Structure of future attributes #419
Comments
|
I think the design constraints are something like this:
It's pretty clear it'll be hard to make this change backwards compatible, so let's plan this for a 2.x branch. It might make sense to think about the recently added |
|
Some discussion from #227:
I think I'm leaning towards 1 as well since it's the most flexible, but I wanted to consider the alternatives. @ATenderholt any other ideas or suggestions at this point? What do you think about this endeavor in general? |
|
I have started working on this, for something to discuss: https://github.com/berquist/cclib/compare/cjson...berquist:qcjson?expand=1 Each attribute should eventually have its own Currently, # Static polarizability from **PROPERTIES/.POLARI.
if line.strip() == "Static polarizabilities (au)":
if not hasattr(self, 'polarizabilities'):
self.polarizabilities = []
polarizability = []
self.skip_lines(inputfile, ['d', 'b', 'directions', 'b'])
for _ in range(3):
line = next(inputfile)
# Separate possibly unspaced huge negative polarizability tensor
# element and the left adjacent column from each other.
line = line.replace('-', ' -')
polarizability.append(line.split()[1:])
self.polarizabilities.append(numpy.array(polarizability))would become # Static polarizability from **PROPERTIES/.POLARI.
if line.strip() == "Static polarizabilities (au)":
if not hasattr(self, 'polarizabilities'):
self.polarizabilities = attributes.polarizabilities()
self.polarizabilities([])
polarizability = []
self.skip_lines(inputfile, ['d', 'b', 'directions', 'b'])
for _ in range(3):
line = next(inputfile)
# Separate possibly unspaced huge negative polarizability tensor
# element and the left adjacent column from each other.
line = line.replace('-', ' -')
polarizability.append(line.split()[1:])
polarizabilities = self.polarizabilities() + [numpy.array(polarizability)]
self.polarizabilities(polarizabilities)This is not good, but I'm still learning about private class methods. I also still need to handle type conversion for "compound" types (like lists of arrays). |
|
Desired behavior notes:
if not hasattr(self, 'polarizabilities'):
self.polarizabilities = []
polarizability = []
for _ in range(3):
line = next(inputfile)
polarizability.append(line.split()[1:])
self.polarizabilities.append(polarizability)where
assert isinstance(attribute_classes, dict)
assert isinstance(attribute_instances, dict)
for attribute_name, attribute_instance in attribute_instances:
assert isinstance(attribute_name, str)
# The keys of attribute_instances should be a subset of the keys of attribute_classes
assert attribute_name in attribute_classes
if not isinstance(attribute_instance, attribute_classes[attribute_name])):
attribute_instances[attribute_name] = attribute_classes[attribute_name](attribute_instance)
else:
# The constructor already does type checking/conversion if passed an argument.
attribute_instance.typecheck()
|
|
I have a general implementation of attribute classes, but still need to modify all the parsers. Here are my current thoughts. If each attribute is a If we want to do this dynamically, as we do now, there is some trickery that needs to be done according to https://stackoverflow.com/a/1355444/, but I haven't tested it. |
|
Can we have a |
|
Poke for #419 (comment). |
|
This is for discussion of how attributes will be represented in the future. Some relevant starting points are #398 (comment) and #227.
The text was updated successfully, but these errors were encountered: