We rely on OrderedDict to keep the order of the parameters. However given the way the code is structured currently, there is really no reason for this. So let's just switch to the standard dictionary. This has two advantages. First the standard dictionary is much faster than OrderedDict (implemented in C rather than in pure python). Then, standard dictionaries are marshalable, which makes them usable as keys for other dictionaries.

Repair the badly broken dustatt_powerlaw module. It did not handle the contribution names properly, which broke the code if the nebular module was called.

Homogeneize the attenuation modules with the latest changes regarding the way we handle filters in compute_fnu(). Also make sure we do not request the key from the filter list as it is not a dictionary anymore.

Use an ordered dictionary to ensure that the attenuations in given bands are always computed in the same order.

During a given run the set of wavelengths will always be the same. Then it means we do not have to recompute the shape of the attenuation curve each time. So we compute it only the first time reuse it later.

Allow pcigale to run without stellar populations so we can carry out SED modelling of dust only for instance.

Remove the possibility to use the same module several times. Simplification of the code ensues. Now sed.info keys are prefixed with the characteristic of the module (stellar, nebular, etc.).

Remove the author email addresses from the source files and add a AUTHORS.rst
file with the addresses and current affiliation to make it easier to update
the information.

Also, give copyright to the University of Cambridge and the University of
Crete for code developed by authors working there.

It will make the documentation clearer.

For the creation and analysis modules, move the contents of the
common.py files to the __init__.py.

Move the SED creation modules to pcigale.creation_modules to be more
consistent with the analysis module location.

Instead of prefixing the key in the SED with the module name, we postfix
them with the same postfix as the module name if any. This makes easier
to inter-change modules adding the same information (for instance
attenuation modules adding the attenuation value).
That implies that when using several modules adding the same
information, one must postfix the name of each.

Instead of prefixing the key in the SED with the module name, we postfix
them with the same postfix as the module name if any. This makes easier
to inter-change modules adding the same information (for instance
attenuation modules adding the attenuation value).
That implies that when using several modules adding the same
information, one must postfix the name of each.

When one value for a parameter is a list (for instance the list of
attenuation keys for re-emission modules) we had to use the "eval [["a",
"b"]]" notation. Convert these module to use a single string with items
separated by a & ("a & b").

WARNING we must still change the way parameters are evaluated not to
return everytime a list.

So that the configuration files and the serialized data have a logical
order.

Prepare to make it possible to get the flux in a filter without taking
into account the contribution from the spectral lines.

The process method was used to complete and check the parameter
dictionary before launching the _process one defined in each module. Now
that the parameters are dealed with at instantiation, this separation is
no more useful.

Also, modify the attenuation modules to read the filter in the database
at instantiation.

To avoid problems that might occur when, for instance, using an integer
instead of a float, we now cast the module parameters to their expected
type. This will also have the advantage to raise an error when a
parameter can't be interpreted with its expected type.

Also, use integer for age parameters as we are now using Myr as unit.