Skip to content

Commit

Permalink
346 TimeSeries interp_time (#353)
Browse files Browse the repository at this point in the history 
* Change interface of interp_time
* Emit warning when interpolating interpolated data
* Add interpolation setter
* Update valid interpolations

Co-authored-by: Cagtay Fabry <43667554+CagtayFabry@users.noreply.github.com>
  • Loading branch information
@vhirtham @CagtayFabry
vhirtham and CagtayFabry authored Jun 25, 2021
1 parent ca354e1 commit 1999c8d
Show file tree
Hide file tree
Showing 6 changed files with 210 additions and 124 deletions.
10 changes: 7 additions & 3 deletions CHANGELOG.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,13 +9,12 @@
structures. [[#328]](https://github.com/BAMWelDX/weldx/pull/328)
- added `WeldxFile` wrapper to handle asdf files with history and schemas more
easily. [[#341]](https://github.com/BAMWelDX/weldx/pull/341).
- added `WeldxFile` wrapper to handle asdf files with history and schemas more
easily. [[#341]](https://github.com/BAMWelDX/weldx/pull/341).
- add `"step"` as additional method to `util.xr_interp_like` [[#363]](https://github.com/BAMWelDX/weldx/pull/363)
- add `util.compare_nested_eq` decorator for dataclasses with array-like
fields [[#378]](https://github.com/BAMWelDX/weldx/pull/378)
- adds a `dataclass_serialization_class` utility function that automatically generates the asdf serialization class for
python dataclasses. [[#380]](https://github.com/BAMWelDX/weldx/pull/380)
- Added method to set the interpolation method to the `TimeSeries` [[#353]](https://github.com/BAMWelDX/weldx/pull/353)

### changes

Expand All @@ -24,7 +23,11 @@
- `get_yaml_header` received a new option parse, which optionally returns the parsed YAML header
as `asdf.tagged.TaggedDict`. [[#338]](https://github.com/BAMWelDX/weldx/pull/338)
- refactor `asdf_json_repr` into `view_tree` [[#339]](https://github.com/BAMWelDX/weldx/pull/339)
- The `MeasurementChain` is now internally based on a `networkx.DiGraph`. New functions are also added to the class to
- `TimeSeries.interp_time` [[#353]](https://github.com/BAMWelDX/weldx/pull/353)
- now returns a new `TimeSeries` instead of a `xarray.DataArray`
- if the data has already been interpolated before, a warning is emitted
- `TimeSeries` supports now all interpolation methods supported by xarray
- The `MeasurementChain` is now internally based on a `networkx.DiGraph`. New functions are also added to the class to
simplify its usage. [[#326]](https://github.com/BAMWelDX/weldx/pull/326)
The following additional changes were applied during the update of the `MeasurementChain`:
- renamed `DataTransformation` class to `SignalTransformation`
Expand All @@ -38,6 +41,7 @@
- Add new tutorial about the `MeasurementChain` [[#326]](https://github.com/BAMWelDX/weldx/pull/326)
- Updated the measurement tutorial [[#326]](https://github.com/BAMWelDX/weldx/pull/326)


### ASDF

- fix inline array serialization for new 64bit inline limit [[#218]](https://github.com/BAMWelDX/weldx/pull/218)
Expand Down
3 changes: 2 additions & 1 deletion tutorials/timeseries_01.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,7 @@
"outputs": [],
"source": [
"def plot_ts(ts):\n",
" ts = ts.data_array\n",
" fig, ax = plt.subplots(1,1)\n",
" x = ts.time.data.astype('timedelta64[ms]').astype(int)/1000.0\n",
" plt.plot(x,ts.data.magnitude,'-o')\n",
Expand Down Expand Up @@ -397,7 +398,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.2"
"version": "3.9.4"
}
},
"nbformat": 4,
Expand Down
2 changes: 1 addition & 1 deletion tutorials/welding_example_02_weaving.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -629,7 +629,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.2"
"version": "3.9.4"
}
},
"nbformat": 4,
Expand Down
275 changes: 169 additions & 106 deletions weldx/core.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
from __future__ import annotations

from typing import TYPE_CHECKING, Any, Dict, List, Tuple, Union
from warnings import warn

import numpy as np
import pandas as pd
Expand Down Expand Up @@ -247,13 +248,21 @@ def evaluate(self, **kwargs) -> Any:
class TimeSeries:
"""Describes the behaviour of a quantity in time."""

_valid_interpolations = ["step", "linear"]
_valid_interpolations = [
"step",
"linear",
"nearest",
"zero",
"slinear",
"quadratic",
"cubic",
]

def __init__(
self,
data: Union[pint.Quantity, MathematicalExpression],
time: Union[None, pd.TimedeltaIndex, pint.Quantity] = None,
interpolation: str = "linear",
interpolation: str = None,
):
"""Construct a TimSeries.
Expand All @@ -276,68 +285,12 @@ def __init__(
self._time_var_name = None
self._shape = None
self._units = None
self._interp_counter = 0

if isinstance(data, pint.Quantity):
if not np.iterable(data): # expand dim for scalar input
data = np.expand_dims(data, 0)
if time is None: # constant value case
time = pd.TimedeltaIndex([0])
interpolation = None
elif interpolation not in self._valid_interpolations:
raise ValueError(
"A valid interpolation method must be specified if discrete "
f'values are used. "{interpolation}" is not supported'
)
if isinstance(time, pint.Quantity):
time = ut.to_pandas_time_index(time)
if not isinstance(time, pd.TimedeltaIndex):
raise ValueError(
'"time" must be a time quantity or a "pandas.TimedeltaIndex".'
)

dax = xr.DataArray(
data=data,
attrs={"interpolation": interpolation},
)
self._data = dax.rename({"dim_0": "time"}).assign_coords({"time": time})

self._initialize_discrete(data, time, interpolation)
elif isinstance(data, MathematicalExpression):

if data.num_variables != 1:
raise Exception(
"The mathematical expression must have exactly 1 free "
"variable that represents time."
)
time_var_name = data.get_variable_names()[0]
try:
eval_data = data.evaluate(**{time_var_name: Q_(1, "second")})
self._units = eval_data.units
if np.iterable(eval_data):
self._shape = eval_data.shape
else:
self._shape = (1,)
except pint.errors.DimensionalityError:
raise Exception(
"Expression can not be evaluated with "
'"weldx.Quantity(1, "seconds")"'
". Ensure that every parameter posses the correct unit."
)

self._data = data
self._time_var_name = time_var_name

try:
self.interp_time(Q_([1, 2], "second"))
self.interp_time(Q_([1, 2, 3], "second"))
except Exception as e:
raise Exception(
"The expression can not be evaluated with arrays of time deltas. "
"Ensure that all parameters that are multiplied with the time "
"variable have an outer dimension of size 1. This dimension is "
"broadcasted during multiplication. The original error message was:"
f' "{str(e)}"'
)

self._init_expression(data)
else:
raise TypeError(f'The data type "{type(data)}" is not supported.')

Expand Down Expand Up @@ -385,6 +338,135 @@ def __repr__(self):
)
return representation + f"Units:\n\t{self.units}\n"

def _initialize_discrete(
self,
data: pint.Quantity,
time: Union[None, pd.TimedeltaIndex, pint.Quantity],
interpolation: str,
):
"""Initialize the internal data with discrete values."""
# set default interpolation
if interpolation is None:
interpolation = "step"

# expand dim for scalar input
if not np.iterable(data):
data = np.expand_dims(data, 0)

# constant value case
if time is None:
time = pd.TimedeltaIndex([0])

if isinstance(time, pint.Quantity):
time = ut.to_pandas_time_index(time)
if not isinstance(time, pd.TimedeltaIndex):
raise ValueError(
'"time" must be a time quantity or a "pandas.TimedeltaIndex".'
)

dax = xr.DataArray(data=data)
self._data = dax.rename({"dim_0": "time"}).assign_coords({"time": time})
self.interpolation = interpolation

def _init_expression(self, data):
"""Initialize the internal data with a mathematical expression."""
if data.num_variables != 1:
raise Exception(
"The mathematical expression must have exactly 1 free "
"variable that represents time."
)

# check that the expression can be evaluated with a time quantity
time_var_name = data.get_variable_names()[0]
try:
eval_data = data.evaluate(**{time_var_name: Q_(1, "second")})
self._units = eval_data.units
if np.iterable(eval_data):
self._shape = eval_data.shape
else:
self._shape = (1,)
except pint.errors.DimensionalityError:
raise Exception(
"Expression can not be evaluated with "
'"weldx.Quantity(1, "seconds")"'
". Ensure that every parameter posses the correct unit."
)

# assign internal variables
self._data = data
self._time_var_name = time_var_name

# check that all parameters of the expression support time arrays
try:
self.interp_time(Q_([1, 2], "second"))
self.interp_time(Q_([1, 2, 3], "second"))
except Exception as e:
raise Exception(
"The expression can not be evaluated with arrays of time deltas. "
"Ensure that all parameters that are multiplied with the time "
"variable have an outer dimension of size 1. This dimension is "
"broadcasted during multiplication. The original error message was:"
f' "{str(e)}"'
)

def _interp_time_discrete(
self, time: Union[pd.TimedeltaIndex, pint.Quantity]
) -> xr.DataArray:
"""Interpolate the time series if its data is composed of discrete values.
See `interp_time` for interface description.
"""
if isinstance(time, pint.Quantity):
time = ut.to_pandas_time_index(time)
if not isinstance(time, pd.TimedeltaIndex):
raise ValueError(
'"time" must be a time quantity or a "pandas.TimedeltaIndex".'
)

return ut.xr_interp_like(
self._data,
{"time": time},
method=self.interpolation,
assume_sorted=False,
broadcast_missing=False,
)

def _interp_time_expression(
self, time: Union[pd.TimedeltaIndex, pint.Quantity], time_unit: str
) -> xr.DataArray:
"""Interpolate the time series if its data is a mathematical expression.
See `interp_time` for interface description.
"""
# Transform time to both formats
if isinstance(time, pint.Quantity) and time.check(UREG.get_dimensionality("s")):
time_q = time
time_pd = ut.to_pandas_time_index(time)
elif isinstance(time, pd.TimedeltaIndex):
time_q = ut.pandas_time_delta_to_quantity(time, time_unit)
time_pd = time
else:
raise ValueError(
'"time" must be a time quantity or a "pandas.TimedeltaIndex".'
)

if len(self.shape) > 1 and np.iterable(time_q):
while len(time_q.shape) < len(self.shape):
time_q = time_q[:, np.newaxis]

# evaluate expression
data = self._data.evaluate(**{self._time_var_name: time_q})
data = data.astype(float).to_reduced_units() # float conversion before reduce!

# create data array
if not np.iterable(data): # make sure quantity is not scalar value
data = np.expand_dims(data, 0)

dax = xr.DataArray(data=data) # don't know exact dimensions so far
return dax.rename({"dim_0": "time"}).assign_coords({"time": time_pd})

@property
def data(self) -> Union[pint.Quantity, MathematicalExpression]:
"""Return the data of the TimeSeries.
Expand Down Expand Up @@ -433,6 +515,18 @@ def interpolation(self) -> Union[str, None]:
return self._data.attrs["interpolation"]
return None

@interpolation.setter
def interpolation(self, interpolation):
if isinstance(self._data, xr.DataArray):
if interpolation not in self._valid_interpolations:
raise ValueError(
"A valid interpolation method must be specified if discrete "
f'values are used. "{interpolation}" is not supported'
)
if self.time is None and interpolation != "step":
interpolation = "step"
self.data_array.attrs["interpolation"] = interpolation

@property
def time(self) -> Union[None, pd.TimedeltaIndex]:
"""Return the data's timestamps.
Expand All @@ -449,7 +543,7 @@ def time(self) -> Union[None, pd.TimedeltaIndex]:

def interp_time(
self, time: Union[pd.TimedeltaIndex, pint.Quantity], time_unit: str = "s"
) -> xr.DataArray:
) -> "TimeSeries":
"""Interpolate the TimeSeries in time.
If the internal data consists of discrete values, an interpolation with the
Expand All @@ -470,55 +564,24 @@ def interp_time(
Returns
-------
xarray.DataArray:
A data array containing the interpolated data.
TimeSeries :
A new `TimeSeries` object containing the interpolated data.
"""
if isinstance(self._data, xr.DataArray):
if isinstance(time, pint.Quantity):
time = ut.to_pandas_time_index(time)
if not isinstance(time, pd.TimedeltaIndex):
raise ValueError(
'"time" must be a time quantity or a "pandas.TimedeltaIndex".'
)
# constant values are also treated by this branch
if self._data.attrs["interpolation"] == "linear" or self.shape[0] == 1:
return ut.xr_interp_like(
self._data,
{"time": time},
assume_sorted=False,
broadcast_missing=False,
)

dax = self._data.reindex({"time": time}, method="ffill")
return dax.fillna(self._data[0])

# Transform time to both formats
if isinstance(time, pint.Quantity) and time.check(UREG.get_dimensionality("s")):
time_q = time
time_pd = ut.to_pandas_time_index(time)
elif isinstance(time, pd.TimedeltaIndex):
time_q = ut.pandas_time_delta_to_quantity(time, time_unit)
time_pd = time
else:
raise ValueError(
'"time" must be a time quantity or a "pandas.TimedeltaIndex".'
if self._interp_counter > 0:
warn(
"The data of the time series has already been interpolated "
f"{self._interp_counter} time(s)."
)

if len(self.shape) > 1 and np.iterable(time_q):
while len(time_q.shape) < len(self.shape):
time_q = time_q[:, np.newaxis]

# evaluate expression
data = self._data.evaluate(**{self._time_var_name: time_q})
data = data.astype(float).to_reduced_units() # float conversion before reduce!

# create data array
if not np.iterable(data): # make sure quantity is not scalar value
data = np.expand_dims(data, 0)
if isinstance(self._data, xr.DataArray):
dax = self._interp_time_discrete(time)
else:
dax = self._interp_time_expression(time, time_unit)

dax = xr.DataArray(data=data) # don't know exact dimensions so far
return dax.rename({"dim_0": "time"}).assign_coords({"time": time_pd})
ts = TimeSeries(data=dax.data, time=time, interpolation=self.interpolation)
ts._interp_counter = self._interp_counter + 1
return ts

@property
def shape(self) -> Tuple:
Expand Down
Loading

0 comments on commit 1999c8d

Please sign in to comment.