Re: [osg-users] Coordinate system in all readerwriters

[Rob Smith]

Hi,

I just stumbled over some rotation-issues with my fbx-files (exported from 
3dsMax, Exporter 2011.3). All of them are rotated -90° around the x-axis. 

So I was wondering if this is the desired behaviour? And if yes, what exactly 
is the prupose? 


Thank you!

Cheers,
Rob

--
Read this topic online here:
http://forum.openscenegraph.org/viewtopic.php?p=30520#30520





___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Paul Martz]

Sukender wrote:

Hi Paul and all,

I agree with you simplification, but I'm not sure it'll keep write(read(A))==A. 
For instance:
- Read a model with a rotation (Y to Z-up for instance). Your +y facing plane 
will be then +z facing.
- Write the model. Model is still rotated and will be +z in the file, thus 
having write(read(A))!=A.


Transforms during write are already entirely under application control 
because the application has access to the scene graph. In the case 
above, if the app wants to restore the original coordinate system, it 
would do this with the inverse transform before it writes the scene graph.


In fact, I don't think we need these options in the plugin for read, 
either. The application would simply transform the scene graph after the 
read completes.


In my opinion, the only situation in which plugins _require_ these 
options is for the few formats that store units, handedness, and 
coordinate system info in the file format, and the app needs to 
"convertTo" a la OpenFlight. For all other formats, the application can 
do the transformation itself, with a NodeVisitor or Transform.


But we should probably support some kind of transform Options for all 
import plugins anyway, for backwards compatibility. If we change all 
import plugins to not transform by default, application developers will 
want an easy switch to get the old default behavior from the plugin.

   -Paul



___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Sukender]

Hi Paul and all,

I agree with you simplification, but I'm not sure it'll keep write(read(A))==A. 
For instance:
- Read a model with a rotation (Y to Z-up for instance). Your +y facing plane 
will be then +z facing.
- Write the model. Model is still rotated and will be +z in the file, thus 
having write(read(A))!=A.

This is why I suggested writers should "undo" what is done during the reading 
by applying the opposite transform given by the read options. Of course one 
could have control over those things by changing options between reading and 
writing, but it's then the user's choice. It's surely the same when writing 
first and then reading.

Am I wrong?

Cheers,

Sukender
PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/

- "Paul Martz"  a écrit :

> To simplify things, I'd propose the following:
> 
>   - I don't think we need options to control writing. Your app can do
> 
> any required transformation with a NodeVisitor before calling osgDB to
> 
> write. The plugin should just write the scene graph passed to it.
>   - For reading, we just need rotation and scale options. "rotate
>  
> " and "scale   ". Those alone will handle changes to
> 
> orientation, handedness, and unit scaling.
>   - If the format stores an up vector, handedness, and unit
> information, 
> then the import plugins for those few formats need some kind of 
> "convertTo" option, sort of like FLT's "convertToMeters" (which scales
> 
> the model from its native units).
> 
> As an addendum to the last point, regarding Ulrich's comment, an
> import 
> plugin can only convert to right-handed if it knows the handedness of
> 
> the model. Which formats store that information? And to clarify, 
> OpenGL's "right-handed" rule is only the default for the definition of
> a 
> front face. Your application and your models may or may not use this 
> convention, and I think it would be wrong for a plugin to do such a 
> conversion automatically, assuming it knows something about the 
> convention you are using.
> 
> Finally, in principle, I agree: write( read( A ) ) should equal A. But
> I 
> want to re-emphasize this can only be true in the context of
> coordinate 
> systems. (100% equality in round-trip model conversions is essentially
> 
> an impossibility.)
> 
> Paul Martz
> Skew Matrix Software LLC
> _http://www.skew-matrix.com_ 
> +1 303 859 9466
> 
> 
> 
> ___
> osg-users mailing list
> osg-users@lists.openscenegraph.org
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Paul Martz]

To simplify things, I'd propose the following:

 - I don't think we need options to control writing. Your app can do 
any required transformation with a NodeVisitor before calling osgDB to 
write. The plugin should just write the scene graph passed to it.
 - For reading, we just need rotation and scale options. "rotate  
" and "scale   ". Those alone will handle changes to 
orientation, handedness, and unit scaling.
 - If the format stores an up vector, handedness, and unit information, 
then the import plugins for those few formats need some kind of 
"convertTo" option, sort of like FLT's "convertToMeters" (which scales 
the model from its native units).


As an addendum to the last point, regarding Ulrich's comment, an import 
plugin can only convert to right-handed if it knows the handedness of 
the model. Which formats store that information? And to clarify, 
OpenGL's "right-handed" rule is only the default for the definition of a 
front face. Your application and your models may or may not use this 
convention, and I think it would be wrong for a plugin to do such a 
conversion automatically, assuming it knows something about the 
convention you are using.


Finally, in principle, I agree: write( read( A ) ) should equal A. But I 
want to re-emphasize this can only be true in the context of coordinate 
systems. (100% equality in round-trip model conversions is essentially 
an impossibility.)


Paul Martz
Skew Matrix Software LLC
_http://www.skew-matrix.com_ 
+1 303 859 9466



___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Sukender]

Hi ulrich,

As Paul said, OpenGL doesn't have any convention. It's just a "common usage" 
that tell us we should use right-handed.
However, I agree the number of proposed options could be reduced (I just fear 
that it may become less explicit... or not?).

Your reflexion make me think we should merge read and write options. That way, 
we may have the two options you mentionned, and writing would the do the 
opposite operation. We thus keep write(read(A)) == A.

Is that okay for you?

Sukender
PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/

- "Ulrich Hertlein"  a écrit :

> Hi all,
> 
> On 30/12/09 11:35 AM, Sukender wrote:
> > - Have plugins do NO rotation by default (reading and writing) when
> the format doesn't
> have coordinate system information (that is to say all but FBX and
> maybe FLT).
> > - Have plugins read and apply coordinate system information if
> available by default when reading.
> > - Have optionnal READING options:
> > - readXup, readYup, readZup to suppose model is X, Y or Z-up
> (overrides format's
> coordinate system information)
> > - readLeftHanded, readRightHanded (similar setting)
> > - When in conflict (two options defined), use the latest one
> > - Have optionnal WRITING options:
> > - writeXup, writeYup, writeZup, writeLeftHanded, writeRightHanded
> (which should almost
> never be supported, in my opinion)
> > - Let plugings optionnally support these options prefixed with their
> name and "_". For
> instance "3ds_readYup" would assume ONLY ".3ds" files are Y-up. These
> would take
> precedence over the general settings.
> > - Also support "pluginName_read/writeAutoUp" and
> "pluginName_read/writeAutoHanded",
> which would specifically deactivate a global setting for a given
> plugin. For instance,
> 
> Do we really want all those options?  At the moment we only have:
> - rotate around X-axis (for Y-up to Z-up conversion) and
> - convert left- to right-hand CS
> 
> While I agree that plugins could leave the orientation alone I still
> think that they
> should do the conversion to a right-handed coordinate system by
> default, because the
> default OpenGL face-winding is right handed.
> 
> Additionally, how would the applied transformation be transmitted to
> the writer?  If this
> were to be stored with the loaded node then it could be automatically
> be undone by the
> writer.  Another possibility might be to just create a transformation
> node above the
> unmodified model.  The transformation node could do all the necessary
> scaling/rotation
> (and FrontFace); this could simply be ignored when writing.  The
> downside of this is that
> it doesn't integrate nicely and might be a performance penalty.
> 
> Cheers,
> /ulrich
> ___
> osg-users mailing list
> osg-users@lists.openscenegraph.org
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Ulrich Hertlein]

Hi all,

On 30/12/09 11:35 AM, Sukender wrote:
> - Have plugins do NO rotation by default (reading and writing) when the 
> format doesn't
have coordinate system information (that is to say all but FBX and maybe FLT).
> - Have plugins read and apply coordinate system information if available by 
> default when reading.
> - Have optionnal READING options:
> - readXup, readYup, readZup to suppose model is X, Y or Z-up (overrides 
> format's
coordinate system information)
> - readLeftHanded, readRightHanded (similar setting)
> - When in conflict (two options defined), use the latest one
> - Have optionnal WRITING options:
> - writeXup, writeYup, writeZup, writeLeftHanded, writeRightHanded (which 
> should almost
never be supported, in my opinion)
> - Let plugings optionnally support these options prefixed with their name and 
> "_". For
instance "3ds_readYup" would assume ONLY ".3ds" files are Y-up. These would take
precedence over the general settings.
> - Also support "pluginName_read/writeAutoUp" and 
> "pluginName_read/writeAutoHanded",
which would specifically deactivate a global setting for a given plugin. For 
instance,

Do we really want all those options?  At the moment we only have:
- rotate around X-axis (for Y-up to Z-up conversion) and
- convert left- to right-hand CS

While I agree that plugins could leave the orientation alone I still think that 
they
should do the conversion to a right-handed coordinate system by default, 
because the
default OpenGL face-winding is right handed.

Additionally, how would the applied transformation be transmitted to the 
writer?  If this
were to be stored with the loaded node then it could be automatically be undone 
by the
writer.  Another possibility might be to just create a transformation node 
above the
unmodified model.  The transformation node could do all the necessary 
scaling/rotation
(and FrontFace); this could simply be ignored when writing.  The downside of 
this is that
it doesn't integrate nicely and might be a performance penalty.

Cheers,
/ulrich
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Sukender]

Hi Rizzen,

Of course! I forgot to add this to the summary.

Sukender
PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/

- "Rizzen"  a écrit :

> Hi all,
> 
> Defining a coordinate system for OSG is a good idea. Though to me the
> most important aspect of reader and writers, is that this expression
> must be true:
> 
> read( write( A ) ) = A
> 
> Rizzen
> 
> Sukender wrote:
> > Hi Ulrich, Paul, and all,
> >
> > I agree we should unify the coordinate system during loading AND
> writing. Here are my suggestions :
> > - Have plugins do NO rotation by default (reading and writing) when
> the format doesn't have coordinate system information (that is to say
> all but FBX and maybe FLT).
> > - Have plugins read and apply coordinate system information if
> available by default when reading.
> > - Have optionnal READING options:
> > - readXup, readYup, readZup to suppose model is X, Y or Z-up
> (overrides format's coordinate system information)
> > - readLeftHanded, readRightHanded (similar setting)
> > - When in conflict (two options defined), use the latest one
> > - Have optionnal WRITING options:
> > - writeXup, writeYup, writeZup, writeLeftHanded,
> writeRightHanded (which should almost never be supported, in my
> opinion)
> > - Let plugings optionnally support these options prefixed with their
> name and "_". For instance "3ds_readYup" would assume ONLY ".3ds"
> files are Y-up. These would take precedence over the general settings.
> > - Also support "pluginName_read/writeAutoUp" and
> "pluginName_read/writeAutoHanded", which would specifically deactivate
> a global setting for a given plugin. For instance, "fbx_readAutoUp"
> would let the plugin decide the up vector (here using coordinate
> system information in the .fbx). "3ds_readAutoUp" would mean "no
> rotation" for .3ds.
> > - Write a portion of code somewhere in osgDB to parse all these
> options and store it in a nice structure (to avoid each plugin
> duplicating code). To be called with something like "myStructure =
> osgDB::readStandardCoordinateSystemOptions(PLUGIN_NAME)".
> > - As OpenGL doesn't have a "etched in stone" orientation standard,
> use the osgGA as an "internal" standard (that is to say +Z up, right
> handed) when a change is needed.
> >
> > 1. Please raise your hand if you agree, and tell about changes you'd
> like to introduce in this summary.
> > 2a. Please mention all plugins you know which break these rules.
> > 2b. Please tell if you agree modifying those plugins.
> >
> > When all is ok, I'll code it.
> >
> > Cheers,
> >
> > Sukender
> > PVLE - Lightweight cross-platform game engine -
> http://pvle.sourceforge.net/
> >
> > - "Ulrich Hertlein"  a écrit :
> >
> >
> >> On 29/12/09 6:52 PM, Paul Martz wrote:
> >>
> >>> Paul Martz wrote:
> >>>
>  I've noticed the same thing with OBJ.
> 
> >>> A little digging reveals that the OBJ plugin has a "noRotation"
> >>>
> >> Option
> >>
> >>> that disables the x axis rotation. The rotation is only done on
> >>>
> >> file
> >>
> >>> load, not on file write.
> >>>
> >>> Rotation about x is on by default. I'd suggest a change to off by
> >>> default, and the Option changed to "doRotation" to enable it.
> >>>
> >> The .x plugin has 'leftHanded' and 'rightHanded'.  Maybe we could
> >> establish a common
> >> option for this type of functionality (one that does 'convert to
> >> right-handed, Z-axis up').
> >>
> >> There already are options with (apparently) related functionality
> >> (courtesy of 'osgconv
> >> --formats'):
> >>
> >> Coordinate system:
> >>obj:noRotation
> >>x:leftHanded
> >>x:rightHanded
> >>
> >> Optimization:
> >>obj:generateFacetNormals
> >>obj:noTriStripPolygons
> >>stl:generateNormals
> >>stl:smooth
> >>stl:tristrip
> >>
> >> Image orientation
> >>hdr:NO_YFLIP
> >>hdr:YFLIP
> >>x:flipTexture
> >>
> >> Cheers,
> >> /ulrich
> >> ___
> >> osg-users mailing list
> >> osg-users@lists.openscenegraph.org
> >>
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
> >>
> > ___
> > osg-users mailing list
> > osg-users@lists.openscenegraph.org
> >
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
> >
> 
> ___
> osg-users mailing list
> osg-users@lists.openscenegraph.org
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Rizzen]

Hi all,

Defining a coordinate system for OSG is a good idea. Though to me the
most important aspect of reader and writers, is that this expression
must be true:

read( write( A ) ) = A

Rizzen

Sukender wrote:
> Hi Ulrich, Paul, and all,
>
> I agree we should unify the coordinate system during loading AND writing. 
> Here are my suggestions :
> - Have plugins do NO rotation by default (reading and writing) when the 
> format doesn't have coordinate system information (that is to say all but FBX 
> and maybe FLT).
> - Have plugins read and apply coordinate system information if available by 
> default when reading.
> - Have optionnal READING options:
> - readXup, readYup, readZup to suppose model is X, Y or Z-up (overrides 
> format's coordinate system information)
> - readLeftHanded, readRightHanded (similar setting)
> - When in conflict (two options defined), use the latest one
> - Have optionnal WRITING options:
> - writeXup, writeYup, writeZup, writeLeftHanded, writeRightHanded (which 
> should almost never be supported, in my opinion)
> - Let plugings optionnally support these options prefixed with their name and 
> "_". For instance "3ds_readYup" would assume ONLY ".3ds" files are Y-up. 
> These would take precedence over the general settings.
> - Also support "pluginName_read/writeAutoUp" and 
> "pluginName_read/writeAutoHanded", which would specifically deactivate a 
> global setting for a given plugin. For instance, "fbx_readAutoUp" would let 
> the plugin decide the up vector (here using coordinate system information in 
> the .fbx). "3ds_readAutoUp" would mean "no rotation" for .3ds.
> - Write a portion of code somewhere in osgDB to parse all these options and 
> store it in a nice structure (to avoid each plugin duplicating code). To be 
> called with something like "myStructure = 
> osgDB::readStandardCoordinateSystemOptions(PLUGIN_NAME)".
> - As OpenGL doesn't have a "etched in stone" orientation standard, use the 
> osgGA as an "internal" standard (that is to say +Z up, right handed) when a 
> change is needed.
>
> 1. Please raise your hand if you agree, and tell about changes you'd like to 
> introduce in this summary.
> 2a. Please mention all plugins you know which break these rules.
> 2b. Please tell if you agree modifying those plugins.
>
> When all is ok, I'll code it.
>
> Cheers,
>
> Sukender
> PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/
>
> - "Ulrich Hertlein"  a écrit :
>
>   
>> On 29/12/09 6:52 PM, Paul Martz wrote:
>> 
>>> Paul Martz wrote:
>>>   
 I've noticed the same thing with OBJ.
 
>>> A little digging reveals that the OBJ plugin has a "noRotation"
>>>   
>> Option
>> 
>>> that disables the x axis rotation. The rotation is only done on
>>>   
>> file
>> 
>>> load, not on file write.
>>>
>>> Rotation about x is on by default. I'd suggest a change to off by
>>> default, and the Option changed to "doRotation" to enable it.
>>>   
>> The .x plugin has 'leftHanded' and 'rightHanded'.  Maybe we could
>> establish a common
>> option for this type of functionality (one that does 'convert to
>> right-handed, Z-axis up').
>>
>> There already are options with (apparently) related functionality
>> (courtesy of 'osgconv
>> --formats'):
>>
>> Coordinate system:
>>obj:noRotation
>>x:leftHanded
>>x:rightHanded
>>
>> Optimization:
>>obj:generateFacetNormals
>>obj:noTriStripPolygons
>>stl:generateNormals
>>stl:smooth
>>stl:tristrip
>>
>> Image orientation
>>hdr:NO_YFLIP
>>hdr:YFLIP
>>x:flipTexture
>>
>> Cheers,
>> /ulrich
>> ___
>> osg-users mailing list
>> osg-users@lists.openscenegraph.org
>> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
>> 
> ___
> osg-users mailing list
> osg-users@lists.openscenegraph.org
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
>   

___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Sukender]

Hi Ulrich, Paul, and all,

I agree we should unify the coordinate system during loading AND writing. Here 
are my suggestions :
- Have plugins do NO rotation by default (reading and writing) when the format 
doesn't have coordinate system information (that is to say all but FBX and 
maybe FLT).
- Have plugins read and apply coordinate system information if available by 
default when reading.
- Have optionnal READING options:
- readXup, readYup, readZup to suppose model is X, Y or Z-up (overrides 
format's coordinate system information)
- readLeftHanded, readRightHanded (similar setting)
- When in conflict (two options defined), use the latest one
- Have optionnal WRITING options:
- writeXup, writeYup, writeZup, writeLeftHanded, writeRightHanded (which 
should almost never be supported, in my opinion)
- Let plugings optionnally support these options prefixed with their name and 
"_". For instance "3ds_readYup" would assume ONLY ".3ds" files are Y-up. These 
would take precedence over the general settings.
- Also support "pluginName_read/writeAutoUp" and 
"pluginName_read/writeAutoHanded", which would specifically deactivate a global 
setting for a given plugin. For instance, "fbx_readAutoUp" would let the plugin 
decide the up vector (here using coordinate system information in the .fbx). 
"3ds_readAutoUp" would mean "no rotation" for .3ds.
- Write a portion of code somewhere in osgDB to parse all these options and 
store it in a nice structure (to avoid each plugin duplicating code). To be 
called with something like "myStructure = 
osgDB::readStandardCoordinateSystemOptions(PLUGIN_NAME)".
- As OpenGL doesn't have a "etched in stone" orientation standard, use the 
osgGA as an "internal" standard (that is to say +Z up, right handed) when a 
change is needed.

1. Please raise your hand if you agree, and tell about changes you'd like to 
introduce in this summary.
2a. Please mention all plugins you know which break these rules.
2b. Please tell if you agree modifying those plugins.

When all is ok, I'll code it.

Cheers,

Sukender
PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/

- "Ulrich Hertlein"  a écrit :

> On 29/12/09 6:52 PM, Paul Martz wrote:
> > Paul Martz wrote:
> >> I've noticed the same thing with OBJ.
> > 
> > A little digging reveals that the OBJ plugin has a "noRotation"
> Option
> > that disables the x axis rotation. The rotation is only done on
> file
> > load, not on file write.
> > 
> > Rotation about x is on by default. I'd suggest a change to off by
> > default, and the Option changed to "doRotation" to enable it.
> 
> The .x plugin has 'leftHanded' and 'rightHanded'.  Maybe we could
> establish a common
> option for this type of functionality (one that does 'convert to
> right-handed, Z-axis up').
> 
> There already are options with (apparently) related functionality
> (courtesy of 'osgconv
> --formats'):
> 
> Coordinate system:
>obj:noRotation
>x:leftHanded
>x:rightHanded
> 
> Optimization:
>obj:generateFacetNormals
>obj:noTriStripPolygons
>stl:generateNormals
>stl:smooth
>stl:tristrip
> 
> Image orientation
>hdr:NO_YFLIP
>hdr:YFLIP
>x:flipTexture
> 
> Cheers,
> /ulrich
> ___
> osg-users mailing list
> osg-users@lists.openscenegraph.org
> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Ulrich Hertlein]

On 29/12/09 6:52 PM, Paul Martz wrote:
> Paul Martz wrote:
>> I've noticed the same thing with OBJ.
> 
> A little digging reveals that the OBJ plugin has a "noRotation" Option
> that disables the x axis rotation. The rotation is only done on file
> load, not on file write.
> 
> Rotation about x is on by default. I'd suggest a change to off by
> default, and the Option changed to "doRotation" to enable it.

The .x plugin has 'leftHanded' and 'rightHanded'.  Maybe we could establish a 
common
option for this type of functionality (one that does 'convert to right-handed, 
Z-axis up').

There already are options with (apparently) related functionality (courtesy of 
'osgconv
--formats'):

Coordinate system:
   obj:noRotation
   x:leftHanded
   x:rightHanded

Optimization:
   obj:generateFacetNormals
   obj:noTriStripPolygons
   stl:generateNormals
   stl:smooth
   stl:tristrip

Image orientation
   hdr:NO_YFLIP
   hdr:YFLIP
   x:flipTexture

Cheers,
/ulrich
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Paul Martz]

Paul Martz wrote:

I've noticed the same thing with OBJ.


A little digging reveals that the OBJ plugin has a "noRotation" Option 
that disables the x axis rotation. The rotation is only done on file 
load, not on file write.


Rotation about x is on by default. I'd suggest a change to off by 
default, and the Option changed to "doRotation" to enable it.

   -Paul


___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Paul Martz]

Sukender wrote:

What I like with you is that your answers are clear!


Ha!

The only thing that bothers me is: should "write( read(A) ) == A"? As 
far as I'm concerned, I would answer "yes". And you?


Yes, for the coordinate system, definitely.

This is not the case when reading FBX and writing, for instance 
(regardless of the output format). So I've been scratching my head to 
figure out where the problem (bug?) is, and I still don't find a 
suitable answer. Maybe something stupid I don't see in my code...


I've noticed the same thing with OBJ. Try this:
  osgconv dumptruck.osg dumptruck.obj
  osgviewer dumptruck.obj
You get a truck on its side. I agree with you that this is incorrect 
behavior. This doesn't happen in conversion to FLT and back to OSG, for 
example.


I'm reluctant to use the term "bug" because I think the plugin was coded 
this way intentionally, so the code is doing exactly what it was 
designed to do. Perhaps it's a "bug" in the design of the plugin, which 
was trying to solve a problem that didn't really exist.


When I encountered this issue with OBJ, I wanted to submit a fix for it, 
but didn't want to waste time creating a fix that would get rejected 
because it might break existing apps. So, thank you for bringing up this 
topic. :-) If we all agree this is wrong, then we can fix the issue, 
accept the consequences, and be better off several months down the road.


I ended up working around the OBJ issue by using the ROT pseudoloader to 
apply a -90 degree rotation around the x axis:

  osgconv dumptruck.osg.-90,1,0,0.rot dumptruck.obj
  osgviewer dumptruck.obj
You should be able to do the same, in lieu of any code change.
   -Paul

___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Sukender]

Le Mon, 28 Dec 2009 19:11:40 +0100, Paul Martz  a écrit:


Hi Sukender --

First, OpenGL doesn't really have a default coordinate system (see the
thread "Cameras View Matrix - I am perplexed" from October). osgGA
camera manipulators use a right-handed +z up convention for the home
position, but this is easily configured to anything else by the app if
desired. Also, OSG apps don't have to use osgGA manipulators, they can
use their own. So although osgGA has a default, I'd argue it's incorrect
for any code to make assumptions about an OSG app's coordinate system.

Regarding OSG plugins, I think their primary job is to either give me
what they read, or write out what I give them. If I modeled an aircraft
with the nose pointing in +y, then I expect the OSG plugin to give me a
scene graph with the nose pointing in +y. If it doesn't do this by
default, that's a bug, isn't it?

That being said, I don't object to OSG plugins having Options to control
application of transformations. In fact, this is required in some cases.
For example, OpenFlight DOF nodes specify range of motion parameters,
and the plugin must scale and rotate these values to match the app's
coordinate system.

Paul Martz
Skew Matrix Software LLC
_http://www.skew-matrix.com_ 
+1 303 859 9466


___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org



Hi Paul,

What I like with you is that your answers are clear! ;) Well you're absolutely 
right. Thanks.

The only thing that bothers me is: should "write( read(A) ) == A"? As far as I'm 
concerned, I would answer "yes". And you?
This is not the case when reading FBX and writing, for instance (regardless of 
the output format). So I've been scratching my head to figure out where the 
problem (bug?) is, and I still don't find a suitable answer. Maybe something 
stupid I don't see in my code...

Cheers,

--
Sukender
PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

Re: [osg-users] Coordinate system in all readerwriters

[Paul Martz]

Hi Sukender --

First, OpenGL doesn't really have a default coordinate system (see the 
thread "Cameras View Matrix - I am perplexed" from October). osgGA 
camera manipulators use a right-handed +z up convention for the home 
position, but this is easily configured to anything else by the app if 
desired. Also, OSG apps don't have to use osgGA manipulators, they can 
use their own. So although osgGA has a default, I'd argue it's incorrect 
for any code to make assumptions about an OSG app's coordinate system.


Regarding OSG plugins, I think their primary job is to either give me 
what they read, or write out what I give them. If I modeled an aircraft 
with the nose pointing in +y, then I expect the OSG plugin to give me a 
scene graph with the nose pointing in +y. If it doesn't do this by 
default, that's a bug, isn't it?


That being said, I don't object to OSG plugins having Options to control 
application of transformations. In fact, this is required in some cases. 
For example, OpenFlight DOF nodes specify range of motion parameters, 
and the plugin must scale and rotate these values to match the app's 
coordinate system.


Paul Martz
Skew Matrix Software LLC
_http://www.skew-matrix.com_ 
+1 303 859 9466


___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

[osg-users] Coordinate system in all readerwriters

[Sukender]

Hi all,

I've got a question/remark: Are all readers supposed to adapt models to the 
OpenGL's coordinate system (I mean that +/- Y/Z-up thing)? If so:
1. It doesn't seem to be with some readers.
2. All writers, AFAIK, directly write the scene without any further processing. 
Thus osgconv-like programs may produce a rotated scene (with respect to the 
original one) if the reader processes coordinate system and writer doesn't. 
When reading an FBX and exporting to 3DS for instance, it may happen.

On the other hand, if readers are supposed to ignore this, we'll loose features 
by removing coordinate system adaptation in existing readers (=FBX).

Other problems are:
- If we change things in readers, it will break many users' apps.
- If we plan to make all writers respect the coordinate systems, it'll also 
break things.

Discussion is opened :)

Cheers,

Sukender
PVLE - Lightweight cross-platform game engine - http://pvle.sourceforge.net/
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org