ExtendedUnit Cell Defects

[Owen Arnold]

Apply fixes below.

First of all, the results in the test are wrong:
[1,0,0] and [0,0,1] are in the horizontal plane (in Mantid convention z direction and x direction). Then the perpendicular to the plane in the up direction is [0,-1,0]. So [1,1,1] is below the plane, and the angle is arcos(-1./sqrt(3.)), not arcos(1./sqrt(3.)).
Minor detail: use numpy.degrees(angle) to get the angle in degrees
Here are the steps that I would use to get this algorithm:
1. Use OrientedLattice instead of Unit Cell
2. Set the U matrix using the two reflections in the plane using ol.setUFromVectors. That function takes any combination of V3Ds, lists, numpy arrays
3. Get the UB matrix and calculate Q: ol.getUB().dot(array(desiredReflection))
4. Calculate the angles a_i=degrees(arccos(Q_i/|Q|)), with i=x,y,z. Note that the z is along beam, x perpendicular to the beam in horizontal plane, and y is vertically up.
The sign thing does not make sense in 3D. a_i will have values from 0 to 180 degrees

[Owen Arnold]

Following a helpful discussion with Andrei, we have come to the following decisions.

• ExtendedUnitCell is redundant and is providing incorrect results anyway. It also isn't using the full UBMatrix. Remove all code and tests for this.
• Pseudo code below does what we want
• The direction cosine calculation below should live on V3D as in V3D::directionAngles(), make it possible to obtain results in degrees or radians
• Expose methods above to python

V3D directionAngles(const V3D & hkl)
{

Q = qFromHKL(hkl);

a_1 = a_i=degrees(arccos(Q[0]/Q.norm()))
a_2 = a_i=degrees(arccos(Q[1]/Q.norm()))
a_3 = a_i=degrees(arccos(Q[2]/Q.norm()))

return V3D(a_1, a_2, a_3);

}

[Owen Arnold]

refs #10019. Extend V3D and test

Changeset: 6b9bdba84d1835d08d39a126b71d404046180fa4

[Owen Arnold]

refs #10019. Expose methods to python and test

Changeset: 5349aa950bde5d7757c69190ae9eeb4b587b9a73

[Owen Arnold]

refs #10019. Remove ExtendedUnitCell class

Changeset: bc19bbdfe5ca637555a284b4c5da2b8eabf05f62

[Owen Arnold]

I suggest that Andrei tests this

Tester: I have provided unit tests throughout for the new functionality. The following would also help.

        
import math
from mantid.geometry import OrientedLattice
from mantid.kernel import V3D     
        
ol = OrientedLattice(1,1,1,90,90,90)
v1 = V3D(1,0,0)
v2 = V3D(0,0,1)
ol.setUFromVectors(v2, v1)
q = ol.qFromHKL(V3D(3,1,1))

angles = q.directionAngles(False)
print "angles: ", angles
print "expected angles: ", [math.acos(math.sqrt(9)/math.sqrt(11)),math.acos(math.sqrt(1)/math.sqrt(11)), math.acos(math.sqrt(1)/math.sqrt(11))] 

[Owen Arnold]

refs #10019. Minor modifications.

Changeset: 5ea0c2f2f03e24bb18e33ac6cd05f2b495d15ea0

[Andrei Savici]

For the python exports for qFromHKL and hklFromQ, one should be able to use lists or numpy arrays, similar to the case of setUFromVectors (see lines 32 to 35 in mantid/Code/Mantid/Framework/PythonInterface/mantid/geometry/src/Exports/OrientedLattice.cpp)

[Owen Arnold]

refs #10019. More helpful about python arguments.

Following Andrei's suggestion. Internally convert to V3D from other types. I've also enabled the default choice of inDegrees for the V3D::directionAngles from the python side. All changes are unit tested.

Changeset: de3e56ae084c3c48ac75dd950b4f2e5803e4330b

[Andrei Savici]

Merge remote-tracking branch 'origin/feature/10019_direction_angles'

Full changeset: ​9d8db81005929e3eb59f2f725983a7d76931336a

[Stuart Campbell]

This ticket has been transferred to github issue ​10861