Organization of parameters

[ALF-Import-Bot]

In GitLab by @jonasschwab on Jul 24, 2020, 15:17 UTC:

I'd like to document the discussion on how to organize the parameters here.

The current structure is as follows: There are a number of "default" namelists:

• VAR_Lattice
• VAR_Model_Generic
• VAR_QMC
• VAR_QMC
• VAR_TEMP
• VAR_Max_Stoch

where VAR_Lattice and VAR_Model_Generic are supposed to be used in every hamiltonian. On top of that are a number of model specific namelists e.g. VAR_Hubbard.

Migrated from GitLab: https://git.physik.uni-wuerzburg.de/ALF/pyALF/-/issues/7

[ALF-Import-Bot]

In GitLab by @jonasschwab on Jul 24, 2020, 15:22 UTC:

The problem I see with this design is, that the structure sometimes implies parameters which are not present in a Hamiltionan. E.g. Hamiltonian_Hubbard_mod.F90 does read the parameter model from VAR_Lattice but doesn't use it.

I suggest a unique namelist for each hamiltonian e.g. VAR_Hubbard that contains all variables read in Hamiltonian_Hubbard_mod.F90.

[ALF-Import-Bot]

In GitLab by @jonasschwab on Jul 24, 2020, 15:28 UTC:

I've added the parameter ham_name to the initialisation of py_alf.Simulation, which replaces executable and partly the model name.

[ALF-Import-Bot]

In GitLab by @jonasschwab on Jul 24, 2020, 18:38 UTC:

Which namelists to write out gets now decided on basis of ham_name.

[ALF-Import-Bot]

In GitLab by @fassaad on Aug 2, 2020, 16:23 UTC:

Lower and upper case for the directory name.

At present if we use
{"Model": "Hubbard", "Lattice_type": "Honeycomb", "L1": 3 , "L2": 3, "ham_T2": 0.0, "NBin": 5}

then the directory name will include T2 since in default_variables.py the variable is declared as "ham_T2" : [1.0, ""]. However if one gets it wrong and types:

{"Model": "Hubbard", "Lattice_type": "Honeycomb", "L1": 3 , "L2": 3, "Ham_T2": 0.0, "NBin": 5}

then T2 will not be included in the directory name. Would be nice to make the generation of then directory name case insensitive by for instance using .upper.

[ALF-Import-Bot]

In GitLab by @jonasschwab on Aug 2, 2020, 21:01 UTC:

This should work now.