Asp Forum - Ruby 3D Math libraries?

Bart Braem

12/7/2006 9:38:00 AM

Hello,

Does anyone know good Ruby 3D Math libraries? I found Math3D
(<http://raa.ruby-lang.org/project/m...) but I can not get it working,
ruby extconf.rb fails:
$ ruby -r mkmf extconf.rb
checking for atan() in -lm... no
No atan() in -lm
*** extconf.rb failed ***

This is after I commented the following line, which made the configuration
fail too:
CFLAGS << " -Wall -Wno-comment"

The Alg3D library is not bad, but it does not contain everything from the 3D
Math library. E.g. rotation is missing, which is something I need.

Regards,
Bart

8 Answers

Paul Lutus

12/7/2006 9:48:00 AM

Bart Braem wrote:

> Hello,
>
> Does anyone know good Ruby 3D Math libraries? I found Math3D
> (<http://raa.ruby-lang.org/project/m...) but I can not get it working,
> ruby extconf.rb fails:
> $ ruby -r mkmf extconf.rb
> checking for atan() in -lm... no
> No atan() in -lm

For some reason, you have a very limited math library on your system.
Chances are you will succeed in the compilation if you will replace your
math library with a complete one.

It is a little difficult to believe there is a math library without
"atan()", and it implies there is something basically wrong with the one
installed on your system.

My point is that the Math3D library is not the problem, it is solely a
problem with your native-code math library, on which Math3D depends.

--
Paul Lutus
http://www.ara...

Bill Kelly

12/7/2006 10:10:00 AM

From: "Bart Braem" <bart.braem@gmail.com>
>
> Does anyone know good Ruby 3D Math libraries? I found Math3D
> (<http://raa.ruby-lang.org/project/m...) but I can not get it working,
> ruby extconf.rb fails:
> $ ruby -r mkmf extconf.rb
> checking for atan() in -lm... no
> No atan() in -lm
> *** extconf.rb failed ***
>
> This is after I commented the following line, which made the configuration
> fail too:
> CFLAGS << " -Wall -Wno-comment"

Are you possibly on Windows? I recall needing to make a few
tweaks here and there to Math3D to get it to compile on
Windows.

Here are the changes:

diff -u orig/rb_math3d.c rb_math3d.c
--- orig/rb_math3d.c 2005-02-08 01:10:14.189772800 -0800
+++ rb_math3d.c 2005-02-08 02:10:09.669819200 -0800
@@ -4,8 +4,15 @@
* yoshi@giganet.net
************************************
*/
+#if defined(_WIN32)
+#define _USE_MATH_DEFINES
+#endif
#include "ruby.h"
#include <math.h>
+#if defined(_WIN32)
+#undef FAR
+#undef NEAR
+#endif
#include "math3d_native.h"

Hope this helps,

Bill

Bart Braem

12/7/2006 10:31:00 AM

Paul Lutus wrote:

> For some reason, you have a very limited math library on your system.
> Chances are you will succeed in the compilation if you will replace your
> math library with a complete one.
>
> It is a little difficult to believe there is a math library without
> "atan()", and it implies there is something basically wrong with the one
> installed on your system.
>
> My point is that the Math3D library is not the problem, it is solely a
> problem with your native-code math library, on which Math3D depends.

Strange, but I noticed my libm.a is not in my path. I'll fix that first...

Bart

Bart Braem

12/7/2006 10:31:00 AM

Bill Kelly wrote:

> Are you possibly on Windows? I recall needing to make a few
> tweaks here and there to Math3D to get it to compile on
> Windows.

I'm working on linux, but thanks for the information!

Bart

Paul Lutus

12/7/2006 10:45:00 AM

Bart Braem wrote:

> Paul Lutus wrote:
>
>> For some reason, you have a very limited math library on your system.
>> Chances are you will succeed in the compilation if you will replace your
>> math library with a complete one.
>>
>> It is a little difficult to believe there is a math library without
>> "atan()", and it implies there is something basically wrong with the one
>> installed on your system.
>>
>> My point is that the Math3D library is not the problem, it is solely a
>> problem with your native-code math library, on which Math3D depends.
>
> Strange, but I noticed my libm.a is not in my path. I'll fix that first...

That should change things, but if this problem persists, write a simple,
short C program that calls "atan()" and that needs to be compiled using the
-lm linker directive (these two don't necessarily follow from each other).
See what happens. If this fails, it will become obvious where the problem
lies.

--
Paul Lutus
http://www.ara...

Bart Braem

12/7/2006 10:52:00 AM

Paul Lutus wrote:

> For some reason, you have a very limited math library on your system.
> Chances are you will succeed in the compilation if you will replace your
> math library with a complete one.

I removed the check for the function and now the Makefile is perfectly
generated and make works flawlessly. Make test fails however:

make test
ruby test.rb
/usr/lib/ruby/site_ruby/1.8/rubygems/custom_require.rb:21:in
`require__': ./tests/Vector.tests.rb:556: no .<digit> floating literal
anymore; put 0 before dot (SyntaxError)
expected = vec.to_a.inject(0) {|x,y| x+(y*y)}.**(.5)
^
../tests/Vector.tests.rb:556: parse error, unexpected '.', expecting ')'
expected = vec.to_a.inject(0) {|x,y| x+(y*y)}.**(.5)
^
from /usr/lib/ruby/site_ruby/1.8/rubygems/custom_require.rb:21:in `require'
from test.rb:30
from /usr/lib/ruby/1.8/find.rb:39:in `find'
from /usr/lib/ruby/1.8/find.rb:38:in `catch'
from /usr/lib/ruby/1.8/find.rb:38:in `find'
from test.rb:20
make: *** [test] Error 1

Turning .5 into 0.5 fixed that. Afterwards I got this error:

make test
ruby test.rb
../tests/Matrix4.tests.rb:32: warning: default `to_a' will be obsolete
../tests/Bound.tests.rb:30: warning: default `to_a' will be obsolete
../tests/Require.tests.rb:27: warning: parenthesize argument(s) for future
version
Required 9 files.
test.rb:45:in `suite': undefined method `add' for
#<Test::Unit::TestSuite:0x404e0b84> (NoMethodError)
from test.rb:44:in `each_object'
from test.rb:44:in `suite'
from /usr/lib/ruby/1.8/test/unit/ui/console/testrunner.rb:27:in
`initialize'
from test.rb:53:in `new'
from test.rb:53
make: *** [test] Error 1

(I'll ignore the to_a warnings, those tests work)
But the suite method add is missing, I replaced it with the << method in
test.rb:
ObjectSpace.each_object( Class ) {|klass|
suite << klass.suite if klass < Test::Unit::TestCase
}

Next my tests ran, without good results:

make test
ruby test.rb
../tests/Matrix4.tests.rb:32: warning: default `to_a' will be obsolete
../tests/Bound.tests.rb:30: warning: default `to_a' will be obsolete
../tests/Require.tests.rb:27: warning: parenthesize argument(s) for future
version
Required 9 files.
Loaded suite Math3d Test Suite
Started
EEE.EEEEEE....EEEEEEEEEEE.EEEEEEEEE.EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE.EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE.EEEEEEFFEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
Finished in 0.207462 seconds.

1) Error:
test_00_7_contains?(OrthoTestCase):
NoMethodError: undefined method `each' for nil:NilClass
./tests/Ortho.tests.rb:65:in `test_00_7_contains?'
./tests/m3dUnitTest.rb:63:in `run'

Somehow the setup methods don't work? It seems there is something wrong with
the testsuite, but I don't see what...

Bart

Paul Lutus

12/7/2006 7:20:00 PM

Bart Braem wrote:

> Paul Lutus wrote:
>
>> For some reason, you have a very limited math library on your system.
>> Chances are you will succeed in the compilation if you will replace your
>> math library with a complete one.
>
> I removed the check for the function and now the Makefile is perfectly
> generated and make works flawlessly.

This is not a very good idea. The resulting library may appear to be intact,
but it may fail at some future point when the missing function is required.
There is a reason the test is present in the configuration procedure, it
should not be removed.

> Make test fails however:

/ ... snip long section of test results

> Somehow the setup methods don't work? It seems there is something wrong
> with the testsuite, but I don't see what...

It's possible the methods used to force the compilation to finish have
ruined the library's functioning. It's also possible that there is
something wrong with your system's math library. But those are minor
points.

I have been having a friendly debate in this newsgroup about the use of
libraries. I am in favor of using whatever method solves the problem at
hand, some others feel that a library, by definition carefully designed and
wrung out, is a better approach. There are good arguments for both sides of
this question.

But if the library is not wrung out or isn't very portable, or if it cannot
be easily compiled and run, the arguments in favor of using it break down.

It turns out that 3D math, that is to say, math involving vectors of at most
three elements, is not very difficult to write. If one executes a lot of
mathematical statements involving 3D vectors, one may encounter a
significant speed penalty if the executed code is written in plain Ruby
code, when compared to using a compiled library such as the one you are
trying to compile.

But I ask you to consider this. From a user's perspective, the total runtime
is the total of the preparation time plus the time spent using the library.
I suspect that the total time for making this library work for you will
soon cross over a threshold beyond which it would have been quicker and
easier to write, test and use your own 3D vector routines, composed in
plain Ruby.

Another advantage to writing your own code -- oops, I mean your
own /library/ -- is that the resulting /library/ will be understood by you,
it will be composed in the very readable and transparent syntax for which
Ruby is justly famous, and you will be able to change it very quickly and
easily to meet new requirements.

In exchange, it will not be nearly as fast as a compiled, native-code
library would have been, if the library had ever been successfully compiled
and run.

It's your time, your requirements, therefore it's your decision.

--
Paul Lutus
http://www.ara...

Bart Braem

12/8/2006 9:42:00 AM

Paul Lutus wrote:

> But I ask you to consider this. From a user's perspective, the total
> runtime is the total of the preparation time plus the time spent using the
> library. I suspect that the total time for making this library work for
> you will soon cross over a threshold beyond which it would have been
> quicker and easier to write, test and use your own 3D vector routines,
> composed in plain Ruby.
>
> Another advantage to writing your own code -- oops, I mean your
> own library -- is that the resulting library will be understood by you,
> it will be composed in the very readable and transparent syntax for which
> Ruby is justly famous, and you will be able to change it very quickly and
> easily to meet new requirements.
>
> In exchange, it will not be nearly as fast as a compiled, native-code
> library would have been, if the library had ever been successfully
> compiled and run.
>

I understand your arguments but I'm looking at this from different
perspectives.

As a programmer I do not mind fixing this to learn more about Ruby libraries
and have a better understanding on how to write and use them. I got this
one fixed (almost, see my most recent post) so I don't mind. Of course when
it would not have worked out soon enough I would have left it.

Another argument pro is that this library is tested. When writing this
myself I am bound to make errors. Here there are a number of tests so
that's less probable.

A last and most important argument is that I need the performance. I am
going to write large programs where lots of 3D calculations are necessary.
Then I need the C performance, but preferably hidden away in a nice Ruby
library.

(So my decision is to use the library, as I have it working.)

Regards,
Bart

comp.lang.ruby

Ruby 3D Math libraries?

Bart Braem

Paul Lutus

Bill Kelly

Bart Braem

Bart Braem

Paul Lutus

Bart Braem

Paul Lutus

Bart Braem

x Login to ForumsZone