Use ConstraintLine rather than redefining data structures

[kronbichler]

While looking at #14905, @bergbauer and I observed some weird std::tuple data structures in the code for AffineConstraints::make_consistent_in_parallel that duplicates the content of AffineConstraints::ConstraintLine. I see no reason not to use this field, and it in fact leads to much more understandable code. For the record, this was introduced in #12430, and I couldn't find a discussion there about the type. Note that also is_consistent_in_parallel uses the more straight-forward type,

dealii/include/deal.II/lac/affine_constraints.templates.h

Lines 114 to 130 in a1b1356

	// Helper to return a ConstraintLine object that belongs to row @p row.
	// If @p row is not constrained or not stored locally, return an empty
	// constraint object that would correspond to a zero constraints
	auto get_line = [&](const size_type line_n) -> ConstraintLine {
	const size_type line_index = calculate_line_index(line_n);
	if (line_index >= lines_cache.size() ||
	lines_cache[line_index] == numbers::invalid_size_type)
	{
	const ConstraintLine empty = {line_n, {}, 0.0};
	return empty;
	}
	else
	return lines[lines_cache[line_index]];
	};
	// identify non-owned rows and send to owner:
	std::map<unsigned int, std::vector<ConstraintLine>> to_send;

[kronbichler]

Note that this will conflict with #14905, so we need to take care when the first of the two PRs is merged.

[peterrum]

observed some weird std::tuple data structures in the code for AffineConstraints::make_consistent_in_parallel that duplicates the content of AffineConstraints::ConstraintLine.

The reason is that the author did not know ConstraintLine^^

Note that this will conflict with #14905, so we need to take care when the first of the two PRs is merged.

I would suggest to merge this one first. The other one will still take some time.

[bergbauer]

I would suggest to merge this one first. The other one will still take some time.

Fine with me. This will make #14905 much easier to read and understand! Thanks @kronbichler

[bangerth]

Nice work!