def test_plots(mck):
x = np.arange(100)
y = np.arange(100)
phase = Phase(x, y)
phase.plot()
plt.show()
mck.assert_called()
def build_GD(self):
"""
Build the GD and return it. It will rebuild every time
this function is called.
"""
delays, positions = self._collect()
self.GD = Phase(positions, delays, GD_mode=True)
return self.GD
def test_phase_gd_mode():
x = np.arange(100)
y = np.arange(100)
phase = Phase(x, y, GD_mode=True)
d, ds, st = phase.fit(reference_point=50, order=2)
np.testing.assert_array_almost_equal(d, [50, 1])
assert phase.order == 1
assert phase.dispersion_order == 2
def test_errorplot2():
x = np.arange(100)
y = np.arange(100)
phase = Phase(x, y)
with pytest.raises(ValueError):
phase.errors
with pytest.raises(ValueError):
phase.errorplot()
def build_phase(self):
"""
Retrieve *only the phase* after the transforms. This will
unwrap the angles and constructs a `~pysprint.core.phase.Phase` object.
Returns
-------
phase : pysprint.core.phase.Phase
The phase object. See its docstring for more info.
"""
if self.nufft_used:
self.shift("y")
y = np.unwrap(np.angle(self.y), axis=0)
self.phase = Phase(self.x, y)
return self.phase # because of inplace ops. we need to return the phase
def test_errorplot():
x = np.arange(100)
y = np.arange(100)
phase = Phase(x, y)
phase.fit(2, 2)
phase.errorplot()
phase.errorplot(percent=True)
def calculate_from_raw(omegas, delays, reference_point, order, show_graph=False):
"""
Calculate the dispersion from matching pairs of delays and SPP positions.
Parameters
----------
omegas : np.ndarray
The SPP positions.
delays : np.ndarray
The delay values in fs.
reference_point : float
The reference point on the x axis.
order : int
Maximum dispersion order to look for. Must be in [2, 6].
show_graph : bool, optional
Whether to show the fitting.
Returns
-------
dispersion : array-like
The dispersion coefficients in the form of:
[GD, GDD, TOD, FOD, QOD, SOD]
dispersion_std : array-like
Standard deviations due to uncertainty of the fit.
It is only calculated if lmfit is installed. The form is:
[GD_std, GDD_std, TOD_std, FOD_std, QOD_std, SOD_std]
fit_report : str
If lmfit is available returns the fit report, else returns an
empty string.
"""
if order == 1:
raise ValueError(
"Order should be greater than 1. Cannot fit constant function to data."
)
GD = Phase(omegas, delays, GD_mode=True)
d, ds, s = GD._fit(reference_point=reference_point, order=order)
if show_graph:
GD.plot()
return d, ds, s
class SPPMethod(metaclass=_DatasetBase):
"""
Interface for Stationary Phase Point Method.
"""
def __init__(self, ifg_names, sam_names=None, ref_names=None, errors="raise", **kwargs):
"""
SPPMethod constructor.
Parameters
----------
ifg_names : list
The list containing the filenames of the interferograms.
sam_names : list, optional
The list containing the filenames of the sample arm's spectra.
ref_names : list, optinal
The list containing the filenames of the reference arm's spectra.
kwargs :
Additional keyword arguments to pass to `parse_raw` function.
"""
if errors not in ("raise", "ignore"):
raise ValueError("errors must be `raise` or `ignore`.")
self.ifg_names = ifg_names
if sam_names:
self.sam_names = sam_names
else:
self.sam_names = None
if ref_names:
self.ref_names = ref_names
else:
self.ref_names = None
if errors == "raise":
self._validate()
if self.sam_names:
if not len(self.ifg_names) == len(self.sam_names):
raise DatasetError(
"Missmatching length of files. Use None if a file is missing."
)
if self.ref_names:
if not len(self.ifg_names) == len(self.ref_names):
raise DatasetError(
"Missmatching length of files. Use None if a file is missing."
)
self.skiprows = kwargs.pop("skiprows", 0)
self.decimal = kwargs.pop("decimal", ",")
self.sep = kwargs.pop("sep", ";")
self.meta_len = kwargs.pop("meta_len", 1)
self.cb = kwargs.pop("callback", defaultcallback)
self.delimiter = kwargs.pop("delimiter", None)
self.comment = kwargs.pop("comment", None)
self.usecols = kwargs.pop("usecols", None)
self.names = kwargs.pop("names", None)
self.swapaxes = kwargs.pop("swapaxes", False)
self.na_values = kwargs.pop("na_values", None)
self.skip_blank_lines = kwargs.pop("skip_blank_lines", True)
self.keep_default_na = kwargs.pop("keep_default_na", False)
if kwargs:
raise TypeError(f"invalid keyword argument:{kwargs}")
self.load_dict = {
"skiprows": self.skiprows,
"decimal": self.decimal,
"sep": self.sep,
"meta_len": self.meta_len,
"callback": self.cb,
"delimiter": self.delimiter,
"comment": self.comment,
"usecols": self.usecols,
"names": self.names,
"swapaxes": self.swapaxes,
"na_values": self.na_values,
"skip_blank_lines": self.skip_blank_lines,
"keep_default_na": self.keep_default_na
}
self._container = {}
self._info = f"Progress: {len(self._container)}/{len(self)}"
self.GD = None
def _collect(self):
# Maybe the dictionary struct can be dropped at this point..
local_delays = {}
local_positions = {}
for idx, (delay, position) in enumerate(self._container.values()):
if idx != 0 and delay.size > 0 and delay.flat[0] in np.concatenate(
[a.ravel() for a in local_delays.values()]
):
raise ValueError(
f"Duplicated delay values found. Delay {delay.flat[0]} fs was previously seen."
)
local_delays[idx] = delay
local_positions[idx] = position
delays = np.concatenate([a.ravel() for a in local_delays.values()])
positions = np.concatenate([a.ravel() for a in local_positions.values()])
return delays, positions
def append(self, newifg, newsam=None, newref=None):
"""
Append a new interferogram to the object.
"""
# ensure padding before trying to append, and also
# we better prevent infinite loop
# TODO
self.ifg_names.append(newifg)
if newsam is not None:
if self.sam_names is not None:
if len(self.ifg_names) > len(self.sam_names):
while len(self.ifg_names) != len(self.sam_names):
self.sam_names.append(None)
self.sam_names.append(newsam)
if newref is not None:
if self.ref_names is not None:
if len(self.ifg_names) > len(self.ref_names):
while len(self.ifg_names) != len(self.ref_names):
self.ref_names.append(None)
self.ref_names.append(newref)
@staticmethod
def calculate_from_ifg(ifg_list, reference_point, order, show_graph=False):
"""
Collect SPP data from a list of `pysprint.Dataset` or child objects
and evaluate them.
Parameters
----------
ifg_list : list
The list containing the interferograms. All member should be
`pysprint.Dataset` or child class type, otherwise TypeError is raised.
reference_point : float
The reference point on the x axis.
order : int
Maximum dispersion order to look for. Must be in [2, 6].
show_graph : bool, optional
Shows a the final graph of the spectral phase and fitted curve.
Default is False.
Returns
-------
dispersion : array-like
The dispersion coefficients in the form of:
[GD, GDD, TOD, FOD, QOD, SOD]
dispersion_std : array-like
Standard deviations due to uncertainty of the fit.
It is only calculated if lmfit is installed. The form is:
[GD_std, GDD_std, TOD_std, FOD_std, QOD_std, SOD_std]
fit_report : str
If lmfit is available returns the fit report, else returns an
empty string.
"""
for ifg in ifg_list:
if not isinstance(ifg, Dataset):
raise TypeError("pysprint.Dataset objects are expected.")
if order == 1:
raise ValueError(
"Order should be greater than 1. Cannot fit constant function to data."
)
local_delays = {}
local_positions = {}
for idx, ifg in enumerate(ifg_list):
delay, position = ifg.emit()
if idx != 0 and delay.size > 0 and delay.flat[0] in np.concatenate(
[a.ravel() for a in local_delays.values()]
):
raise ValueError(
f"Duplicated delay values found. Delay {delay.flat[0]} fs was previously seen."
)
local_delays[idx] = delay
local_positions[idx] = position
delays = np.concatenate([a.ravel() for a in local_delays.values()])
positions = np.concatenate([a.ravel() for a in local_positions.values()])
GD = Phase(positions, delays, GD_mode=True)
d, ds, s = GD._fit(reference_point=reference_point, order=order)
if show_graph:
GD.plot()
return d, ds, s
def __len__(self):
return len(self.ifg_names)
# TODO
def __str__(self):
return f"{type(self).__name__}\nInterferogram count : {len(self)}"
def _repr_html_(self):
alive = [i for i in Dataset._get_instances() if i.parent == self]
s = f"""
<table style="border:1px solid black;float:top;">
<tbody>
<tr>
<td colspan=2 style="text-align:center">
<font size="5">{type(self).__name__}</font>
</td>
</tr>
<tr>
<td style="text-align:center"><b>Interferograms accumulated<b></td>
<td style="text-align:center"> {len(self)}</td>
</tr>
<tr>
<td style="text-align:center"><b>Interferograms cached<b></td>
<td style="text-align:center"> {len(alive)}</td>
</tr>
<tr>
<td style="text-align:center"><b>Eagerly calculating<b></td>
<td style="text-align:center"> {self.is_eager}</td>
</tr>
<tr>
<td style="text-align:center"><b>Data recorded from<b></td>
<td style="text-align:center"> {len(self._container)}</td>
</tr>
</table>
"""
return s
@lru_cache(500)
def __getitem__(self, key):
try:
dataframe = Dataset.parse_raw(
self.ifg_names[key],
self.sam_names[key],
self.ref_names[key],
**self.load_dict,
parent=self
)
except (TypeError, ValueError):
dataframe = Dataset.parse_raw(
self.ifg_names[key],
**self.load_dict,
parent=self
)
return dataframe
def _validate(self):
for filename in self.ifg_names:
if filename is not None and not os.path.exists(filename):
raise FileNotFoundError(f"""File named '{filename}' is not found.""")
if self.sam_names:
for sam in self.sam_names:
if sam is not None and not os.path.exists(sam):
raise FileNotFoundError(f"""File named '{sam}' is not found.""")
if self.ref_names:
for ref in self.ref_names:
if ref is not None and not os.path.exists(ref):
raise FileNotFoundError(f"""File named '{ref}' is not found.""")
def flush(self):
"""
Reset the state of recorded delays and positions, even on
active objects that have been constructed on the runtime.
"""
self._container = {}
for ifg in self:
ifg._delay = None
ifg.delay = None
ifg._positions = None
ifg.positions = None
def save_data(self, filename):
"""
Save the currectly stored SPP data.
Parameters
----------
filename : str
The filename to save as. If not ends with ".txt" it's
appended by default.
"""
if not filename.endswith(".txt"):
filename += ".txt"
delays, positions = self._collect()
np.savetxt(
f"{filename}", np.column_stack((positions, delays)), delimiter=",",
)
@staticmethod
def calculate_from_raw(omegas, delays, reference_point, order, show_graph=False):
"""
Calculate the dispersion from matching pairs of delays and SPP positions.
Parameters
----------
omegas : np.ndarray
The SPP positions.
delays : np.ndarray
The delay values in fs.
reference_point : float
The reference point on the x axis.
order : int
Maximum dispersion order to look for. Must be in [2, 6].
show_graph : bool, optional
Whether to show the fitting.
Returns
-------
dispersion : array-like
The dispersion coefficients in the form of:
[GD, GDD, TOD, FOD, QOD, SOD]
dispersion_std : array-like
Standard deviations due to uncertainty of the fit.
It is only calculated if lmfit is installed. The form is:
[GD_std, GDD_std, TOD_std, FOD_std, QOD_std, SOD_std]
fit_report : str
If lmfit is available returns the fit report, else returns an
empty string.
"""
if order == 1:
raise ValueError(
"Order should be greater than 1. Cannot fit constant function to data."
)
GD = Phase(omegas, delays, GD_mode=True)
d, ds, s = GD._fit(reference_point=reference_point, order=order)
if show_graph:
GD.plot()
return d, ds, s
def build_GD(self):
"""
Build the GD and return it. It will rebuild every time
this function is called.
"""
delays, positions = self._collect()
self.GD = Phase(positions, delays, GD_mode=True)
return self.GD
def calculate(self, reference_point, order, show_graph=False):
"""
This function should be used after setting the SPP data in
the interactive matplotlib editor or other way.
Parameters
----------
reference_point : float
The reference point on the x axis.
order : int, optional
Maximum dispersion order to look for. Must be in [2, 6].
show_graph : bool, optional
Shows a the final graph of the spectral phase and fitted curve.
Default is False.
Returns
-------
dispersion : array-like
The dispersion coefficients in the form of:
[GD, GDD, TOD, FOD, QOD, SOD]
dispersion_std : array-like
Standard deviations due to uncertainty of the fit.
It is only calculated if lmfit is installed. The form is:
[GD_std, GDD_std, TOD_std, FOD_std, QOD_std, SOD_std]
fit_report : str
If lmfit is available returns the fit report, else returns an
empty string.
"""
if order == 1:
raise ValueError(
"Order should be greater than 1. Cannot fit constant function to data."
)
self.build_GD()
d, ds, s = self.GD._fit(reference_point=reference_point, order=order)
if show_graph:
self.GD.plot()
return d, ds, s
@property
def info(self):
"""
Return how many interferograms were processed.
"""
self._info = f"Progress: {len(self._container)}/{len(self)}"
return self._info
@property
def is_eager(self):
"""
Returns if eager execution is enabled.
"""
# TODO
if self.cb.__name__ == "inner":
return True
return False
def calculate_from_ifg(ifg_list, reference_point, order, show_graph=False):
"""
Collect SPP data from a list of `pysprint.Dataset` or child objects
and evaluate them.
Parameters
----------
ifg_list : list
The list containing the interferograms. All member should be
`pysprint.Dataset` or child class type, otherwise TypeError is raised.
reference_point : float
The reference point on the x axis.
order : int
Maximum dispersion order to look for. Must be in [2, 6].
show_graph : bool, optional
Shows a the final graph of the spectral phase and fitted curve.
Default is False.
Returns
-------
dispersion : array-like
The dispersion coefficients in the form of:
[GD, GDD, TOD, FOD, QOD, SOD]
dispersion_std : array-like
Standard deviations due to uncertainty of the fit.
It is only calculated if lmfit is installed. The form is:
[GD_std, GDD_std, TOD_std, FOD_std, QOD_std, SOD_std]
fit_report : str
If lmfit is available returns the fit report, else returns an
empty string.
"""
for ifg in ifg_list:
if not isinstance(ifg, Dataset):
raise TypeError("pysprint.Dataset objects are expected.")
if order == 1:
raise ValueError(
"Order should be greater than 1. Cannot fit constant function to data."
)
local_delays = {}
local_positions = {}
for idx, ifg in enumerate(ifg_list):
delay, position = ifg.emit()
if idx != 0 and delay.size > 0 and delay.flat[0] in np.concatenate(
[a.ravel() for a in local_delays.values()]
):
raise ValueError(
f"Duplicated delay values found. Delay {delay.flat[0]} fs was previously seen."
)
local_delays[idx] = delay
local_positions[idx] = position
delays = np.concatenate([a.ravel() for a in local_delays.values()])
positions = np.concatenate([a.ravel() for a in local_positions.values()])
GD = Phase(positions, delays, GD_mode=True)
d, ds, s = GD._fit(reference_point=reference_point, order=order)
if show_graph:
GD.plot()
return d, ds, s
class FFTMethod(Dataset):
"""
Basic interface for the Fourier transform method.
"""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
# making sure it's not normalized
if self._is_normalized:
self.y_norm = self.y
self._is_normalized = False
if np.any(self.sam) or np.any(self.ref):
warnings.warn("This method doesn't require arms' spectra.",
PySprintWarning)
self.original_x = self.x
self.at = None
self.std = None
self.fwhm = None
self.window_order = None
self.phase = None
self._ifft_called_first = False
self.nufft_used = False
@inplacify
def shift(self, axis="x"):
"""
Equivalent to `scipy.fftpack.fftshift`, but it's easier to
use this function instead, because we don't need to explicitly
call the class' x and y attribute.
Parameters
----------
axis : str
Must be 'x', 'y', 'both', 'xy' or 'yx'.
"""
if axis == "x":
self.x = fftshift(self.x)
elif axis == "y":
self.y = fftshift(self.y)
elif axis == "both" or axis == "xy" or axis == "yx":
self.y = fftshift(self.y)
self.x = fftshift(self.x)
else:
raise ValueError("axis should be either `x`, `y` or `both`.")
return self
@inplacify
def ifft(
self,
interpolate=True,
usenifft=False,
eps=1e-12,
exponent="positive",
):
"""
Applies inverse Fast Fourier Transfrom to the dataset.
Parameters
----------
interpolate : bool, default is True -- WILL BE REMOVED
Whether to apply linear interpolation on the dataset
before transforming.
usenifft : bool, optional
Whether to use non uniform fft. It uses the algorithm
described in the references. This means the interferogram
will *not* be linearly interpolated. Default is False.
eps : float, optional
The desired approximate error for the non uniform FFT result. Must be
in range 1E-33 < eps < 1E-1, though be aware that the errors are
only well calibrated near the range 1E-12 ~ 1E-6. Default is 1E-12.
exponent : str, optional
if 'negative', compute the transform with a negative exponent.
if 'positive', compute the transform with a positive exponent.
Default is `positive`.
Notes
-----
The basic scheme is ifft -> windowing -> fft, so you should call
these functions in this order. Otherwise the transforms may be
inconsistent.
If numba is not installed the non uniform FTT is approximately
5x times slower, but still remains comparable to `np.fft.ifft`.
References
----------
[1] Dutt A., Rokhlin V. : Fast Fourier Transforms for Nonequispaced Data II,
Applied and Computational Harmonic Analysis
Volume 2, Issue 1, January 1995, Pages 85-100
(1995)
[2] Greengard, Leslie & Lee, June-Yub.: Accelerating the
Nonuniform Fast Fourier Transform,
Society for Industrial and Applied Mathematics.
46. 443-454. 10.1137/S003614450343200X.
(2004)
"""
self.nufft_used = usenifft
self._ifft_called_first = True
if usenifft:
x_spaced = np.linspace(self.x[0], self.x[-1], len(self.x))
timestep = np.diff(x_spaced)[0]
x_axis = np.fft.fftfreq(len(self.x), d=timestep / (2 * np.pi))
y_transform = nuifft(
self.x,
self.y,
gl=len(self.x),
df=(x_axis[1] - x_axis[0]),
epsilon=eps,
exponent=exponent,
)
self.x, self.y = x_axis, np.fft.fftshift(y_transform)
else:
self.x, self.y = ifft_method(self.x,
self.y,
interpolate=interpolate)
return self
@inplacify
def fft(self):
"""
Applies fft to the dataset.
If ifft was not called first, inaccurate results might happen.
"""
if not self._ifft_called_first:
warnings.warn("This module is designed to call ifft before fft",
FourierWarning)
self.x, self.y = fft_method(self.original_x, self.y)
return self
@inplacify
def window(self, at, fwhm, window_order=6, plot=True):
"""
Draws a gaussian window on the plot with the desired parameters.
The maximum value is adjusted for the dataset's maximum value,
mostly for visibility.
Parameters
----------
at : float
The maximum of the gaussian curve.
fwhm : float
Full width at half maximum of the gaussian
window_order : int, optional
Order of the gaussian curve.
If not even, it's incremented by 1.
Default is 6.
plot : bool, optional
Whether to immediately show the window with the data.
Default is `True`.
"""
self.at = at
self.fwhm = fwhm
self.window_order = window_order
gaussian = gaussian_window(self.x, self.at, self.fwhm,
self.window_order)
self.plt.plot(self.x, gaussian * max(abs(self.y)), "k--")
if plot:
self.plot(overwrite="$t\,[fs]$")
self.show()
return self
@inplacify
def apply_window(self):
"""
If window function is set, applies window on the dataset.
"""
self.plt.clf()
self.plt.cla()
self.plt.close()
self.y = cut_gaussian(
self.x,
self.y,
spike=self.at,
fwhm=self.fwhm,
win_order=self.window_order,
)
return self
def build_phase(self):
"""
Retrieve *only the phase* after the transforms. This will
unwrap the angles and constructs a `~pysprint.core.phase.Phase` object.
Returns
-------
phase : pysprint.core.phase.Phase
The phase object. See its docstring for more info.
"""
if self.nufft_used:
self.shift("y")
y = np.unwrap(np.angle(self.y), axis=0)
self.phase = Phase(self.x, y)
return self.phase # because of inplace ops. we need to return the phase
def calculate(self, reference_point, order, show_graph=False):
"""
FFTMethod's calculate function. It will unwrap the phase by changing
deltas _between values to 2*pi complement. After that, fit a curve to
determine dispersion coefficients.
Parameters
----------
reference_point : float
The reference point on the x axis.
order : int
Polynomial (and maximum dispersion) order to fit. Must be in [1, 5].
show_graph : bool, optional
Shows a the final graph of the spectral phase and fitted curve.
Default is False.
Returns
-------
dispersion : array-like
The dispersion coefficients in the form of:
[GD, GDD, TOD, FOD, QOD, SOD]
dispersion_std : array-like
Standard deviations due to uncertainty of the fit.
It is only calculated if lmfit is installed. The form is:
[GD_std, GDD_std, TOD_std, FOD_std, QOD_std, SOD_std]
fit_report : str
If lmfit is available returns the fit report, else returns an
empty string.
Notes
-----
Decorated with pprint_disp, so the results are immediately
printed without explicitly saying so.
Developer-commentary
--------------------
Currently the x-axis transformation is sloppy, because we cache the
original x axis and not transforming it backwards.
In addition we need to keep track of interpolation and
zero-padding too. Currently the transforms are correct only if
first ifft was used. For now it's doing okay: giving good results.
For consistency we should still implement that a better way later.
"""
self.build_phase()
dispersion, dispersion_std, fit_report = self.phase._fit(
reference_point=reference_point, order=order)
if show_graph:
self.phase.plot()
self._dispersion_array = dispersion
return -dispersion, dispersion_std, fit_report
def autorun(
self,
reference_point=None,
order=None,
*,
enable_printing=True,
skip_domain_check=False,
only_phase=False,
show_graph=True,
usenifft=False,
):
"""
Automatically run the Fourier Transfrom based evaluation on the dataset.
It's not as reliable as I want it to be, so use it carefully. I'm working
on making it as competent and useful as possible.
Parameters
----------
reference_point : float, optional
The reference point on the x axis. If not given, only_phase mode
will be activated. Default is None.
order : int, optional
Polynomial (and maximum dispersion) order to fit. Must be in [1, 6].
If not given, only_phase mode will be activated. Default is None.
only_phase : bool, optional
If True, activate the only_phase mode, which will retrieve the phase
without fitting a curve, and return a `pysprint.core.Phase.phase` object.
Default is False (also not giving enough information for curve fitting
will automatically activate it).
enable_printing : bool, optional
If True enable printing the detailed results. Default is True.
skip_domain_check : bool, optional
If True skip the interferogram domain check and force the algorithm
to perform actions without changing domain. If False, check for potential
wrong domains and change for an appropriate one. Default is False.
show_graph : bool, optional
If True show the graph with the phase and the fitted curve, if there is any.
Default is True.
usenifft : bool, optional
If True use the Non Uniform Fast Fourier Transform algorithm. For more details
see `help(pysprint.FFTMethod.ifft)`. Default is False.
References
----------
[1] Dutt A., Rokhlin V. : Fast Fourier Transforms for Nonequispaced Data II,
Applied and Computational Harmonic Analysis
Volume 2, Issue 1, January 1995, Pages 85-100
(1995)
[2] Greengard, Leslie & Lee, June-Yub.: Accelerating the
Nonuniform Fast Fourier Transform,
Society for Industrial and Applied Mathematics.
46. 443-454. 10.1137/S003614450343200X.
(2004)
"""
if not reference_point or not order:
only_phase = True
if not enable_printing:
with open(os.devnull, "w") as g, contextlib.redirect_stdout(g):
_run(
self,
skip_domain_check=skip_domain_check,
show_graph=show_graph,
usenifft=usenifft,
)
if only_phase:
y = np.unwrap(np.angle(self.y), axis=0)
self.phase = Phase(self.x, y)
return self.phase
self.calculate(reference_point=reference_point,
order=order,
show_graph=True)
else:
_run(
self,
skip_domain_check=skip_domain_check,
show_graph=show_graph,
usenifft=usenifft,
)
if only_phase:
y = np.unwrap(np.angle(self.y), axis=0)
self.phase = Phase(self.x, y)
return self.phase
self.calculate(reference_point=reference_point,
order=order,
show_graph=True)
# TODO: add interpolation
def get_pulse_shape_from_array(self,
x_sample,
y_sample,
truncate=True,
tol=None):
"""
Find out the shape of the pulse in the time domain I(t).
Parameters
----------
x_sample : np.ndarray
The x values of the sample arm.
y_sample : np.ndarray
The y values of the sample arm.
truncate : bool, optional
Whether to truncate the phase and sample spectra
to the longest_common_subsequence (imeplemented at
pysprint.core.bases.algorithms). Default is True.
tol : float or None, optional
The tolerance which determines how big difference is allowed
_between x values to interpret them as the same datapoint.
"""
if self.phase is None:
raise NotCalculatedException("Must calculate phase first.")
if not len(y_sample) == len(x_sample):
raise ValueError("Missmatching shapes.")
# quick check if we're able to broadcast
y_sample = np.asarray(y_sample, dtype=float)
x_phase, y_phase = self.phase.data[0], self.phase.data[1]
if len(y_sample) != len(self.phase.data[0]):
if truncate:
x_sample, y_sample, x_phase, y_phase = longest_common_subsequence(
x_sample, y_sample, x_phase, y_phase, tol=tol)
logger.info(
f"Shapes were truncated from {np.min(x_sample)} to {np.max(x_sample)} with length {len(x_sample)}."
)
else:
raise ValueError(
f"Shapes differ with {len(x_sample)} and {len(self.phase.data[0])}."
)
E_field = np.sqrt(y_sample) * np.exp(-1j * y_phase)
E_pulse = np.abs(np.fft.ifft(E_field))**2
x_spaced = np.linspace(x_phase[0], x_phase[-1], len(x_phase))
timestep = np.diff(x_spaced)[0]
x_axis = np.fft.fftfreq(len(x_phase), d=timestep / (2 * np.pi))
return x_axis, E_pulse
def get_pulse_shape_from_file(self,
filename,
truncate=True,
tol=None,
**kwargs):
"""
Find out the shape of the pulse in the time domain I(t).
The sample arm's spectra is loaded from file.
Parameters
----------
filename : str
The file containing the sample arm's spectra.
truncate : bool, optional
Whether to truncate the phase and sample spectra
to the longest_common_subsequence (imeplemented at
pysprint.core.bases.algorithms). Default is True.
tol : float or None, optional
The tolerance which determines how big difference is allowed
_between x values to interpret them as the same datapoint.
kwargs : dict, optional
The additional keyword arguments for parsing. Same as
`pysprint.Dataset.parse_raw`. If `chdomain=True`, then
change the domain after loading.
"""
if isinstance(filename, str):
ch = kwargs.pop("chdomain", False)
df = pd.read_csv(filename, names=["x", "y"], **kwargs)
x_sample = df["x"].values
y_sample = df["y"].values
if ch:
x_sample = self.wave2freq(x_sample)
return self.get_pulse_shape_from_array(x_sample,
y_sample,
truncate=truncate,
tol=tol)
def errorplot(self, *args, **kwargs):
"""
Plot the errors of fitting.
Parameters
----------
ax : matplotlib.axes.Axes, optional
An axis to draw the plot on. If not given, it will plot
of the last used axis.
percent : bool, optional
Whether to plot percentage difference. Default is False.
title : str, optional
The title of the plot. Default is "Errors".
kwargs : dict, optional
Additional keyword arguments to pass to plot function.
"""
try:
getattr(self.phase, "errorplot", None)(*args, **kwargs)
except TypeError:
raise NotCalculatedException(
"Must calculate before plotting errors.")
@property
def get_phase(self):
"""
Return the phase if it is already calculated.
"""
if self.phase is not None:
return self.phase
raise NotCalculatedException("Must retrieve the phase first.")
@property
def errors(self):
"""
Return the fitting errors as np.ndarray.
"""
errors = getattr(self.phase, "errors", None)
if errors is not None:
return errors
raise NotCalculatedException("Must calculate the fit first.")
# redefinition to ensure proper attributes are changed
@inplacify
def resample(self, N, kind="linear", **kwds):
"""
Resample the interferogram to have `N` datapoints.
Parameters
----------
N : int
The number of datapoints required.
kind : str, optional
The type of interpolation to use. Default is `linear`.
kwds : optional
Additional keyword argument to pass to `scipy.interpolate.interp1d`.
Raises
------
PySprintWarning, if trying to subsample to lower `N` datapoints than original.
"""
f = interp1d(self.x, self.y_norm, kind, **kwds)
if N < len(self.x):
N = len(self.x)
warnings.warn(
"Trying to resample to lower resolution, keeping shape..",
PySprintWarning)
xnew = np.linspace(np.min(self.x), np.max(self.x), N)
ynew = f(xnew)
setattr(self, "x", xnew)
setattr(self, "y_norm", ynew)
setattr(self, "y", ynew)
return self
def autorun(
self,
reference_point=None,
order=None,
*,
enable_printing=True,
skip_domain_check=False,
only_phase=False,
show_graph=True,
usenifft=False,
):
"""
Automatically run the Fourier Transfrom based evaluation on the dataset.
It's not as reliable as I want it to be, so use it carefully. I'm working
on making it as competent and useful as possible.
Parameters
----------
reference_point : float, optional
The reference point on the x axis. If not given, only_phase mode
will be activated. Default is None.
order : int, optional
Polynomial (and maximum dispersion) order to fit. Must be in [1, 6].
If not given, only_phase mode will be activated. Default is None.
only_phase : bool, optional
If True, activate the only_phase mode, which will retrieve the phase
without fitting a curve, and return a `pysprint.core.Phase.phase` object.
Default is False (also not giving enough information for curve fitting
will automatically activate it).
enable_printing : bool, optional
If True enable printing the detailed results. Default is True.
skip_domain_check : bool, optional
If True skip the interferogram domain check and force the algorithm
to perform actions without changing domain. If False, check for potential
wrong domains and change for an appropriate one. Default is False.
show_graph : bool, optional
If True show the graph with the phase and the fitted curve, if there is any.
Default is True.
usenifft : bool, optional
If True use the Non Uniform Fast Fourier Transform algorithm. For more details
see `help(pysprint.FFTMethod.ifft)`. Default is False.
References
----------
[1] Dutt A., Rokhlin V. : Fast Fourier Transforms for Nonequispaced Data II,
Applied and Computational Harmonic Analysis
Volume 2, Issue 1, January 1995, Pages 85-100
(1995)
[2] Greengard, Leslie & Lee, June-Yub.: Accelerating the
Nonuniform Fast Fourier Transform,
Society for Industrial and Applied Mathematics.
46. 443-454. 10.1137/S003614450343200X.
(2004)
"""
if not reference_point or not order:
only_phase = True
if not enable_printing:
with open(os.devnull, "w") as g, contextlib.redirect_stdout(g):
_run(
self,
skip_domain_check=skip_domain_check,
show_graph=show_graph,
usenifft=usenifft,
)
if only_phase:
y = np.unwrap(np.angle(self.y), axis=0)
self.phase = Phase(self.x, y)
return self.phase
self.calculate(reference_point=reference_point,
order=order,
show_graph=True)
else:
_run(
self,
skip_domain_check=skip_domain_check,
show_graph=show_graph,
usenifft=usenifft,
)
if only_phase:
y = np.unwrap(np.angle(self.y), axis=0)
self.phase = Phase(self.x, y)
return self.phase
self.calculate(reference_point=reference_point,
order=order,
show_graph=True)
def constructor2():
Phase.from_disperion_array([1, 2, 3, 4])
Phase.from_disperion_array([1, 2, 3, 4], domain=np.arange(5611))
def constructor1():
Phase.from_coeff([1, 2, 3, 4])
Phase.from_coeff([1, 2, 3, 4], domain=np.arange(5611))
def build_GD(self,
silent=False,
fastmath=True,
usenifft=False,
parallel=False,
errors="ignore"):
"""
Build the GD.
Parameters
----------
silent : bool, optional
Whether to print progressbar. By default it will print.
fastmath : bool, optional
Whether to build additional arrays to display heatmap.
Default is True.
usenifft : bool, optional
Whether to use Non-unfirom FFT when calculating GD.
Default is False. **Not stable.**
parallel : bool, optional
Whether to use parallel computation. Only availabe if `Dask`
is installed. The speedup is about 50-70%. Default is False.
errors : str, optional
Whether to raise an error is the algorithm couldn't find the
center of the peak.
Returns
-------
GD : pysprint.core.phase.Phase
The phase object with `GD_mode=True`. See its docstring for more info.
"""
if parallel:
if not CAN_PARALLELIZE:
raise ModuleNotFoundError(
"Module `dask` not found. Please install it in order to use parallelism."
)
else:
self.fastmath = fastmath
self._apply_window_seq_parallel(fastmath=fastmath,
usenifft=usenifft,
errors=errors)
if not silent:
with ProgressBar():
computed = compute(*self.found_centers.values())
else:
computed = compute(*self.found_centers.values())
cleaned_delays = [
k for i, k in enumerate(self.found_centers.keys())
if computed[i] is not None
]
delay = np.fromiter(cleaned_delays, dtype=float)
omega = np.fromiter([c for c in computed if c is not None],
dtype=float)
if not silent:
print(
f"Errors found: {len(self.window_seq) - sum(1 for _ in filter(None.__ne__, computed))}"
)
else:
self.fastmath = fastmath
self._apply_window_sequence(silent=silent,
fastmath=fastmath,
usenifft=usenifft)
self._clean_centers()
delay = np.fromiter(self.found_centers.keys(), dtype=float)
omega = np.fromiter(self.found_centers.values(), dtype=float)
self.GD = Phase(delay, omega, GD_mode=True)
return self.GD