How can I resize a std::vector<Functions<dim>::ParsedFunction>?

[Alex Zimmerman]

Fundamentally I am trying to allow for a variable number of ParsedFunction objects to be specified in a parameter input file. Maybe there is a better approach which circumvents my issue below. I can continue my work for some time with this being a constant; but as soon as I want to extend to 3D or different geometries, this will be a roadblock.

I've documented my issue on my GitHub repostiory: https://github.com/alexanderzimmerman/nsb-pcm/issues/10 . Here's a copy of the text from my issue, which pretty much explains it:

Attempting to resize or push_back into a std::vector<Functions::ParsedFunction> throws an error within TBB and/or mu::Parser, e.g.:

In file included from /usr/include/c++/4.8/vector:64:0,

from /usr/include/c++/4.8/bits/random.h:34,

from /usr/include/c++/4.8/random:50,

from /usr/include/c++/4.8/bits/stl_algo.h:65,

from /usr/include/c++/4.8/algorithm:62,

from /home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/bundled/tbb/concurrent_vector.h:48,

from /home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/bundled/tbb/enumerable_thread_specific.h:32,

from /home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/base/thread_local_storage.h:23,

from /home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/base/logstream.h:23,

from /home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/lac/vector.h:21,

from /mnt/c/Users/Alex/UbuntuShared/nsb-pcm/tests/peclet_data.cc:1:

/usr/include/c++/4.8/bits/stl_vector.h: In instantiation of ‘std::_Vector_base<_Tp, _Alloc>::~_Vector_base() [with _Tp = mu::Parser; _Alloc = std::allocatormu::Parser]’:

/usr/include/c++/4.8/bits/stl_vector.h:416:33: required from ‘std::vector<_Tp, _Alloc>::~vector() [with _Tp = mu::Parser; _Alloc = std::allocatormu::Parser]’

/home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/bundled/tbb/enumerable_thread_specific.h:659:17: required from ‘void tbb::interface6::internal::ets_element<U, ModularSize>::unconstruct() [with U = std::vectormu::Parser; long unsigned int ModularSize = 24ul]’

/home/zimmerman/installed/deal.ii-candi/deal.II-v8.4.2/include/deal.II/bundled/tbb/enumerable_thread_specific.h:729:17: required from ‘void tbb::interface6::enumerable_thread_specific<T, Allocator, ETS_key_type>::unconstruct_locals() [with T = std::vectormu::Parser; Allocator = tbb::cache_aligned_allocator<std::vectormu::Parser >; tbb::ets_key_usage_type ETS_key_type = (tbb::ets_key_usage_type)1u]’

Because of this issue, the number of boundaries on the domain must be known at compile time (i.e. they can't be set in the input parameter file), so that the proper number of ParsedFunction objects can be allocated.

Maybe ParsedFunction does not mean these requirements: http://stackoverflow.com/questions/12251368/type-requirements-for-stdvectortype

This might be a bug, or rather missing functionality, in ParsedFunction. But I'm still fairly new to deal.II and to C++; so I think it's more likely that I'm doing something wrong.

The repository includes a branch with test code that reproduces the compiler error. The test source code is linked in the issue (on the bugs branch).

Any ideas?


Thanks,

Alex

[Alex Zimmerman]

Correction: Of course I mean a std::vector<dealii::Functions::ParsedFunction<dim>>. My question title has a <dim> in the wrong place.

[luca.heltai]

Hi Alex,

I’d use a vector of

std::vector<std_cxx11::shared_pointer<ParsedFunction<dim> > > v;

which are light objects, and can be resized and reshaped. Whenever you do a push_back, you’d have to use

v.push_back(std_cxx11::shared_pointer<ParsedFunction<dim>(new ParsedFunction<dim>(…) );

then

v[i]->value(…)

would work.

> --
> The deal.II project is located at http://www.dealii.org/
> For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en
> ---
> You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to dealii+un...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Alex Zimmerman]

Luca,

Thanks for the idea. Conceptually I don't understand how a shared_ptr solves my problem. I can resize std::vector<std::shared_ptr<dealii::Functions::ParsedFunction<dim>>>; but all of the shared pointers are empty. I imagine that they are supposed to point to ParsedFunction<dim> objects that I've instantiated somewhere.

That being said, I think you are leading me in the correct direction. In the following test code, "default.prm" is successfully generated with the four parsed function sections. It seg faults when I call parse_parameters; so I'm still debugging:

#include <deal.II/base/parsed_function.h>

#include <iostream>
#include <assert.h>

int main(int /*argc*/, char** /*argv*/)
{
    const unsigned int dim = 2;
    const unsigned int function_count = 4;

    std::vector<std::shared_ptr<dealii::Functions::ParsedFunction<dim>>>
        function_pointers;
   
    function_pointers.resize(function_count);

    dealii::ParameterHandler prm;

    for (unsigned int f = 0; f < function_count; ++f)
    {
        prm.enter_subsection("parsed_function_"+std::to_string(f));
        {
            function_pointers[f]->declare_parameters(prm);
        }
        prm.leave_subsection();
    }

    prm.read_input("default.prm", false, true);

    for (unsigned int f = 0; f < function_count; ++f)
    {
        prm.enter_subsection("parsed_function_"+std::to_string(f));
        {
            function_pointers[f]->parse_parameters(prm);
        }
        prm.leave_subsection();
    }

    return 0;
}

Thanks,

Alex

[Alex Zimmerman]

Also for the record, I verified that the test passes if I use a std::vector<dealii::Functions::ParsedFunction<dim>> without resizing: https://github.com/alexanderzimmerman/nsb-pcm/blob/bugs/tests/bug_10_pass.cc

[Timo Heister]

Your call to resize() will create 4 shared_ptrs, but they are pointing
to nothing (NULL) because you are never allocating any objects and
assigning them.

Think of this example:
std::vector<int*> v;
v.resize(4);
now v[0] is a pointer to an int, but it is NULL unless you do something like
v[0] = new int;

On Wed, Mar 22, 2017 at 3:09 PM, Alex Zimmerman
<zimmerma...@gmail.com> wrote:
> Also for the record, I verified that the test passes if I use a
> std::vector<dealii::Functions::ParsedFunction<dim>> without resizing:

> https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_alexanderzimmerman_nsb-2Dpcm_blob_bugs_tests_bug-5F10-5Fpass.cc&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=v1-30ZnlLX9dscEMGIPAxToOfyXrgOcTjUUS54eVdJA&e=

>>> > https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_alexanderzimmerman_nsb-2Dpcm_issues_10&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=a-21yeSa2EGcnF7p79Efq_xBzJ8qavKiAtJHROzgbX0&e= . Here's a copy of

>>> > https://urldefense.proofpoint.com/v2/url?u=http-3A__stackoverflow.com_questions_12251368_type-2Drequirements-2Dfor-2Dstdvectortype&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=vWPe9Cb9rr1lr_D0ElWWOlaFRepLECEGgOsA5om9cYM&e=

>>> >
>>> >
>>> > This might be a bug, or rather missing functionality, in
>>> > ParsedFunction. But I'm still fairly new to deal.II and to C++; so I think
>>> > it's more likely that I'm doing something wrong.
>>> >
>>> > The repository includes a branch with test code that reproduces the
>>> > compiler error. The test source code is linked in the issue (on the bugs
>>> > branch).
>>> >
>>> > Any ideas?
>>> >
>>> >
>>> > Thanks,
>>> >
>>> > Alex
>>> >
>>> > --

>>> > The deal.II project is located at https://urldefense.proofpoint.com/v2/url?u=http-3A__www.dealii.org_&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=tVQj2WHLTosvw2iFgktRW28HcOYr79Dy3FU9NdDBA5Y&e=

>>> > For mailing list/forum options, see

>>> > https://urldefense.proofpoint.com/v2/url?u=https-3A__groups.google.com_d_forum_dealii-3Fhl-3Den&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=CX85qMkHRHJYO670sdPrmLde-ikUtH3C6pXznMZchKQ&e=

>>> > ---
>>> > You received this message because you are subscribed to the Google
>>> > Groups "deal.II User Group" group.
>>> > To unsubscribe from this group and stop receiving emails from it, send
>>> > an email to dealii+un...@googlegroups.com.

>>> > For more options, visit https://urldefense.proofpoint.com/v2/url?u=https-3A__groups.google.com_d_optout&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=oM93B3-DRw_tDnTE4Ftec3BqqGO3vpgbmK9PBEm3aXc&e= .
>>>
> --
> The deal.II project is located at https://urldefense.proofpoint.com/v2/url?u=http-3A__www.dealii.org_&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=tVQj2WHLTosvw2iFgktRW28HcOYr79Dy3FU9NdDBA5Y&e=

> For mailing list/forum options, see

> https://urldefense.proofpoint.com/v2/url?u=https-3A__groups.google.com_d_forum_dealii-3Fhl-3Den&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=CX85qMkHRHJYO670sdPrmLde-ikUtH3C6pXznMZchKQ&e=

> ---
> You received this message because you are subscribed to the Google Groups
> "deal.II User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to dealii+un...@googlegroups.com.

> For more options, visit https://urldefense.proofpoint.com/v2/url?u=https-3A__groups.google.com_d_optout&d=DwIFaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=4k7iKXbjGC8LfYxVJJXiaYVu6FRWmEjX38S7JmlS9Vw&m=bMNkhuzVlxuFjrw_sjeYQVtM1sMLS6wmKIkP9ILH7d4&s=oM93B3-DRw_tDnTE4Ftec3BqqGO3vpgbmK9PBEm3aXc&e= .



--
Timo Heister
http://www.math.clemson.edu/~heister/

[Alex Zimmerman]

Timo, thanks for the extra clarification. As I mentioned in my reply to Luca, I don't understand why declare_parameters appeared to succeed in my test, which failed at parse_parameters.

My issue is solved using the vector<shared_ptr<ParsedFunction>>.push_back(new ParsedFunction) approach. My attempt at the resize approach isn't working; but push_back is fine. 

Related test code is here: https://github.com/alexanderzimmerman/nsb-pcm/tree/bugs/tests
ctest passes all tests except for issue_10_resize.cc. I'm guessing that I have the syntax wrong somewhere.

Thanks for the help!

[Alex Zimmerman]

Also for the record, here's a copy of issue_10_push_back.cc from that repository, which is the example of how to solve my issue:

#include <deal.II/base/parsed_function.h>

#include <iostream>
#include <assert.h>

int main(int /*argc*/, char** /*argv*/)
{
    const unsigned int dim = 2;
    const unsigned int function_count = 4;

    std::vector<std::shared_ptr<dealii::Functions::ParsedFunction<dim>>>
        function_pointers;

    dealii::ParameterHandler prm;
    dealii::Point<dim> point;
    double value;

    for (unsigned int f = 0; f < function_count; ++f)
    {

        function_pointers.push_back(
            std::shared_ptr<dealii::Functions::ParsedFunction<dim>>(new dealii::Functions::ParsedFunction<dim>()));

           
        prm.enter_subsection("parsed_function_"+std::to_string(f));
        {
            function_pointers[f]->declare_parameters(prm);

            function_pointers[f]->parse_parameters(prm);
        }
        prm.leave_subsection();

        function_pointers[f]->value(point, value);

        std::cout << "f(" << point << ") = " << value << std::endl;
    }

    return 0;
}

[luca.heltai]

Resizing is fine. It is the equality that won’t work:

function_pointers[f] =

std::shared_ptr<dealii::Functions::ParsedFunction<dim>>(new dealii::Functions::ParsedFunction<dim>()));

is the correct way to do it.

L.

> The deal.II project is located at http://www.dealii.org/
> For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en

> ---
> You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to dealii+un...@googlegroups.com.

> For more options, visit https://groups.google.com/d/optout.

[Alexander Gary Zimmerman]

Luca, I finally got around to testing your fix, and now resize works as you said it would :) Thanks! Of course it was really easy to do, but I had moved on with another branch. Sorry for the delay.