-
Notifications
You must be signed in to change notification settings - Fork 34
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
VariableMultipole
compatibility between matlab AT and pyat
#839
base: master
Are you sure you want to change the base?
Conversation
Not totally related, but I think it is also an small bug in In Matlab, I get this weird behavior: KIDDK = atvariablemultipole('KIDDK','arbitrary','VariableThinMPolePass',... gives: struct with fields:
If MaxOrder is fixed to a value then it is ok. |
Dear @ZeusMarti , |
Dear @swhite2401 , |
VariableMultipole
compatibility between matlab AT and pyat
@oscarxblanco , I can do it. In fact there are some changes that you did that I don't understand. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hi @oscarxblanco , here is a first round of comments
pyat/at/lattice/variable_elements.py
Outdated
Periodic=bool, | ||
) | ||
|
||
def __init__(self, family_name: str, **kwargs: dict[str, any]): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You don't need to specify the type of kwargs this is standard python
Could you please restore the defaults for AmplitudeA/B and the mode? I think it is better if they appear in the API doc, in fact I would be tempted to add more of these keyword arguments rather than removing them
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Dear @swhite2401 , if I remove the annotation I get the following flake8 warning. This is the reason why I included, should I ignore it ? I may be using flake8 with a wrong configuration or I may use another tool ?
/variable_elements.py:42:44: ANN003 Missing type annotation for **kwargs
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could you please restore the defaults for AmplitudeA/B and the mode? I think it is better if they appear in the API doc, in fact I would be tempted to add more of these keyword arguments rather than removing them
Dear @swhite2401 , the reason why I remove it is because of an incompatibility between AT matlab and this implementation of the element in pyat.
If I save a lattice in matlab and try to read it in pyat I get the error here below, I'll try to explain the reason in detail, but, in few words: one cannot have args and kwargs with the same name
When reading an element from matlab it is initially put in a dictionary elem_dict
, and then the _BUILD_ATTRIBUTES
are stripped to pass them to the class instantiation as elem_args
. See the lines below
Lines 368 to 369 in 268994f
elem_args = [elem_dict.pop(attr, None) for attr in cls._BUILD_ATTRIBUTES] | |
element = cls(*(arg for arg in elem_args if arg is not None), **elem_dict) |
I could not put the AmplutudeA, AmplitudeB and Mode in the element BUILD_ATTRIBUTES because they are not garanteed to exist in matlab, see the following line
function elem=atvariablemultipole(fname,varargin) |
therefore, the call to cls with args in necessarily incomplete and the AmplitudeA, AmplitudeB and Mode are passed in a dictionary together with args that have the same name. Python does not handle this mix well, it removes the elements from the kwargs and leaves unset the args.
There are ways to workaround this situation when args and kwargs have the same name, but they require changing the way the class is instantiated in utils.py. I decided not to do it because that will change how all elements are instantiated.
A second possibility was to rename the element args in pyat, for example amplitudeA instead of AmplitudeA i.e. lower case in the first letter to make one small change. However, this would break compatibility.
I decided to keep the element without any requirement as in AT matlab, which forced me to do checks on the input parameters before assigning them inside __init__
. This wouldn't break compatibility with previous versions of pyat, and AT matlab files could be read because the element also has no requirements except for the FamName.
If you prefer, I could do as you asked. Keep the args and break backwards compatibility. Or force the user to provide AmplitudeA, AmplitudeB and Mode in matlab, which will make possible to include them in BUILD_ATTRIBUTES that will be removed in the input dictionary and set before the class is instantiated.
Or maybe there is another solution ?
>>> ring = at.load_mat('test1.mat')
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/load/matfile.py", line 202, in load_mat
return Lattice(
^^^^^^^^
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/lattice/lattice_object.py", line 191, in __init__
super(Lattice, self).__init__(elems)
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/lattice/lattice_object.py", line 1478, in params_filter
for elem in elem_filter(params, *args):
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/load/matfile.py", line 140, in ringparam_filter
for elem in elem_iterator(params, *args):
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/load/matfile.py", line 105, in _matfile_generator
yield Element.from_dict(kwargs, index=index, check=check, quiet=quiet)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/load/utils.py", line 365, in element_from_dict
element = cls(*(arg for arg in elem_args if arg is not None), **elem_dict)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/at/lattice/variable_elements.py", line 111, in __init__
raise AtError("Please provide at least one amplitude for A or B")
at.lattice.utils.AtError: Please provide at least one amplitude for A or B
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A second possibility was to rename the element args in pyat, for example amplitudeA instead of AmplitudeA i.e. lower case in the first letter to make one small change. However, this would break compatibility.
Hi @oscarxblanco , thanks for the detailed answers! My personnal opinion is that this was poorly implemented to start with (which is my fault in fact, sorry...) and that in such situation we should be allowed to break backward compatibility.
In all other pyAT element, args seem to have in fact different name that kwargs and we should probably stick to that convention to keep things simple. Similarly we could modify MATLAB to preserve compatibility with the new pyAT version.
This feature is relatively recent and it could well be that @ZeusMarti is the only user... we have used it at ESRF but I don't think we have saved lattice files including it. Would it be a big effort on your side to switch to a new signature?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Dear @swhite2401 , I asked @ZeusMarti and he was not able to use it effectively because he intends to have a custom function where the time of the particle is also considered, i.e. not only a kick every turn for all particles.
I have started modifying the C code today to make some test, is this feature something you would accept for this element ? Or should I keep the custom function option as it is ?
o
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have started modifying the C code today to make some test, is this feature something you would accept for this element ?
Sure! In fact I developed this for @ZeusMarti initially, so any improvement is perfectly fine.
You can already do a kick every n turns by using the table by the way, just fill zeros where you do not want any kick.
Do you want help for this?
pyat/at/lattice/variable_elements.py
Outdated
super(VariableMultipole, self).__init__(family_name, **kwargs) | ||
|
||
def _setmaxorder(self, ampa, ampb): | ||
if len(kwargs) > 0: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why is this needed?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This was needed in order to separate the case when the element is called inside pyat, therefore the object already exists and there are no kwargs, from the case when the element is instantiated using a dictionary (the dictionary comming from reading a .mat, or from user settings).
Sometimes the object is read, for example when printing in a terminal, the element is passed again (I guess this happens because it inherits __str__
from other element) and no modifications in the object are needed, I just need to pass it to super and find its way to __str__
.
pyat/at/lattice/variable_elements.py
Outdated
|
||
def _setmaxorder(self, ampa, ampb): | ||
if len(kwargs) > 0: | ||
self.FamName = family_name |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
family_name = kwargs.pop('FamName', family_name)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I explained before, but, when the element comes from AT matlab, a dictionary is created but all BUILD_ATTRIBUTES are stripped. 'FamName' is a a BUILD_ATTRIBUTE of the class element and therefore it is not in kwargs. One has to take it from args.
pyat/at/lattice/variable_elements.py
Outdated
raise AtError("Please provide at least one amplitude for A or B") | ||
# start setting up Amplitudes and modes | ||
# fist modes are called differently | ||
modepyatinput = kwargs.pop("mode", ACMode.SINE) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You could replace all these lines by:
Self.Mode = kwargs.pop('Mode', int(mode))
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I cannot.
The issue here is that the argument is called 'mode' in pyat and 'Mode' in MATLAB.
I need to check in advance which is correct.
pyat/at/lattice/variable_elements.py
Outdated
def _setmaxorder(self, ampa, ampb): | ||
if len(kwargs) > 0: | ||
self.FamName = family_name | ||
if "AmplitudeA" not in kwargs and "AmplitudeB" not in kwargs: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
AmplitudeA/B présence in kwargs also needs to be checked
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sorry @swhite2401 , I don't understand. What you say is implemented in line 103 above.
You mean individually ? I thought this element should accept to have at least one of the two.
pyat/at/lattice/variable_elements.py
Outdated
self.MaxOrder = kwargs.get("MaxOrder", 0) | ||
amplitudea = None | ||
amplitudeb = None | ||
if "AmplitudeA" in kwargs: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This lol ne and the ones below can be removed
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What do you mean by lol ?
pyat/at/lattice/variable_elements.py
Outdated
# fill in super class | ||
super().__init__(family_name, **kwargs) | ||
|
||
def _setmaxorder(self, ampa: np.ndarray, ampb: np.ndarray): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In what cases did this end up to be none? When ampa/b was filled with only zeros?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have to repeat the test but suppose you have the current implementation available in the master. You will have this behaviour with zeros.
>>> numpy.max(numpy.nonzero([0,0,0,0]))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/numpy/core/fromnumeric.py", line 2810, in max
return _wrapreduction(a, np.maximum, 'max', axis, None, out,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/oblanco/Documents/public/progs/pyenv/pyenvat/lib/python3.11/site-packages/numpy/core/fromnumeric.py", line 88, in _wrapreduction
return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: zero-size array to reduction operation maximum which has no identity
I changed to mxa = np.max(np.append(np.nonzero(ampa), 0))
which is the maximum between the nonzero in ampa and zero, and I get sure ampa exists before doing it.
pyat/at/lattice/variable_elements.py
Outdated
|
||
def _set_params(self, amplitude, mode, ab, **kwargs): | ||
def _set_params( | ||
self, amplitude: int or str, mode, a_b: str, **kwargs: dict[str, any] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No need to define the type of kwargs
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is flake8 who suggested. Should I ignore it ? Do you work with some custom flake8 settings for pyat development ?
@oscarxblanco , finally I found some time to test this. All seems to be working fine, here a test script and output. `import at filename = '../lattice/hmba.mat' variable_mpole = at.VariableThinMultipole('VMULT', at.ACMode.SINE, p0 = np.zeros(6) plt.subplot(211) nbunch = 992 plt.xlabel("x'") |
I have a few comments on the implementation: |
@swhite2401 ,
@swhite2401 , @lfarv . I also ignore translations (T1,T2) and rotations(R1,R2) are ignored. They might be implemented in a future PR if they are necessary. I did not try to implement the buffer as @lfarv considered that this will slow-down the pass method. Should I revert this and do the buffer ? Also, the element name was changed to |
I mean the number of coefficient, presently you stop at 4, but it could go to |
I agree these are not need for the moment |
I don't think it slows down the passmethod, it is just a bit more complicated as you have to handle memory allocation in C, however for me this is a non-issue since it is already implemented elsewhere |
This pull request address issue #838 .
The pyat element class
VariableMultipole
is modified.The attribute
to_dict
is checked in order to determine if the element is already created, if it has not been created the attributes are assigned as before.This also fixes a bug in the determination of
MaxOrder
. Before it could be None, now it is limited to be at least zero.