PyAT Matlab comparison test failures about at HOT 4 CLOSED

Hi @T-Nicholls. Are you sure that:

• you use the last version of AT
• you call the python get_mcf with ddp=1.E-6, which is what Matlab uses
• you run on the same machine

I get an agreement with 15 digits for dp=0, and with 10 digits for dp!= 0

 0.00  8.506669244225616e-05
 0.01  8.820785279e-05
-0.01  8.2317973893e-05

Which agrees with your Matlab results !!!

(running on MacOSX 10.14, python 2.7 and 3.6, Matlab R2018a)

from at.

@lfarv my apologies, a change I made caused the difference.

Ok, I have found the source of the problem:

• I was using a branch where the py_hmba fixture had a scope session, i.e.

  @pytest.fixture(scope='session')
  def py_hmba():
      return Lattice(load_mat(utils.hmba_ring, key='RING'), keep_all=True)
  

• We turn on radiation in the lattice in test_ohmi_envelope:

  # Python call
  py_lattice.radiation_on()
  py_emit0, py_beamdata, py_emit = physics.ohmi_envelope(py_lattice, refpts)
  

• The same lattice is used and the radiation is not turned off. Adding a line to turn it off fixes the issue:

  # test momentum compaction factor
  py_lattice.radiation_off()
  py_mcf = py_lattice.get_mcf(dp, ddp=1.E-6)  # Matlab uses ddp=1.E-6
  

Should we be ensuring the radiation state of the lattice, as we do for linopt and ohmi_envelope, or is get_mcf designed to be run in both states and the difference to be expected?

from at.

get_mcf is based on find_orbit4. Strictly speaking, this is valid if there is no longitudinal motion (radiation off), though many people use it in any condition (as you could see, the results are not very different). The get_mcf function works for any sequence of elements and cannot be protected, but the get_mcf method of the Lattice object could be, as it is done for linopt.

from at.

@lfarv thank you for the clarification, I will protect the get_mcf method on Lattice then.

from at.