translate MGF file(s) */
int     argc;
char    **argv;
{
        int     i;
                                        /* initialize dispatch table */
        mg_ehand[MG_E_FACE] = myfaceh;          /* ours */
        mg_ehand[MG_E_VERTEX] = c_hvertex;              /* parser lib */
        mg_ehand[MG_E_POINT] = c_hvertex;               /* parser lib */
        mg_ehand[MG_E_XF] = xf_handler;         /* parser lib */
        mg_init();                              /* initialize parser */
        for (i = 1; i < argc; i++)              /* load each file argument */
                if (mg_load(argv[i]) != MG_OK)  /* and check for error */
                        exit(1);
        exit(0);                        /* all done! */
}
.DE
Hopefully, this example demonstrates just how easy it is to
write an MGF translator.
Of course, translators get more complicated the more entity
types they support, but the point is that one does not
.I have
to support every entity -- the parser handles what the translator
does not.
Also, the library includes many general entity handlers,
further reducing the burden on the programmer.
This same principle means that it is not necessary to modify an
existing program to accommodate a new version of MGF -- one need only
link to the new parser library to comply with the new standard.
.SH
Division of Labor
.LP
As seen in the previous example, there are two parser routines that
are normally called directly in an MGF translator or loader program.
The first is
.I mg_init,
which takes no arguments but relies on the program having
initialized those parts of the global
.I mg_ehand
array it cares about.
The second routine is
.I mg_load,
which is called once on each input file.
(A third routine,
.I mg_clear,
may be called to free the parser data structures after each file or
after all files, if the program plans to continue rather than
exit.)\0
.LP
The rest of the routines in a translator or loader program are
called indirectly through the
.I mg_ehand
dispatch table, and they are the ones that do the real work of
supporting the MGF entities.
In addition to converting or discarding entities that the calling
program does not know or care about, the parser library includes a
set of context handlers that greatly simplify the translation
process.
There are three handlers for each of the three named contexts and
their constituents, and two handlers for the two hierarchical
context entities.
To use these handlers, one simply sets the appropriate positions in the
.I mg_ehand
dispatch table to point to these functions.
Additional functions and global data structures provide convenient
access to the relevant contexts, and all of these are detailed in
the following manual pages.
.ds LH Basic Parser Routines
.ds RH MG_INIT
.bp
.SH
NAME
.LP
mg_init, mg_ehand, mg_uhand - initialize MGF entity handlers
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B void
mg_init(
.B void
)
.LP
.B int
mg_defuhand(
.B int
argc,
.B char
**argv )
.LP
.B "extern int"
(*mg_ehand[MG_NENTITIES])(
.B int
argc,
.B char
**argv )
.LP
.B "extern int"
(*mg_uhand)(
.B int
argc,
.B char
**argv )
.LP
.B "extern unsigned"
mg_nunknown
.SH
DESCRIPTION
.LP
The parser dispatch table,
.I mg_ehand
is initially set to all NULL pointers, and it
is the duty of the calling program to assign entity handler functions to
each of the supported entity positions in the array.
The entities are given in the include file "parser.h" as the
following:
.DS
#define MG_E_COMMENT    0               /* #            */
#define MG_E_COLOR              1               /* c            */
#define MG_E_CCT                2               /* cct          */
#define MG_E_CONE               3               /* cone */
#define MG_E_CMIX               4               /* cmix */
#define MG_E_CSPEC              5               /* cspec        */
#define MG_E_CXY                6               /* cxy          */
#define MG_E_CYL                7               /* cyl          */
#define MG_E_ED         8               /* ed           */
#define MG_E_FACE               9               /* f            */
#define MG_E_INCLUDE    10              /* i            */
#define MG_E_IES                11              /* ies          */
#define MG_E_IR                 12              /* ir           */
#define MG_E_MATERIAL   13              /* m            */
#define MG_E_NORMAL     14              /* n            */
#define MG_E_OBJECT     15              /* o            */
#define MG_E_POINT              16              /* p            */
#define MG_E_PRISM              17              /* prism        */
#define MG_E_RD         18              /* rd           */
#define MG_E_RING               19              /* ring */
#define MG_E_RS         20              /* rs           */
#define MG_E_SIDES              21              /* sides        */
#define MG_E_SPH                22              /* sph  */
#define MG_E_TD         23              /* td           */
#define MG_E_TORUS              24              /* torus        */
#define MG_E_TS         25              /* ts           */
#define MG_E_VERTEX     26              /* v            */
#define MG_E_XF         27              /* xf           */

#define MG_NENTITIES    28              /* total # entities */
.DE
.LP
Once the
.I mg_ehand
array has been set by the program, the
.I mg_init
routine must be called to complete the initialization process.
This should be done once and only once per invocation, before any other
parser routines are called.
.LP
The
.I mg_uhand
variable points to the current handler for unknown entities
encountered on the input.
Its default value points to the
.I mg_defuhand
function, which simply increments the global variable
.I mg_nunknown,
printing a warning message on the standard error on the first
offense.
(This message may be avoided by incrementing
.I mg_nunknown
before processing begins.)\0
If
.I mg_uhand
is assigned a value of NULL, then an unknown entity will return an
.I MG_EUNK
error, which will cause the parser to abort.
(See the
.I mg_load
page for a list of errors.)\0
If the
.I mg_uhand
pointer is assigned to another function, that function will receive
any unknown entities and their arguments, and the parsing will
abort if the new function returns a non-zero error value.
This offers a convenient way to customize the language by adding
non-standard entities.
.SH
DIAGNOSTICS
.LP
If an inconsistent set of entities has been set for support, the
.I mg_init
routine will print an informative message to standard error and abort
the calling program with a call to
.I exit.
This is normally unacceptable behavior for a library routine, but since
such an error indicates a fault with the calling program itself,
recovery is impossible.
.SH
SEE ALSO
.LP
mg_load, mg_handle
.ds RH MG_LOAD
.bp
.SH
NAME
.LP
mg_load, mg_clear, mg_file, mg_err - load MGF file, clear data structures
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B int
mg_load(
.B char
*filename )
.LP
.B void
mg_clear(
.B void
)
.LP
.B extern
MG_FCTXT *mg_file
.LP
.B "extern char"
*mg_err[MG_NERRS]
.SH
DESCRIPTION
.LP
The
.I mg_load
function loads the named file, or standard input if
.I filename
is the NULL pointer.
Calls back to the appropriate MGF handler routines are made through the
.I mg_ehand
dispatch table.
.LP
The global
.I mg_file
variable points to the current file context structure, which
may be useful for the interpretation of certain entities, such as
.UL ies,
which must know the directory path of the enclosing file.
This structure is of the defined type
.I MG_FCTXT,
given in "parser.h" as:
.DS
typedef struct mg_fctxt {
        char    fname[96];                      /* file name */
        FILE    *fp;                            /* stream pointer */
        int     fid;                            /* unique file context id */
        char    inpline[4096];          /* input line */
        int     lineno;                 /* line number */
        struct mg_fctxt *prev;  /* previous context */
} MG_FCTXT;
.DE
.SH
DIAGNOSTICS
.LP
If an error is encountered during parsing,
.I mg_load
will print an appropriate error message to the standard error stream
and return one of the non-zero values from "parser.h" listed below:
.DS
#define MG_OK                   0               /* normal return value */
#define MG_EUNK         1               /* unknown entity */
#define MG_EARGC                2               /* wrong number of arguments */
#define MG_ETYPE                3               /* argument type error */
#define MG_EILL                         4               /* illegal argument value */
#define MG_EUNDEF               5               /* undefined reference */
#define MG_ENOFILE              6               /* cannot open input file */
#define MG_EINCL                7               /* error in included file */
#define MG_EMEM         8               /* out of memory */
#define MG_ESEEK                9               /* file seek error */
#define MG_EBADMAT      10              /* bad material specification */
#define MG_ELINE                11              /* input line too long */

#define MG_NERRS        12
.DE
If it is inappropriate to send output to standard error, the calling
program should use the routines listed under
.I mg_open
for better control over the parsing process.
.LP
The
.I mg_err
array contains error messages corresponding to each of the values
listed above in the native country's language.
.SH
SEE ALSO
.LP
mg_fgetpos, mg_handle, mg_init, mg_open
.ds RH MG_OPEN
.bp
.SH
NAME
.LP
mg_open, mg_read, mg_parse, mg_close - MGF file loading subroutines
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B int
mg_open( MG_FCTXT *fcp,
.B char
*filename )
.LP
.B int
mg_read(
.B void
)
.LP
.B int
mg_parse(
.B void
)
.LP
.B void
mg_close(
.B void
)
.SH
DESCRIPTION
.LP
Most loaders and translators will call the
.I mg_load
routine to handle the above operations, but some programs or
entity handlers require tighter control over the loading process.
.LP
The
.I mg_open
routine takes an uninitialized
.I MG_FCTXT
structure and a file name as its arguments.
If
.I filename
is the NULL pointer, the standard input is "opened."
The
.I fcp
structure will be set by
.I mg_open
prior to its return, and the global
.I mg_file
pointer will be assigned to point to it.
This variable must not be destroyed until after the file is closed
with a call to
.I mg_close.
(See the
.I mg_load
page for a definition of
.I mg_file
and the
.I MG_FCTXT
type.)\0
.LP
The
.I mg_read
function reads the next input line from the current file,
returning the number of characters in the line, or zero if the
end of file is reached or there is a file error.
If the last character read in the input line is not a newline,
then the input line was too long, and you
should return an MG_ELINE error.
The function skips over escaped newlines, and keeps track of the
line number in the current file context
.I mg_file,
which also contains the line that was read.
.LP
The
.I mg_parse
function breaks the current line in the
.I mg_file
structure into words and calls the appropriate handler routine, if
any.
Blank lines and unsupported entities cause a quick return.
.LP
The
.I mg_close
routine closes the current input file (unless it is the standard
input) and returns to the previous file context (if any).
.SH
DIAGNOSTICS
.LP
The
.I mg_open
function returns
.I MG_OK
(0) normally, or
.I MG_ENOFILE
if the open fails for some reason.
.LP
The
.I mg_parse
function returns
.I MG_OK
if the current line was successfully interpreted, or one of the
defined error values if there is a problem.
(See the
.I mg_load
page for the defined error values.)\0
.SH
SEE ALSO
.LP
mg_fgetpos, mg_handle, mg_init, mg_load
.ds RH MG_FGETPOS
.bp
.SH
NAME
.LP
mg_fgetpos, mg_fgoto - get current file position and seek to pointer
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B void
mg_fgetpos( MG_FPOS *pos )
.LP
.B int
mg_fgoto( MG_FPOS *pos )
.SH
DESCRIPTION
.LP
The
.I mg_fgetpos
gets the current MGF file position and loads it into the passed
.I MG_FPOS
structure,
.I pos.
.LP
The
.I mg_fgoto
function seeks to the position
.I pos,
taken from a previous call to
.I mg_fgetpos.
.SH
DIAGNOSTICS
.LP
If
.I mg_fgoto
is passed an illegal pointer or one that does not correspond to the
current
.I mg_file
context, it will return the
.I MG_ESEEK
error value.
Normally, it returns
.I MG_OK
(0).
.SH
SEE ALSO
.LP
mg_load, mg_open
.ds RH MG_HANDLE
.bp
.SH
NAME
.LP
mg_handle, mg_entity, mg_ename, mg_nqcdivs - entity assistance and control
.SH
SYNOPSIS
.LP
.B int
mg_handle(
.B int
en,
.B int
ac,
.B char
*av )
.LP
.B int
mg_entity(
.B char
*name )
.LP
.B "extern char"
mg_ename[MG_NENTITIES][MG_MAXELEN]
.LP
.B "extern int"
mg_nqcdivs
.SH
DESCRIPTION
.LP
The
.I mg_handle
routine may be used to pass entities back to the parser
to be redirected through the
.I mg_ehand
dispatch table.
This method is recommended rather than calling through
.I mg_ehand
directly, since the parser sometimes has its own support routines
that it needs to call for specific entities.
The first argument,
.I en,
is the corresponding entity number, or -1 if
.I mg_handle
should figure it out from the first
.I av
argument.
.LP
The
.I mg_entity
function gets an entity number from its name, using a hash
table on the
.I mg_ename
list.
.LP
The
.I mg_ename
table contains the string names corresponding to each MGF entity in
the designated order.
(See the
.I mg_init
page for the list of MGF entities.)\0
.LP
The global integer variable
.I mg_nqcdivs
tells the parser how many subdivisions to use per quarter circle (90
degrees) when tesselating curved geometry.
The default value is 5, and it may be reset at any time by the
calling program.
.SH
DIAGNOSTICS
.LP
The
.I mg_handle
function returns
.I MG_OK
if the entity is handled correctly, or one of the predefined error
values if there is a problem.
(See the
.I mg_load
page for a list of error values.)\0
.LP
The
.I mg_entity
function returns -1 if the passed name does not appear in the
.I mg_ename
list.
.SH
SEE ALSO
.LP
mg_init, mg_load, mg_open
.ds RH ISINT, ISFLT, ISNAME
.bp
.SH
NAME
.LP
isint, isflt, isname - determine if string fits integer or real format,
or is legal identifier
.SH
SYNOPSIS
.LP
.B int
isint(
.B char
*str )
.LP
.B int
isflt(
.B char
*str )
.LP
.B int
isname(
.B char
*str )
.SH
DESCRIPTION
.LP
The
.I isint
function checks to see if the passed string
.I str
matches a decimal integer format (positive or negative),
and returns 1 or 0 based on whether it does or does not.
.LP
The
.I isflt
function checks to see if the passed string
.I str
matches a floating point format (positive or negative with optional
exponent), and returns 1 or 0 based on whether it does or does not.
.LP
The
.I isname
function checks to see if the passed string
.I str
is a legal identifier name.
In MGF, a legal identifier must begin with a letter and contain only
visible ASCII characters (those between decimal 33 and 127 inclusive).
The one caveat to this is that names may begin with one or more
underscores ('_'), but this is a trick employed by the parser to
maintain a separate name space from the user, and is not legal usage
otherwise.
.LP
Note that a string that matches an integer format is also a valid
floating point value.
Conversely, a string that is not a floating point number cannot be a
valid integer.
.LP
These routines are useful for checking arguments passed to entity
handlers that certain types in certain positions.
If an invalid argument is passed, the handler should return an
.I MG_ETYPE
error.
.SH
SEE ALSO
.LP
mg_init, mg_load
.ds LH Entity Support Routines
.ds RH C_HVERTEX
.bp
.SH
NAME
.LP
c_hvertex, c_getvert, c_cvname, c_cvertex - vertex entity support
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B int
c_hvertex(
.B int
argc,
.B char
**argv )
.LP
C_VERTEX *c_getvert(
.B char
*name )
.LP
.B "extern char"
*c_vname
.LP
.B extern
C_VERTEX *c_cvertex
.SH
DESCRIPTION
.LP
The
.I c_hvertex
function handles the MGF vertex entities,
.UL v,
.UL p
and
.UL n.
If either
.UL p
or
.UL n
is supported, then
.UL v
must be also.
The assignments are normally made to the
.I mg_ehand
array prior to parser initialization, like so:
.DS
mg_ehand[MG_E_VERTEX] = c_hvertex;              /* support "v" entity */
mg_ehand[MG_E_POINT] = c_hvertex;               /* support "p" entity */
mg_ehand[MG_E_NORMAL] = c_hvertex;      /* support "n" entity */
/* other entity handler assignments... */
mg_init();                      /* initialize parser */
.DE
If vertex normals are not understood by any of the program-supported
entities, then the
.I MG_E_NORMAL
entry may be left with its original NULL assignment.
.LP
The
.I c_getvert
call takes the name of a defined vertex and returns a pointer to its
.I C_VERTEX
structure, defined in "parser.h" as:
.DS
typedef FLOAT  FVECT[3];        /* a 3-d real vector */

typedef struct {
        int     clock;                  /* incremented each change -- resettable */
        FVECT   p, n;           /* point and normal */
} C_VERTEX;             /* vertex context */
.DE
The
.I clock
member will be incremented each time the value gets changed by a
.UL p
or
.UL n
entity, and may be reset by the controlling program if desired.
This is a convenient way to keep track of whether or not a vertex has
changed since its last use.
To link identical vertices, one must also check that the current
transform has not changed, which is uniquely identified by the
global
.I xf_context->xid
variable, but only if one is using the parser libraries transform
handler.
(See the
.I xf_handler
page.)\0
.LP
It is possible but not recommended to alter the contents of the
vertex structure returned by
.I c_getvert.
Normally it is read during the
interpretation of entities using named vertices.
.LP
The name of the current vertex is given by the global
.I c_cvname
variable, which is set to NULL if the unnamed vertex is current.
The current vertex value is pointed to by the global variable
.I c_cvertex,
which should never be NULL.
.SH
DIAGNOSTICS
.LP
The
.I c_hvertex
function returns
.I MG_OK
(0) if the vertex is handled correctly, or one of the predefined
error values if there is a problem.
(See the
.I mg_load
page for a list of errors.)\0
.LP
The
.I c_getvert
function returns NULL if the specified vertex name is undefined, at
which point the calling function should return an
.I MG_EUNDEF
error.
.SH
SEE ALSO
.LP
c_hcolor, c_hmaterial, mg_init, mg_load, xf_handler
.ds RH C_HCOLOR
.bp
.SH
NAME
.LP
c_hcolor, c_getcolor, c_ccname, c_ccolor, c_ccvt, c_isgrey -
color entity support
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B int
c_hcolor(
.B int
argc,
.B char
**argv )
.LP
C_COLOR *c_getcolor(
.B char
*name )
.LP
.B "extern char"
*c_ccname
.LP
.B extern
C_COLOR *c_ccolor
.LP
.B void
c_ccvt( C_COLOR *cvp,
.B int
cflags )
.LP
.B int
c_isgrey( C_COLOR *cvp )
.SH
DESCRIPTION
.LP
The
.I c_hcolor
function supports the MGF entities,
.UL c,
.UL cxy,
.UL cspec,
.UL cct
and
.UL cmix.
It is an error to support any of the color field entities without
supporting the
.UL c
entity itself.
The assignments are normally made to the
.I mg_ehand
array prior to parser initialization, like so:
.DS
mg_ehand[MG_E_COLOR] = c_hcolor;        /* support "c" entity */
mg_ehand[MG_E_CXY] = c_hcolor;  /* support "cxy" entity */
mg_ehand[MG_E_CSPEC] = c_hcolor;        /* support "cspec" entity */
mg_ehand[MG_E_CCT] = c_hcolor;          /* support "cct" entity */
mg_ehand[MG_E_CMIX] = c_hcolor; /* support "cmix" entity */
/* other entity handler assignments... */
mg_init();                      /* initialize parser */
.DE
If the loader/translator has no use for spectral data, the entries for
.UL cspec
and
.UL cct
may be left with their original NULL assignments and these entities will
be re-expressed appropriately as tristimulus values.
.LP
The
.I c_getcolor
function takes the name of a defined color and returns a pointer to its
.I C_COLOR
structure, defined in "parser.h" as:
.DS
#define C_CMINWL        380                                     /* minimum wavelength */
#define C_CMAXWL        780                                     /* maximum wavelength */
#define C_CNSS          41                                      /* number of spectral samples */
#define C_CWLI          ((C_CMAXWL-C_CMINWL)/(C_CNSS-1))
#define C_CMAXV         10000                   /* nominal maximum sample value */
#define C_CLPWM         (683./C_CMAXV)  /* peak lumens/watt multiplier */

typedef struct {
        int     clock;                  /* incremented each change */
        short   flags;                  /* what's been set */
        short   ssamp[C_CNSS];  /* spectral samples, min wl to max */
        long    ssum;                   /* straight sum of spectral values */
        float   cx, cy;         /* xy chromaticity value */
        float   eff;                    /* efficacy (lumens/watt) */
} C_COLOR;              /* color context */
.DE
The
.I clock
member will be incremented each time the value gets changed by a
color field entity, and may be reset by the calling program if
desired.
This is a convenient way to keep track of whether or not a color has
changed since its last use.
The
.I flags
member indicates which color representations have been assigned,
and is an inclusive OR of one or more of the following:
.DS
#define C_CSSPEC        01              /* flag if spectrum is set */
#define C_CDSPEC        02              /* flag if defined w/ spectrum */
#define C_CSXY          04              /* flag if xy is set */
#define C_CDXY          010             /* flag if defined w/ xy */
#define C_CSEFF         020             /* flag if efficacy set */
.DE
.LP
It is possible but not recommended to alter the contents of the
color structure returned by
.I c_getcolor.
Normally, this routine is never called directly, since there are no
entities that access colors by name other than
.UL c.
.LP
The global variable
.I c_ccname
points to the name of the current color, or NULL if it is unnamed.
The variable
.I c_ccolor
points to the current color value, which should never be NULL.
.LP
The
.I c_ccvt
routine takes a
.I C_COLOR
structure and a set of desired flag settings and computes the
missing color representation(s).
.LP
The
.I c_isgrey
function returns 1 if the passed color is very close to neutral
grey, or 0 otherwise.
.SH
DIAGNOSTICS
.LP
The
.I c_hcolor
function returns
.I MG_OK
(0) if the color is handled correctly, or one of the predefined
error values if there is a problem.
(See the
.I mg_load
page for a list of errors.)\0
.LP
The
.I c_getcolor
function returns NULL if the specified color name is undefined, at
which point the calling function should return an
.I MG_EUNDEF
error.
.SH
SEE ALSO
.LP
c_hmaterial, c_hvertex, mg_init, mg_load
.ds RH C_HMATERIAL
.bp
.SH
NAME
.LP
c_hmaterial, c_getmaterial, c_cmname, c_cmaterial -
material entity support
.SH
SYNOPSIS
.LP
#include "parser.h"
.LP
.B int
c_hmaterial(
.B int
argc,
.B char
**argv )
.LP
C_MATERIAL *c_getmaterial(
.B char
*name )
.LP
.B "extern char"
*c_cmname
.LP
.B extern
C_MATERIAL *c_cmaterial
.SH
DESCRIPTION
.LP
The
.I c_hmaterial
function supports the MGF entities,
.UL m,
.UL ed,
.UL ir,
.UL rd,
.UL rs,
.UL sides,
.UL td,
and
.UL ts.
It is an error to support any of the material field entities without
supporting the
.UL m
entity itself.
The assignments are normally made to the
.I mg_ehand
array prior to parser initialization, like so:
.DS
mg_ehand[MG_E_MATERIAL] = c_hmaterial;  /* support "m" entity */
mg_ehand[MG_E_ED] = c_hmaterial;                /* support "ed" entity */
mg_ehand[MG_E_IR] = c_hmaterial;                /* support "ir" entity */
mg_ehand[MG_E_RD] = c_hmaterial;                /* support "rd" entity */
mg_ehand[MG_E_RS] = c_hmaterial;                /* support "rs" entity */
mg_ehand[MG_E_SIDES] = c_hmaterial;             /* support "sides" entity */
mg_ehand[MG_E_TD] = c_hmaterial;                /* support "td" entity */
mg_ehand[MG_E_TS] = c_hmaterial;                /* support "ts" entity */
/* other entity handler assignments... */
mg_init();                      /* initialize parser */
.DE
Any of the above entities besides
.UL m
may be unsupported, but the parser will not attempt to include their
effect into other members, e.g. an unsupported
.UL rs
component will not be added back into the
.UL rd
member.
It is therefore safer to support all of the relevant material
entities and make final approximations from the complete
.I C_MATERIAL
structure.
.LP
The
.I c_getmaterial
function takes the name of a defined material and returns a pointer to its
.I C_MATERIAL
structure, defined in "parser.h" as:
.DS
#define C_1SIDEDTHICK   0.005           /* assumed thickness of 1-sided mat. */

typedef struct {
        int     clock;          /* incremented each change -- resettable */
        int     sided;          /* 1 if surface is 1-sided, 0 for 2-sided */
        float   nr, ni;         /* index of refraction, real and imaginary */
        float   rd;             /* diffuse reflectance */
        C_COLOR rd_c;   /* diffuse reflectance color */
        float   td;             /* diffuse transmittance */
        C_COLOR td_c;   /* diffuse transmittance color */
        float   ed;             /* diffuse emittance */
        C_COLOR ed_c;   /* diffuse emittance color */
        float   rs;             /* specular reflectance */
        C_COLOR rs_c;   /* specular reflectance color */
        float   rs_a;           /* specular reflectance roughness */
        float   ts;             /* specular transmittance */
        C_COLOR ts_c;   /* specular transmittance color */
        float   ts_a;           /* specular transmittance roughness */
} C_MATERIAL;   /* material context */
.DE
The
.I clock
member will be incremented each time the value gets changed by a
material field entity, and may be reset by the calling program if
desired.
This is a convenient way to keep track of whether or not a material has
changed since its last use.
.LP
All reflectance and transmittance values correspond to normal
incidence, and may vary as a function of angle depending on the
index of refraction.
A solid object is normally represented with a one-sided material.
A two-sided material is most appropriate for thin surfaces, though
it may be used also when the surface normal orientations in a model
are unreliable.
.LP
If a transparent or translucent surface is one-sided, then the
absorption will change as a function of distance through the
material, and a single value for diffuse or specular transmittance is
ambiguous.
We therefore define a standard thickness,
.I C_1SIDEDTHICK,
which is the object thickness to which the given values correspond,
so that one may compute the isotropic absorptance of the material.
.LP
It is possible but not recommended to alter the contents of the
material structure returned by
.I c_getmaterial.
Normally, this routine is never called directly, since there are no
entities that access materials by name other than
.UL m.
.LP
The global variable
.I c_cmname
points to the name of the current material, or NULL if it is unnamed.
The variable
.I c_cmaterial
points to the current material value, which should never be NULL.
.SH
DIAGNOSTICS
.LP
The
.I c_hmaterial
function returns
.I MG_OK
(0) if the color is handled correctly, or one of the predefined
error values if there is a problem.
(See the
.I mg_load
page for a list of errors.)\0
.LP
The
.I c_getmaterial
function returns NULL if the specified material name is undefined, at
which point the calling function should return an
.I MG_EUNDEF
error.
.SH
SEE ALSO
.LP
c_hcolor, c_hvertex, mg_init, mg_load
.ds RH OBJ_HANDLER
.bp
.SH
NAME
.LP
obj_handler, obj_clear, obj_nnames, obj_name - object name support
.SH
SYNOPSIS
.LP
.B int
obj_handler(
.B int
argc,
.B char
**argv )
.LP
.B void
obj_clear(
.B void
)
.LP
.B "extern int"
obj_nnames
.LP
.B "extern char"
**obj_name
.SH
DESCRIPTION
.LP
The
.I obj_handler
routine should be assigned to the
.I MG_E_OBJECT
entry of the parser's
.I mg_ehand
array prior to calling
.I mg_load
if the loader/translator wishes to support hierarchical object
names.
.LP
The
.I obj_clear
function may be used to clear the object name stack and free any
associated memory, but this is usually not necessary since
.UL o
begin and end entities are normally balanced in the input.
.LP
The global
.I obj_nnames
variable indicates the number of names currently in the object
stack, and the
.I obj_name
list contains the name strings in the same order as they were
encountered on the input.
(I.e. the most recently pushed name is last.)\0
.SH
DIAGNOSTICS
.LP
The
.I obj_handler
function returns
.I MG_OK
(0) if the color is handled correctly, or one of the predefined
error values if there is a problem.
(See the
.I mg_load
page for a list of errors.)\0
.SH
SEE ALSO
.LP
mg_init, mg_load, xf_handler
.ds RH XF_HANDLER
.bp
.SH
NAME
.LP
xf_handler, xf_clear, xf_context, xf_argend - transformation support
.SH
SYNOPSIS
.LP
.B int
xf_handler(
.B int
argc,
.B char
**argv )
.LP
.B void
xf_clear(
.B void
)
.LP
.B extern
XF_SPEC *xf_context
.LP
.B "extern char"
**xf_argend
.SH
DESCRIPTION
.LP
The
.I xf_handler
routine should be assigned to the
.I MG_E_XF
entry of the parser's
.I mg_ehand
array prior to calling
.I mg_load
if the loader/translator wishes to support hierarchical
transformations.
(Note that all MGF geometric entities require this support.)\0
.LP
The
.I xf_clear
function may be used to clear the transform stack and free any
associated memory, but this is usually not necessary since
.UL xf
begin and end entities are normally balanced in the input.
.LP
The global
.I xf_context
variable points to the current transformation context, which is of
the type
.I XF_SPEC,
described in "parser.h":
.DS
typedef struct xf_spec {
        long    xid;                            /* unique transform id */
        short   xac;                            /* context argument count */
        short   rev;                            /* boolean true if vertices reversed */
        XF      xf;                             /* cumulative transformation */
        struct xf_array *xarr;          /* transformation array pointer */
        struct xf_spec  *prev;  /* previous transformation context */
} XF_SPEC;              /* followed by argument buffer */
.DE
The
.I xid
member is a identifier associated with this transformation,
which should be the same for identical transformations, as an aid to
vertex sharing.
(See also the
.I c_hvertex
page.)\0
The
.I xac
member indicates the total number of transform arguments, and is
used to indicate the position of the first argument relative to the
last one pointed to by the global
.I xf_argend
variable.
.LP
The first transform argument starts at
.I xf_argv,
which is a macro defined in "parser.h" as:
.DS
#define xf_argv         (xf_argend - xf_context->xac)
.DE
Note that accessing this macro will result in a segmentation violation
if the current context is NULL, so one should first test the second macro
.I xf_argc
against zero.
This macro is defined as:
.DS
#define xf_argc         (xf_context==NULL ? 0 : xf_context->xac)
.DE
.LP
Normally, neither of these macros will be used, since there are
routines for transforming points, vectors and scalars directly based
on the current transformation context.
(See the
.I xf_xfmpoint
page for details.)\0
.LP
The
.I rev
member of the
.I XF_SPEC
structure indicates whether or not this transform reverses the order
of polygon vertices.
This member will be 1 if the transformation mirrors about an odd
number of coordinate axes, thus inverting faces.
The usual thing to do in this circumstance is to interpret the
vertex arguments in the reverse order, so as to bring the face back
to its original orientation in the new position.
.LP
The
.I xf
member contains the transformation scalefactor (in xf.sca)
and 4x4 homogeneous matrix (in xf.xfm), but these will usually not
be accessed directly.
Likewise, the
.I xarr
and
.I prev
members point to data that should not be needed by the calling
program.
.SH
DIAGNOSTICS
.LP
The
.I xf_handler
function returns
.I MG_OK
(0) if the color is handled correctly, or one of the predefined
error values if there is a problem.
(See the
.I mg_load
page for a list of errors.)\0
.SH
SEE ALSO
.LP
mg_init, mg_load, obj_handler, xf_xfmpoint
.ds RH XF_XFMPOINT
.bp
.SH
NAME
.LP
xf_xfmpoint, xf_xfmvect, xf_rotvect, xf_scale - apply current
transformation
.SH
SYNOPSIS
.LP
.B void
xf_xfmpoint( FVECT pnew, FVECT pold )
.LP
.B void
xf_xfmvect( FVECT vnew, FVECT vold )
.LP
.B void
xf_rotvect( FVECT nnew, FVECT nold )
.LP
.B double
xf_scale(
.B double
sold )
.SH
DESCRIPTION
.LP
The
.I xf_xfmpoint
routine applies the current transformation defined by
.I xf_context
to the point
.I pold,
scaling, rotating and moving it to its proper location, which is put in
.I pnew.
(As for
.I xf_xfmvect
and
.I xf_rotvect,
the two arguments may point to the same vector.)\0
.LP
The
.I xf_xfmvect
routine applies the current transformation to the vector
.I vold,
scaling and rotating it to its proper location, which is put in
.I vnew.
The only difference between
.I xf_xfmpoint
and
.I xf_xfmvect
is that in the latter, the final translation is not applied.
.LP
The
.I xf_rotvect
routine rotates the vector
.I nold
using the current transformation, and stores the result in
.I nnew.
No translation or scaling is applied, which is the appropriate
action for surface normal vectors for example.
.LP
The
.I xf_scale
function takes a scalar argument
.I sold
and applies the current scale factor, returning the result.
.SH
SEE ALSO
.LP
xf_handler
.ds LH
.ds RH
.bp
.NH
Application Notes
.NH 2
Relation to Standard Practices in Computer Graphics
.LP
For those coming from a computer graphics background, some of the
choices in the material model may seem strange or even capricious.
Why not simply stick with RGB colors and a Phong specular component
like everyone else?
What is the point in choosing the number of sides to a material?
.LP
In the real world, a surface can have only one side,
defining the interface between one volume and another.
Many object-space rendering packages (e.g. z-buffer algorithms) take
advantage of this fact by culling back-facing polygons and thus saving
as much as 50% of the preprocessing time.
However, many models rely on an
approximation whereby a single surface is used to represent a very thin
volume, such as a pane of glass, and this also can provide significant
calculational savings in an image-space algorithm (such as
ray-tracing).
Also, many models are created in such a way that the front vs. back
information is lost or confused, so that the back side of one or
more surfaces may have to serve as the front side during rendering.
(AutoCAD is one easily identified culprit in this department.)\0
Since both types of surface models are useful and any
rendering algorithm may ultimately be applied, MGF provides a way
to specify sidedness rather than picking one interpretation or the other.
.LP
The problem with RGB is that there is no accepted standard, and even
if we were to set one it would either be impossible to realize (i.e.
impossible to create phosphors with the chosen colors) or it would
have a gamut that excludes many saturated colors.
The CIE color system was very carefully conceived and developed,
and is the standard to which all photometric measurements adhere.
It is therefore the logical choice in any standard format, though it
has been too often ignored by the computer graphics community.
.LP
Regarding Phong shading, this was never a physical model and making it
behave basic laws of reciprocity and energy balance is difficult.
More to the point, specular power has almost nothing to do with
surface microstructure, and is difficult to set properly
even if every physical characteristic of a material has
been carefully measured.
This is the ultimate indictment of any physical model -- that it
is incapable of reproducing any measurement whatsoever.
.LP
Admittedly, the compliment of diffuse and specular component plus
surface roughness and index of refraction used in MGF is less than a
perfect model, but it is serviceable for most materials and
relatively simple to incorporate into a rendering algorithm.
In the long term, MGF shall probably include full spectral
scattering functions, though the sheer quantity of data involved
makes this burdensome from both the measurement side and the
simulation side.
.NH 3
Converting between Phong Specular Power and Gaussian Roughness
.LP
So-called specular reflection and transmission are modeled using a
Gaussian distribution of surface facets.
The roughness parameters to the
.UL rs
and
.UL ts
entities specify
the root-mean-squared (RMS) surface facet slope, which varies from 0
for a perfectly smooth surface to around .2 for a fairly rough one.
The effect this will have on the reflected component distribution is
well-defined, but predicting the behavior of the transmitted
component requires further assumptions.
We assume that the surface
scatters light passing through it just as much as it scatters
reflected light.
This assumption is approximately correct for a
two-sided transparent material with an index of refraction of 1.5
(like glass) and both sides having the given RMS facet slope.
.LP
Oftentimes, one is translating from a Phong exponent on the cosine
of the half-vector-to-normal angle to the more physical but less
familiar Gaussian model of MGF.
The hardest part is translating the specular power to a roughness value.
For this, we recommend the following approximation:
.IP
roughness = 0.6/sqrt(specular_power)
.LP
It is not a perfect correlation, but it is about as close as one can get.
.NH 3
Converting between RGB and CIE Colors
.LP
Unlike most graphics languages, MGF does not use an RGB color model,
simply because there is no recognized definition for this model.
It is based on computer monitor phosphors, which vary from one
CRT to the next.
(There is an RGB standard defined in the TV
industry, but this has a rather poor correlation to most computer
monitors and it is impossible to express many real-world colors
within its limited gamut.)\0
.LP
MGF uses two alternative, well-defined standards, spectral power
distributions and the 1931 CIE 2 degree standard observer.
With the CIE standard, any viewable
color may be exactly represented as an (x,y) chromaticity value.
Unfortunately, the interaction between
colors (i.e. colored light sources and interreflections) cannot be
specified exactly with any finite coordinate set, including CIE
chromaticities.
So, MGF offers the ability to give reflectance,
transmittance or emittance as a function of wavelength over the visible
spectrum.
This function is still discretized, but at a user-selectable
resolution.
Furthermore, spectral colors may be mixed, providing (nearly)
arbitrary basis functions, which can produce more accurate results in
some cases and are merely a convenience for translation in others.
.LP
Conversion back and forth between CIE chromaticity coordinates and spectral
samples is provided within the MGF parser.
Unfortunately, conversion
to and from RGB values depends on a particular RGB definition, and as we
have said, there is no recognized standard.
We therefore recommend that
you decide yourself what chromaticity values to use for each RGB primary,
and adopt the following code to convert between CIE and RGB coordinates.
.LP
.nf
#ifdef  NTSC
#define  CIE_x_r                0.670           /* standard NTSC primaries */
#define  CIE_y_r                0.330
#define  CIE_x_g                0.210
#define  CIE_y_g                0.710
#define  CIE_x_b                0.140
#define  CIE_y_b                0.080
#define  CIE_x_w                0.3333          /* monitor white point */
#define  CIE_y_w                0.3333
#else
#define  CIE_x_r                0.640           /* nominal CRT primaries */
#define  CIE_y_r                0.330
#define  CIE_x_g                0.290
#define  CIE_y_g                0.600
#define  CIE_x_b                0.150
#define  CIE_y_b                0.060
#define  CIE_x_w                0.3333          /* monitor white point */
#define  CIE_y_w                0.3333
#endif

#define CIE_D           (       CIE_x_r*(CIE_y_g - CIE_y_b) + \\
                                CIE_x_g*(CIE_y_b - CIE_y_r) + \\
                                CIE_x_b*(CIE_y_r - CIE_y_g)     )
#define CIE_C_rD        ( (1./CIE_y_w) * \\
                                ( CIE_x_w*(CIE_y_g - CIE_y_b) - \\
                                  CIE_y_w*(CIE_x_g - CIE_x_b) + \\
                                  CIE_x_g*CIE_y_b - CIE_x_b*CIE_y_g     ) )
#define CIE_C_gD        ( (1./CIE_y_w) * \\
                                ( CIE_x_w*(CIE_y_b - CIE_y_r) - \\
                                  CIE_y_w*(CIE_x_b - CIE_x_r) - \\
                                  CIE_x_r*CIE_y_b + CIE_x_b*CIE_y_r     ) )
#define CIE_C_bD        ( (1./CIE_y_w) * \\
                                ( CIE_x_w*(CIE_y_r - CIE_y_g) - \\
                                  CIE_y_w*(CIE_x_r - CIE_x_g) + \\
                                  CIE_x_r*CIE_y_g - CIE_x_g*CIE_y_r     ) )

#define CIE_rf          (CIE_y_r*CIE_C_rD/CIE_D)
#define CIE_gf          (CIE_y_g*CIE_C_gD/CIE_D)
#define CIE_bf          (CIE_y_b*CIE_C_bD/CIE_D)

float  xyz2rgbmat[3][3] = {     /* XYZ to RGB */
        {(CIE_y_g - CIE_y_b - CIE_x_b*CIE_y_g + CIE_y_b*CIE_x_g)/CIE_C_rD,
         (CIE_x_b - CIE_x_g - CIE_x_b*CIE_y_g + CIE_x_g*CIE_y_b)/CIE_C_rD,
         (CIE_x_g*CIE_y_b - CIE_x_b*CIE_y_g)/CIE_C_rD},
        {(CIE_y_b - CIE_y_r - CIE_y_b*CIE_x_r + CIE_y_r*CIE_x_b)/CIE_C_gD,
         (CIE_x_r - CIE_x_b - CIE_x_r*CIE_y_b + CIE_x_b*CIE_y_r)/CIE_C_gD,
         (CIE_x_b*CIE_y_r - CIE_x_r*CIE_y_b)/CIE_C_gD},
        {(CIE_y_r - CIE_y_g - CIE_y_r*CIE_x_g + CIE_y_g*CIE_x_r)/CIE_C_bD,
         (CIE_x_g - CIE_x_r - CIE_x_g*CIE_y_r + CIE_x_r*CIE_y_g)/CIE_C_bD,
         (CIE_x_r*CIE_y_g - CIE_x_g*CIE_y_r)/CIE_C_bD}
};

float  rgb2xyzmat[3][3] = {     /* RGB to XYZ */
        {CIE_x_r*CIE_C_rD/CIE_D,CIE_x_g*CIE_C_gD/CIE_D,CIE_x_b*CIE_C_bD/CIE_D},
        {CIE_y_r*CIE_C_rD/CIE_D,CIE_y_g*CIE_C_gD/CIE_D,CIE_y_b*CIE_C_bD/CIE_D},
        {(1.-CIE_x_r-CIE_y_r)*CIE_C_rD/CIE_D,
         (1.-CIE_x_g-CIE_y_g)*CIE_C_gD/CIE_D,
         (1.-CIE_x_b-CIE_y_b)*CIE_C_bD/CIE_D}
};


cie_rgb(rgbcolor, ciecolor)             /* convert CIE to RGB */
register float  *rgbcolor, *ciecolor;
{
        register int  i;

        for (i = 0; i < 3; i++) {
                rgbcolor[i] =   xyz2rgbmat[i][0]*ciecolor[0] +
                                xyz2rgbmat[i][1]*ciecolor[1] +
                                xyz2rgbmat[i][2]*ciecolor[2] ;
                if (rgbcolor[i] < 0.0)          /* watch for negative values */
                        rgbcolor[i] = 0.0;
        }
}


rgb_cie(ciecolor, rgbcolor)             /* convert RGB to CIE */
register float  *ciecolor, *rgbcolor;
{
        register int  i;

        for (i = 0; i < 3; i++)
                ciecolor[i] =   rgb2xyzmat[i][0]*rgbcolor[0] +
                                rgb2xyzmat[i][1]*rgbcolor[1] +
                                rgb2xyzmat[i][2]*rgbcolor[2] ;
}
.fi
.LP
An alternative to adopting the above code is to use the MGF "cmix"
entity to convert from RGB directly by naming the three primaries in
terms of their chromaticities, e.g:
.DS
c R =
        cxy 0.640 0.330
c G =
        cxy 0.290 0.600
c B =
        cxy 0.150 0.060
.DE
.LP
Then, converting from RGB to MGF colors is as simple as multiplying each
component by its relative luminance in a cmix statement, for instance:
.DS
c white =
        cmix 0.265 R 0.670 G 0.065 B
.DE
.LP
For the chosen RGB standard, the above specification would result a pure
white.
The reason the coefficients are not all 1 as you might expect is
that cmix uses relative luminance as the standard for its weights.
Since
blue is less luminous for the same energy than red, which is in turn
less luminous than green, the weights cannot be the same to achieve an
even spectral balance.
Unfortunately, computing these relative weights
is not straightforward, though it is given in the above macros as CIE_rf,
CIE_gf and CIE_bf.
(The common factors in these macros may of course
be removed since
.UL cmix
weights are all relative.)\0
Alternatively, one could measure the actual full scale luminance of
the phosphors with a luminance probe and get the same relative
values.
.NH 2
Relation to IESNA LM-63 and Luminaire Catalogs
.LP
Recently, the Illuminating Engineering Society of North America
(IESNA) adopted MGF as the official standard for
representing luminaire geometry and materials.
The way this works in an IES luminaire data file is through the
addition of a keyword called LUMINOUSGEOMETRY, which is given on a
line in the header portion of a file (before the TILT specification)
like so:
.LP
.B [LUMINOUSGEOMETRY]
.I mgf_file
.LP
The given MGF file must exist relative to the directory containing
the IES file (i.e. the same stipulations and restrictions on pathnames
apply as for the MGF
.UL i
entity).
Furthermore, the position of the MGF geometry must be
such that the gross geometric specification of emitting surfaces
in the IES file completely
blocks or encloses the luminous portions of the MGF description.
Specifically, any ray traced towards the MGF geometry must strike
the IES gross geometry before it strikes any luminous surface in the
MGF description.
This provides a convenient way of preventing overcounting in the
illumination calculation, while still allowing for accurate fixture
appearance.
.LP
To give two examples, let us consider first a recessed can, followed
by a hanging direct/indirect fluorescent fixture.
.LP
The most appropriate IES geometric specification for the emitting
area of a can light would be a circular disk.
Since the IES gross geometry gives only the diameter of the disk, the
actual 3-dimensional placement is implicitly defined as having a
center at the origin, with the radiating disk facing in the
negative Z direction (nadir, downwards).
The MGF geometry would then be placed such that any luminous portion
was above this disk, and no portion of it would obstruct the IES
geometry.
The most sensible position therefore has the IES disk flush with the
MGF can opening, as shown in Figure 3.
.bp
Replace this page with the second page from "figures.ps".
.bp
.LP
In the case of a direct/indirect fluorescent fixture, light will
exit both the top and the bottom sides, and the IES geometry must
enclose the radiating portion of the fixture entirely.
It is acceptable to have additional MGF geometry above the 
fixture so long as it does not radiate, which is what we must do if
we wish to include the support rods, as shown in Figure 4.
.LP
Note that the origin is always in the exact center of the IES
geometry.
.LP
Not all fixtures will fit the simple IES geometry specification so
nicely.
For odd-shaped fixtures, it may be necessary to use an IES geometry
that does not match the radiating area terribly well in order that
it completely block or enclose the required MGF specification.
.LP
The unit of length in the MGF file is always meters, regardless of
the units specified in the enclosing IES file.
However, any and all multipliers applied to the candlepower data in the
IES file will also be applied to the emittance of surfaces in the
MGF specification, so that one MGF file may serve similar
luminaires that differ in their total output.
.NH
Credits
.LP
The MGF language grew out of a joint investigation into physical
representations for rendering undertaken by the author
(Greg Ward of LBL) and Holly Rushmeier of the National
Institute of Standards and Technology.
After deciding that a complete and robust specification was
an extreme challenge, we shelved the project for another time.
A few months later, the author spoke with Ian Ashdown and Robert
Shakespeare, who are both members of the IES Computing Committee,
about the need for extending the existing data standard to
include luminaire geometry and near-field photometry.
We then moved forward as a team towards a somewhat less ambitious
approach to physical materials and geometry that had the advantage
of simplicity and the possibility of support with a standard parser
library.
The author went to work over the next two months
on the detailed design of the language
and an ANSI-C parser, with regular feedback from the other three
team members.
Several months and several versions later, we arrived at release
1.0, which is the occasion of this document's creation.
.LP
Funding for this work... would be nice.
Revision:	1.9
Committed:	Wed Nov 22 12:23:27 1995 UTC (28 years, 5 months ago) by greg
Content type:	application/x-troff
Branch:	MAIN
Changes since 1.8:	+5 -1 lines
Log Message:	made mgfilt expand included files for version conformancy