Revise the file manager for on-the-fly binning.

[legordian]

Revise the file manager and all the attached scripts to implement a binning logic including on-the-fly binning. At the present stage, this is in a rather rough state: all the scripts should work, but the functionality has not been fully exploited yet. In particular, the fitting scripts in this version are supplied with a integer bin id, which corresponds to an integral bin. This should be replaced by an interface actually passing in the bin and the script system using the integral bins to create an new integral for that bin, if possible. Also, a number of checks are not implemented yet, e.g. it is not checked if every event is actually in a bin and the metadata compatibility is not verified. However, this changeset should provide a good basis to implement these things.

[suhlatwork]

• the conversion of MeV to GeV also needs to be done for:
  • userAnalysisWorkspace/3pi.--+/keyfiles/wavelist.compass.2004
  • userAnalysisWorkspace/3pi.--+/keyfiles/wavelist.compass.2008.42waves
  • userAnalysisWorkspace/3pi.--+/keyfiles/wavelist.compass.2008.53waves
  • userAnalysisWorkspace/3pi.-00/keyfiles/wavelist.compass.2004
  • userAnalysisWorkspace/3pi.-00/keyfiles/wavelist.compass.2008.42waves
  • userAnalysisWorkspace/3pi.-00/keyfiles/wavelist.compass.2008.53waves
  • userAnalysisWorkspace/3pi.-00/keyfiles/wavelist.compass.2008.88waves
  • storageFormats/convertEventFile.py (argument help text, example for --binning)
  • genPseudoData.py (should not multiply by 1000 to get the mass to search the fit result)
  • partialWaveFit/pwaLikelihood.cc (adjust messages showing thresholds)
  • highLevelInterface/pwaNloptFit.cc (adjust message showing mass range of fit)
  • highLevelInterface/pwaFit.cc (adjust message showing mass range of fit)
• my comments are based on commit 06d6d7a7d86a5fe039866fe5c1f8fef47bdb662b
• no actual testing done so far

[legordian]

All implemented. Note that merging this pull request will also merge #167, so either merge that one first or only merge this one if #167 is also worthy of inclusion.

[suhlatwork]

• file binning does not work as the each file is assigned to an individual bin, so there is one bin containing only real data, one bin with only generated PS, and one bin with only accepted PS, despite the binning maps being the same in these files. A fit cannot be executed.
• calcIntegrals.py checks whether all events were used to calculate the amplitudes, this makes usage of the -n switch to calcAmplitudes.py impossible (even if also calcIntegrals.py is called with -n. -n should probably be removed from both scripts
• in my test case of two event files for two neighbouring mass ranges (1.22 to 1.26 and 1.26 to 1.3) over the full t' range (0.1 to 1.0, divided into eight bins for the integrals) (x3 for real data, generated PS, accepted PS) the integrals for the higher mass bin (1.26 to 1.3) also contain the event metadata of the lower mass bin (1.22 to 1.26). I did not check whether the integrals also include events from this bin. All integral files for the lower mass bin also contain only the metadata of the lower mass bin. This probably also cases warnings (??? bool rpwa::ampIntegralMatrixMetadata::addKeyFileContent(const string&) [/nfs/freenas/tuph/e18/project/compass/analysis/suhl/rootpwa/rootpwa-devel/decayAmplitude/ampIntegralMatrixMetadata.cc:168]: warning: cannot add keyfile content: already present.) when running calcIntegrals.py. Fitting of the lower mass bin works, fitting of the higher mass bin crashes with a segfault:

#6  0x00007f4d626723f9 in SetBranchAddress<double> (ptr=0x0, add=0x423ba98, bname=0x3fac8a8 "mass", this=0x4141c70) at /nfs/mnemosyne/sys/slc6/sw/root/x86-64/6.04.02/root-6.04.02/include/TTree.h:517
#7  rpwa::pwaLikelihood<std::complex<double> >::setOnTheFlyBinning (this=this
entry=0x3c83f80, binningMap=std::map with 2 elements, evtMetas=std::vector of length 2, capacity 2 = {...}) at /nfs/freenas/tuph/e18/project/compass/analysis/suhl/rootpwa/rootpwa-devel/partialWaveFit/pwaLikelihood.cc:1327
#8  0x00007f4d677dc4ca in (anonymous namespace)::pwaLikelihood_setOnTheFlyBinning (self=..., pyBinningMap=..., pyEvtMetas=...) at /nfs/freenas/tuph/e18/project/compass/analysis/suhl/rootpwa/rootpwa-devel/pyInterface/partialWaveFit/pwaLikelihood_py.cc:230
#9  0x00007f4d677dd04b in invoke<boost::python::to_python_value<bool const&>, bool (*)(rpwa::pwaLikelihood<std::complex<double> >&, boost::python::dict, boost::python::list), boost::python::arg_from_python<rpwa::pwaLikelihood<std::complex<double> >&>, boost::python::arg_from_python<boost::python::dict>, boost::python::arg_from_python<boost::python::list> > (ac2=<synthetic pointer>, ac1=<synthetic pointer>, ac0=<synthetic pointer>, f=
0x1598788: 0x7f4d677dbd70 <(anonymous namespace)::pwaLikelihood_setOnTheFlyBinning(rpwa::pwaLikelihood<std::complex<double> >&, boost::python::dict, boost::python::list)>, rc=...) at /nfs/freenas/tuph/e18/project/compass/analysis/suhl/software/boost/boost_1_59_0/include/boost/python/detail/invoke.hpp:75

[suhlatwork]

Concerning the crash in the likelihood class, the garbage collector of my python version seems to be much more aggressive and deletes the TFile from which the event metadata were read, so the event tree is an invalid pointer. A similar thing happens for the amplitudes.