Re: [eigen] Forward declarations

[David Lonie <loniedavid@xxxxxxxxx>]

When I try that, AutoAlign, RowMajor, ColMajor, and
EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION are all undefined. I could
swap out the integer values for the constants and have a fragile
solution, but AIUI EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION is
build-specific and cannot be known without including Eigen. Should I
assume that the class cannot be forward declared and start looking at
my other options, or is there a trick to get around this?

Thanks,

Dave

On Tue, Jul 19, 2011 at 11:17 AM, Gael Guennebaud
<gael.guennebaud@xxxxxxxxx> wrote:
> look at the file Eigen/src/Core/util/ForwardDeclarations.h, here is
> how it is declared:
>
> template<typename _Scalar, int _Rows, int _Cols,
>         int _Options = AutoAlign |
> #if defined(__GNUC__) && __GNUC__==3 && __GNUC_MINOR__==4
>    // workaround a bug in at least gcc 3.4.6
>    // the innermost ?: ternary operator is misparsed. We write it slightly
>    // differently and this makes gcc 3.4.6 happy, but it's ugly.
>    // The error would only show up with EIGEN_DEFAULT_TO_ROW_MAJOR is defined
>    // (when EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION is RowMajor)
>                          ( (_Rows==1 && _Cols!=1) ? RowMajor
>                          : !(_Cols==1 && _Rows!=1) ?
> EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION
>                          : ColMajor ),
> #else
>                          ( (_Rows==1 && _Cols!=1) ? RowMajor
>                          : (_Cols==1 && _Rows!=1) ? ColMajor
>                          : EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION ),
> #endif
>         int _MaxRows = _Rows,
>         int _MaxCols = _Cols
>> class Matrix;
>
> gael
>
> On Tue, Jul 19, 2011 at 5:08 PM, David Lonie <loniedavid@xxxxxxxxx> wrote:
>> Hi Gael,
>>
>> On Tue, Jul 19, 2011 at 3:21 AM, Gael Guennebaud
>> <gael.guennebaud@xxxxxxxxx> wrote:
>>> Hi,
>>>
>>> that's not so easy because I see that OpenQube headers typically
>>> contains functions returning Eigen::Vector3d objects by value or
>>> classes having Eigen's objects as members. That means the compiler has
>>> to know what is, e.g., a Eigen::Vector3d so I'm afraid simple forward
>>> declarations won't help.
>>
>> The OpenQube library is still fairly new, and I may be able to rewrite
>> the API to work around those issues -- at this point, I'm just trying
>> to gather information to figure out how to best handle this. My
>> options right now are (a) Forward declare Eigen classes (and rewrite
>> OpenQube API appropriately), (b) Remove Eigen from OpenQube's API
>> altogether, or (c) Make Eigen an optional dependency of VTK.
>>
>> (b) would lose some flexibility and be a decent bit of work, and (c)
>> may be difficult, while (a), if possible, would be the best solution
>> that I can see.
>>
>> Is it possible to forward declare the Eigen::Matrix class? This would
>> be useful for a few other projects I'm working on as well ;-)
>>
>>> Gael
>>>
>>> On Mon, Jul 18, 2011 at 4:09 PM, David Lonie <loniedavid@xxxxxxxxx> wrote:
>>>> Hi List,
>>>>
>>>> I'm currently integrating a library that uses Eigen vectors/matrices
>>>> in its API into VTK, and I have a quick question: Is it possible to
>>>> forward-declare Eigen::Matrix? The library (OpenQube
>>>> http://github.com/cryos/openqube) currently includes Eigen/Core in
>>>> it's headers, but I'd like to replace these includes with forward
>>>> declarations so that I wouldn't have to add Eigen as a dependency for
>>>> VTK.
>>>>
>>>> I've tried doing this myself, but it looks like the last three
>>>> template arguments are dependent on compiler options. Is there a trick
>>>> to forward declaring this class?
>>>>
>>>> Thanks,
>>>>
>>>> Dave
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>
>>
>>
>
>
>