I understand now, thank you. Would you see if the change in the svn
head fixes the problem?
Thanks,
---
Roy
On Wed, 23 Mar 2011, Loris Nagler wrote:
> Sorry, in my previous post I was not too specific. I need to write out a
> complex valued vector. I write out my data using MeshData::write(**),
> which in my case calls MeshData::write_xdr(**), which calls Xdr::data(**).
> By doing so in the former libmesh version, the vector entries were
> written out by this loop:
>
> for (unsigned int i=0; i<v.size(); i++)
> {
> libmesh_assert (out.get() != NULL); libmesh_assert (out->good());
> OFSNumberscientific(*out,17,v[i]) << " ";
> }
>
> Hence, two consecutive complex valued entries were written out like this:
> 1.00000000000000000e+00 2.00000000000000000e+00 3.00000000000000000e+00
> 4.00000000000000000e+00
>
> Now, the call of MeshData::write(**) leads to the following loop
> (xdr_cxx.C line 522):
>
> for (unsigned int i=0; i<a.size(); i++)
> {
> libmesh_assert (out.get() != NULL); libmesh_assert (out->good());
> this->do_write(a[i]);
> }
> with
> do_write(std::complex<T>& a) { *out << a.real() << "\t " << a.imag(); }
>
> The same two consectutive entries as above are now written out like this:
> 1 23 4
> (no space between 2 and 3, no scientific notation)
>
> But you're right, the Xdr::data_stream(**) function avoids this problem,
> however, by calling MeshData::write(**) I don't reach
> Xdr::data_stream(**) but only Xdr::data(**). In example ex10, the data
> is written out over EquationSystems::write(**), which in the end leads
> to System::write_SCALAR_dofs(**) which calls Xdr::data_stream(**) and
> everything is fine.
>
> As far as I can tell, adding << " "; after a.imag() should do the trick.
> Or am I missing something? Sorry to bother you!
>
> best reagrds
> Loris
>
> On 03/23/2011 04:14 PM, Roy Stogner wrote:
>>
>> On Wed, 23 Mar 2011, Loris Nagler wrote:
>>
>>> I recently updated libmesh to version 0.7.0.4. In a former version,
>>> the function
>>>
>>> void Xdr::data (std::complex<double>& a, const char* comment)
>>>
>>> which is defined in xdr_cxx.C, wrote out complex numbers like this:
>>>
>>> *out << a.real() << "\t " << a.imag() << "\t " << comment << '\n';
>>>
>>> whereas now they are written out like this (now using
>>> do_write(std::complex<T>& a) ):
>>>
>>> *out << a.real() << "\t " << a.imag();
>>
>> Take a look at Xdr::data again in the new version - it calls
>> do_write() to do the above, but then adds the final
>> tab/comment/newline itself.
>>
>> For Xdr::data(vector<*>), on the other hand... it does look like we're
>> missing separators between data points... but we manage to avoid
>> triggering it because all the library code uses the older
>> Xdr::data_stream method instead.
>>
>>> Due to a missing space or tab after a.imag(), the imaginary part of
>>> one entry is directly attached to the real part of the next entry.
>>> No wonder my input files are pretty messed up since the update ;) .
>>
>> Could you give us an example, or better yet a test case that generates
>> an example? I tried to replicate the problem with ex10 using
>> --enable-complex, but saved_solution.xda from that run looks just
>> fine.
>> ---
>> Roy
>>
>>
>> ------------------------------------------------------------------------------
>> Enable your software for Intel(R) Active Management Technology to meet the
>> growing manageability and security demands of your customers. Businesses
>> are taking advantage of Intel(R) vPro (TM) technology - will your software
>> be a part of the solution? Download the Intel(R) Manageability Checker
>> today! http://p.sf.net/sfu/intel-dev2devmar
>>
>>
>> _______________________________________________
>> Libmesh-users mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>>
>
> ------------------------------------------------------------------------------
> Enable your software for Intel(R) Active Management Technology to meet the
> growing manageability and security demands of your customers. Businesses
> are taking advantage of Intel(R) vPro (TM) technology - will your software
> be a part of the solution? Download the Intel(R) Manageability Checker
> today! http://p.sf.net/sfu/intel-dev2devmar
> _______________________________________________
> Libmesh-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>
------------------------------------------------------------------------------
Enable your software for Intel(R) Active Management Technology to meet the
growing manageability and security demands of your customers. Businesses
are taking advantage of Intel(R) vPro (TM) technology - will your software
be a part of the solution? Download the Intel(R) Manageability Checker
today! http://p.sf.net/sfu/intel-dev2devmar
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
