tangential position range inconsistency

[SomayehSaghamanesh]

There are two different definitions (as below), which result in inconsistency in tangential position range. This is problematic at least in cases handling full tangential positions.
In ProjDataInfo.cxx:
min_tangential_pos_num = -(num_tang_poss / 2);
max_tangential_pos_num = min_tangential_pos_num + num_tang_poss - 1;

In ProjDataInfoCylindricalNoArcCorr.cxx and ProjDataInfoGenericNoArcCorr.cxx :
const int min_tang_pos_num = -(num_detectors / 2) + 1;
const int max_tang_pos_num = -(num_detectors / 2) + num_detectors;

[robbietuk]

The class inheritance is as follows

I suggest using this

STIR/src/buildblock/ProjDataInfo.cxx

Lines 123 to 129 in e168b07

	void
	ProjDataInfo::set_num_tangential_poss(const int num_tang_poss)
	{
	min_tangential_pos_num = -(num_tang_poss / 2);
	max_tangential_pos_num = min_tangential_pos_num + num_tang_poss - 1;
	}

in the child classes for consistency?

Maybe worth noting, the definition of the max/min tang poss in ProjDataInfoCylindricalNoArcCorr is very old (circa 2003). Somethings may break because they were built upon this methodology.

[SomayehSaghamanesh]

Yes, I have already followed it in PR #1508, changed two classes to be consistent with ProjDataInfo.cxx.

[robbietuk]

I am sorry, I reread the full function. These values are used for slightly odd checks in initialise_uncompressed_view_tangpos_to_det1det2. I'm not sure what these

STIR/src/buildblock/ProjDataInfoGenericNoArcCorr.cxx

Lines 152 to 163 in e168b07

	const int min_tang_pos_num = -(num_detectors / 2) + 1;
	const int max_tang_pos_num = -(num_detectors / 2) + num_detectors;
	if (this->get_min_tangential_pos_num() < min_tang_pos_num || this->get_max_tangential_pos_num() > max_tang_pos_num)
	{
	error("The tangential_pos range (%d to %d) for this projection data is too large.\n"
	"Maximum supported range is from %d to %d",
	this->get_min_tangential_pos_num(),
	this->get_max_tangential_pos_num(),
	min_tang_pos_num,
	max_tang_pos_num);
	}

checks are, but since both classes' methods are a copy and paste of one another. Your fix in #1508 seems correct to me.

[SomayehSaghamanesh]

Just not sure why 8 tests (including FBP, OSMAPOSL, etc) are failed after these changes. Haven't gone through them yet. Any comment is appreciated.