Replace the ".uid" in polarpoissonlikesolver.hpp.

[PaulineVidal]

I noticed that the static_analysis often fails on the polarpoissonlikesolver.hpp file because of the remaining .uid(). I tried to replace them.

[tpadioleau]

Indeed this should silence the warnings but I am afraid we still have the same underlying problem. Calling the constructor https://ddc.mdls.fr/classddc_1_1DiscreteElement.html#af6776e89752caf2077e70f690502a5e5 should be done as a last resort. One should see uid() and this constructor as DDC exits.

Isn't there a function in splines that could get you the Idx<RCellDim> for example ? Is RCellDim the dimension of knots ?

[EmilyBourne]

RCellDim is the dimension of the cells on which the bsplines are defined. It is declared here:

gyselalibxx/src/pde_solvers/polarpoissonlikesolver.hpp

Lines 80 to 82 in 0647e3d

	struct RCellDim
	{
	};

As it is inside this class I don't think it is dangerous to call this constructor.

That said in the places where - Idx<RCellDim>(0) is used it seems to me that some of the operations could be carried out directly with an Idx.

[EmilyBourne]

Suggested change

	assert((ib_test_r - Idx<RBasisSubset>(0)).value() < BSplinesR::degree() + 1);
	assert(ib_test_r < Idx<RBasisSubset>(BSplinesR::degree() + 1));

[EmilyBourne]

Are integers needed here? Is it not possible to do something like:

Suggested change

	const std::size_t idx_trial_r_val = (idx_trial_r - m_idx_bsplines_r_front).value();
	const std::size_t idx_trial_theta_val
	= (idx_trial_theta - m_idx_bsplines_theta_front).value();
	const std::size_t first_overlap_element_r_val
	= idx_trial_r_val < BSplinesR::degree()
	? 0
	: idx_trial_r_val - BSplinesR::degree();
	const Idx<RCellDim> first_overlap_element_r(first_overlap_element_r_val);
	const Idx<RCellDim> first_overlap_element_r(
	idx_trial_r < Idx<RCellDim>(BSplinesR::degree())
	? Idx<RCellDim>(0)
	: idx_trial_r - BSplinesR::degree());

?
Or is idx_trial_r the wrong type for this?

[EmilyBourne]

If you want to fix the warnings in this file then the 2 remaining warnings (which are easier to fix) should be addressed.

That said I was under the impression that we had agreed to leave this file as it is (with the warnings) for now as this code should be replaced in the coming months by a new implementation written by @etiennemlb . I think that a clean solution to these warnings involves more work than simply writing (ddc::select<BSplinesR>(idx_test) - m_idx_bsplines_r_front).value() instead of ddc::select<BSplinesR>(idx_test).uid() to ensure that the index types are used as intended