Merge FreeMagnetismInterface feature enhancement #232
Conversation
Also fixed issue with magnetic profile calculation when drhoM[-1] = 0.
…rface FIX: length of vectors not equal when dthetaM is not defined
Removed conditions at lines 514 and 521. Also added same statement at line 520 as is used for rhoM. This should now behave in the same way as FreeInterface.
Removed divide by zero when z[-1]=0.
…ue when not specified If only tabove or tbelow are specified then they are made equal. If neither are specified they are made the same as DEFAULT_THETA_M. This is to get around the issue when interfacing with a non-magnetic layer where the thetaM value should not change.
Additional edits as suggested by Paul K.
…erface Magnetism interface
Also fixed issue with magnetic profile calculation when drhoM[-1] = 0.
Removed conditions at lines 514 and 521. Also added same statement at line 520 as is used for rhoM. This should now behave in the same way as FreeInterface.
Removed divide by zero when z[-1]=0.
…ue when not specified If only tabove or tbelow are specified then they are made equal. If neither are specified they are made the same as DEFAULT_THETA_M. This is to get around the issue when interfacing with a non-magnetic layer where the thetaM value should not change.
Additional edits as suggested by Paul K.
…metry/refl1d into magnetism-interface-AJC
Also fixed issue with magnetic profile calculation when drhoM[-1] = 0.
Removed conditions at lines 514 and 521. Also added same statement at line 520 as is used for rhoM. This should now behave in the same way as FreeInterface.
Removed divide by zero when z[-1]=0.
…ue when not specified If only tabove or tbelow are specified then they are made equal. If neither are specified they are made the same as DEFAULT_THETA_M. This is to get around the issue when interfacing with a non-magnetic layer where the thetaM value should not change.
Additional edits as suggested by Paul K.
Removed conditions at lines 514 and 521. Also added same statement at line 520 as is used for rhoM. This should now behave in the same way as FreeInterface.
…ue when not specified If only tabove or tbelow are specified then they are made equal. If neither are specified they are made the same as DEFAULT_THETA_M. This is to get around the issue when interfacing with a non-magnetic layer where the thetaM value should not change.
Additional edits as suggested by Paul K.
…gnetism-interface-AJC # Conflicts: # refl1d/mono.py
Magnetism interface webview version
When updating the branch to the latest version from the master branch a lot of merge conflicts had to be gone through. Here I am correcting where I accepted minor changes that do not matter to the code but deviate from the master branch unnecessarily.
refl1d/magnetism.py
Outdated
| """ | ||
|
|
||
| name: Optional[str] | ||
| mbelow: Optional[Union[Parameter, float]] |
There was a problem hiding this comment.
@bmaranville I am not sure if this is the current style that you are using for typing. I.e. would you prefer to remove Any and Optional?
There was a problem hiding this comment.
It's not really a question of style - the dataclass attribute types should be exactly what will be stored in those attributes after init + post_init. There can be different types for the arguments of the init function, if they are going to be coerced before being stored in the attribute. For example, an argument will often be of type Union[float, Parameter] but then we store it as Parameter.default(value) which coerces it to a Parameter, so Parameter should be the type of the attribute (not the Union)
There was a problem hiding this comment.
OK so in this case, we could drop the Optional since we probably don't want None as one of the types? That said tabove and tbelow do take None as the default arg.
There was a problem hiding this comment.
The typing of the arg is not the same as the typing of the attribute - tbelow and tabove are both initialized as Parameter objects during init, even if they are None as an argument. The type of the attributes should be Parameter
|
@acaruana2009 have you tried the new class out? Does it work as you expect? Can you serialize/deserialize it? |
Yes, it all seems to work. I haven't done any fits with it yet, but I can load in a file, save out into a session and reload into a session and it keeps chisq intact. |
…meter after init, even ones that are supplied as float to the init function)
refl1d/mono.py
Outdated
| dp: List[Union[float, Par]] | ||
| dz: List[Par] | ||
| dp: List[Par] | ||
| magnetism: Optional[BaseMagnetism] = None |
There was a problem hiding this comment.
Can this really be any BaseMagnetism object, or does it have to be specifically a FreeMagnetismInterface?
bmaranville
force-pushed
the
magnetism-interface-AJC-webview
branch
from
4f50fbc to
689653a
Compare
… into magnetism-interface-AJC-webview # Conflicts: # refl1d/names.py # refl1d/sample/mono.py
for more information, see https://pre-commit.ci
FreeMagnetismInterfaceacts in the same way thatFreeInterfacedoes. It allows for a freeform connecting spline between magnetic layers to deal with non-trivial magnetic interfaces.Additional modifications have been put into
mono.pyin order to make it more useable, however, more work is probably needed in all of the spline functions we have in refl1d - this may be out of scope of this PR.