The RunPythonScript should not have special behaviour for a WorkspaceGroup input

[Russell Taylor]

Basically, Algorithm::checkGroups() needs to be overridden to return false, but it's a Python algorithm so I don't know how to do that.

[Russell Taylor]

My thinking has moved on since I wrote this. My feeling is that (unless there's a technical reason why it can't be) RunPythonScript should be converted to a C++ algorithm. If you don't see why that can't be the case, do send the ticket back my way.

[Martyn Gigg]

I'll have a look. I think Janik tried initially and there was a problem but I'm not sure what it was

[Martyn Gigg]

Batch move to 2.7

[Martyn Gigg]

Rewrite RunPythonScript in C++.

This is a core algorithm for the LiveData so needs away from possible user modification and allows custom behaviour for WorkspaceGroups. Refs #7019

Changeset: fb72c2375c708111c4cc7c09abd0d4e88aabc6f4

[Martyn Gigg]

Branch: feature/7019_cpp_runpythonscript

Tester: This is best tested through the StartLiveData algorithm: http://www.mantidproject.org/StartLiveData. The behaviour should be the same as before with the exception that the algorithm will only be run once for a WorkspaceGroup input rather than once per member.

[Russell Taylor]

I'll grab this for testing as I have a strong interest in making certain all the live stuff is stable. I won't test it until tomorrow though.

[Martyn Gigg]

There seems to be a problem with one of the tests on fedora 18 that I'm looking into at the moment. If I don't sort it by the end of the day I'll reopen the ticket.

[Martyn Gigg]

Disable a test while it is diagnosed. Refs #7019

Changeset: 0cfb9b2573ff9fabe95317dad2ab12313022f7a5

[Martyn Gigg]

Switch from Poco->boost regex engine.

There is a bug in Poco::RegularExpression that causes and out of memory error on some regular expressions. The boost version does not suffer from this and actually has cleaner syntax for the eol conversion. Refs #7019

Changeset: 103522339b9e51a828268d30acef5fe70ceced89

[Martyn Gigg]

Switch out usage of Poco regex for boost in GUI ScriptCode class

Not technically related to this ticket but the regexes were identical so the ScriptCode class would have suffered the same out of memory error. Refs #7019

Changeset: 65b17b38e13df051386d457e53933ea2a4b75c76

[Russell Taylor]

When I try this out using the DGS Reduction GUI (see the to_live_script method in scripts/Interface/reduction_gui/reduction/inelastic/dgs_reduction_script.py), I get the following back the first time MonitorLiveData goes through it (i.e. the second time overall as StartLiveData already ran it):

MonitorLiveData-[Error] MonitorLiveData: UNKNOWN Exception is caught in exec()

This comes out of line 177: boost::python::exec(script.c_str(), globals, locals);, but that's all I could figure out (the exception appears not to be derived from std::exception or boost::exception).

The code that's being run in case you can spot anything wrong with it is:

from mantid.simpleapi import *
def PyExec(input=None,output=None):
  if mtd.doesExist("live_spe"): input.run().addProperty("EnergyRequest",mtd["live_spe"].getRun()["EnergyRequest"],True)
  DgsReduction(SampleInputWorkspace=input,OutputWorkspace=output,UseIncidentEnergyGuess=True,IncidentBeamNormalisation="None",)
  return input,output
input,output = PyExec(input,output)

Note that the "if mtd.doesExist("live_spe"):" test will be true the second time around where it was false the first time.

[Martyn Gigg]

Backwards-compatible check for Python none object in C++.

Refs #7019

Changeset: 766c49f0676ce1172cc446163137094efb550b32

[Martyn Gigg]

Refactor Python error handling code into separate header.

Refs #7019

Changeset: 91b512e51769774e6ab29cfb100ab535e7c2491f

[Martyn Gigg]

Translate Python errors to C++ exception in RunPythonScript

Refs #7019

Changeset: 0dad518a5808683b3832b471ca138646a8f82d91

[Martyn Gigg]

Adjust exported WorkspaceCreationHelper module to only uses interfaces

The unit tests will now behave much more like a real situation where the concrete types are not exposed. Refs #7019

Changeset: ec34864fb1ba9b97de2b363cc14ad7cd68ad9eca

[Martyn Gigg]

Ensure RunPythonScript sees proper workspace types.

Refs #7019

Changeset: 003fb4f55fd9da88daa38db2fa72dc228b3a8328

[Martyn Gigg]

This went to the wrong ticket:

Update class name as a result of change elsewhere. Refs #7109

Changeset: 4f9825ae34e42260919bb8a626864cd20e80bba9

[Martyn Gigg]

This should be fixed now and ready to test again.

[Martyn Gigg]

Only gcc 4.6 accepts typename outside template. Refs #7019

Changeset: f829c2d94371d7f07fe66b9ac9321127a585eb93

[Martyn Gigg]

Please leave this for Russell to test

[Russell Taylor]

All my testing checked out OK now. The DGS reduction GUI runs OK, as do other live scripts I have. Sensible messages come back if I insert deliberate errors into scripts. The algorithm now only runs once if it gets a group as input.

I did observe some strange behaviour for which tickets #8007 & #8008 have been created, though these are not regressions due to this work.

There is one thing I noticed though - PythonInterface/test/cpp/RunPythonScriptTest.h is not being build and run because it's not included in the CMakeLists. I'm reopening so that can be fixed (some method calls in the test need changing so that it will compile).

[Martyn Gigg]

Add C++ RunPythonScript test to CMakeLists...

and fix the test. Refs #7019

Changeset: 53855f98a0b92bf3ed51000f47f48f7286d80b49

[Martyn Gigg]

I'd been focussing on the RunPythonScript test written in Python as the C++ is basically there so that there is a matching test.

It should be ready to test again now.

[Russell Taylor]

Merge remote branch 'origin/feature/7019_cpp_runpythonscript'

[Russell Taylor]

BTW, there is one way in which this behaves differently to before - it won't allow you not to have an output workspace whereas before you could (e.g. your processing script could just be something like print "Hello"). However, I don't think the previous behaviour was correct.

[Stuart Campbell]

This ticket has been transferred to github issue ​7865