tsflex.processing
Processing module.
Processing guideThe following sections will explain the processing module in detail.
Jump to API reference Working example ✅tsflex is built to be intuitive, so we encourage you to copy-paste this code and toy with some parameters!
This executable example creates a processing pipeline that contains 3 processing steps (abs, median filter, and detreding, each applied on a different subset of series).
import pandas as pd; import scipy.signal as ss; import numpy as np
from tsflex.processing import SeriesProcessor, SeriesPipeline
# 1. -------- Get your time-indexed data --------
# Data contains 3 columns; ["ACC_x", "ACC_y", "ACC_z"]
url = "https://github.com/predict-idlab/tsflex/raw/main/examples/data/empatica/"
data = pd.read_parquet(url + "acc.parquet").set_index("timestamp")
# 2 -------- Construct your processing pipeline --------
processing_pipe = SeriesPipeline(
processors=[
SeriesProcessor(function=np.abs, series_names=["ACC_x", "ACC_y"]),
SeriesProcessor(ss.medfilt, ["ACC_y", "ACC_z"], kernel_size=5)
]
)
# -- 2.1. Append processing steps to your processing pipeline
processing_pipe.append(SeriesProcessor(ss.detrend, ["ACC_x", "ACC_z"]))
# 3 -------- Process the data --------
processing_pipe.process(data=data, return_df=True)
# which outputs:
The processing functionality of tsflex is provided by a SeriesPipeline that contains SeriesProcessor steps. The processing steps are applied sequentially on the data that is passed to the processing pipeline.
As shown above, there are 2 relevant classes for processing.
function: the Callable processing-function - e.g. scipy.signal.detrendseries_names: the name(s) of the series on which the processing function should be applied**kwargs: the keyword arguments for the function.The snippet below shows how the SeriesPipeline & SeriesProcessor components work:
import numpy as np; import scipy.signal as ss
from tsflex.processing import SeriesProcessor, SeriesPipeline
# The SeriesPipeline takes a List[SeriesProcessor] as input
processing_pipe = SeriesPipeline(processors=[
SeriesProcessor(np.abs, ["series_a", "series_b"]),
SeriesProcessor(ss.medfilt, "series_b", kernel_size=5) # (with kwargs)
]
)
# We can still append processing steps after instantiating.
processing_pipe.append(processor=SeriesProcessor(ss.detrend, "series_a"))
# Apply the processing steps
processing_pipe.process(...)
The function that processes the series should match this prototype:
function(*series: pd.Series, **kwargs)
-> Union[np.ndarray, pd.Series, pd.DataFrame, List[pd.Series]]
Hence, the processing function should take one (or multiple) series as input, these may be followed by some keyword arguments. The output of a processing function can be rather versatile.
Note
A function that processes a np.ndarray instead of pd.Series should work just fine.
In this section you can find more info on advanced usage of processing functions.
Important notes 📢In a SeriesPipeline it is common behavior that series are transformed (i.e., replaced). Hence, it is important to keep the following principles in mind when:
pd.Series with a different name.SeriesProcessors) in the processing pipeline (SeriesPipeline) might affect the output.As explained above tsflex is rather versatile in terms of function input and output.
tsflex does not just allow one-to-one processing functions, but also many-to-one, one-to-many, and many-to-many functions are supported in a convenient way:
many-to-one; the processing function should
Example
def abs_diff(s1: pd.Series, s2: pd.Series) -> pd.Series:
return pd.Series(np.abs(s1-s2), name=f"abs_diff_{s1.name}-{s2.name}")
one-to-many; the processing function should
Example
def abs_square(s1: pd.Series) -> List[pd.Series]:
s1_abs = pd.Series(np.abs(s1), name=f"abs_{s1.name}")
s1_square = pd.Series(np.square(s1), name=f"square_{s1.name}")
return [s1_abs, s1_square]
many-to-many; (combination of the above) the processing function should
Example
def abs_square_diff(s1: pd.Series, s2: pd.Series) -> List[pd.Series]:
s_abs = pd.Series(np.abs(s1-s2), name=f"abs_{s1.name}-{s2.name}")
s_square = pd.Series(np.square(s1-s2), name=f"square_{s1.name}-{s2.name}")
return [s_abs, s_square]
In some (rare) cases a processing function requires a pd.DataFrame as input. For these cases we provide the dataframe_func decorator. This decorator wraps the processing function in the SeriesPipeline, provided a pd.DataFrame as input instead of multiple pd.Series.
Note
In most cases series arguments are sufficient; you can perform column-based operations on multiple pd.Series (e.g., subtract 2 series). Only when row-based operations are required (e.g., df.dropna(axis=0)), a pd.DataFrame is unavoidable.
When a logging_file_path is passed to the SeriesPipeline its process method, the execution times of the processing steps will be logged.
"""Processing module.
.. include:: ../../docs/pdoc_include/processing.md
"""
__author__ = "Jonas Van Der Donckt, Emiel Deprost, Jeroen Van Der Donckt"
from .. import __pdoc__
from .logger import get_processor_logs
from .series_pipeline import SeriesPipeline
from .series_processor import SeriesProcessor, dataframe_func
__pdoc__["SeriesProcessor.__call__"] = True
__all__ = [
"dataframe_func",
"SeriesProcessor",
"SeriesPipeline",
"get_processor_logs",
].logger
Contains the used variables and functions to provide logging functionality …
.series_pipeline
SeriesPipeline class for time-series data (pre-)processing pipeline.
.series_processor
Code for time-series data (pre-)processing.
.utils
(Advanced) utilities for the processing pipelines.
def dataframe_func(func)
def dataframe_func(func: Callable) -> Callable:
"""Decorate function to use a DataFrame instead of multiple series (as argument).
This decorator can be used for functions that need to work on a whole
`pd.DataFrame`. It will convert the required series into a DataFrame using an
**outer merge**.
The function's prototype should be:
func(df : pd.DataFrame, **kwargs)
-> Union[np.ndarray, pd.Series, pd.DataFrame, List[pd.Series]]
So the decorated `func` has to take a DataFrame as first argument.
Notes
-----
Only when you want to perform row-based operations, such as `df.dropna(axis=0)`,
this wrapper is needed.
Hence, in most cases that `func` requires a `pd.DataFrame`, series arguments would
be sufficient; as you can perform column-based operations on multiple `pd.Series`
(e.g., subtract 2 series) and most dataframe operations are also available for a
`pd.Series`.
"""
def wrapper( # type: ignore[no-untyped-def]
*series: pd.Series,
**kwargs,
) -> Union[np.ndarray, pd.Series, pd.DataFrame, List[pd.Series]]:
series_dict = {s.name: s for s in series}
df = series_dict_to_df(series_dict)
res = func(df, **kwargs)
return res
wrapper.__name__ = "dataframe_func: " + func.__name__
return wrapperDecorate function to use a DataFrame instead of multiple series (as argument).
This decorator can be used for functions that need to work on a whole pd.DataFrame. It will convert the required series into a DataFrame using an outer merge.
The function's prototype should be:
func(df : pd.DataFrame, **kwargs)
-> Union[np.ndarray, pd.Series, pd.DataFrame, List[pd.Series]]
So the decorated func has to take a DataFrame as first argument.
Only when you want to perform row-based operations, such as df.dropna(axis=0), this wrapper is needed. Hence, in most cases that func requires a pd.DataFrame, series arguments would be sufficient; as you can perform column-based operations on multiple pd.Series (e.g., subtract 2 series) and most dataframe operations are also available for a pd.Series.
def get_processor_logs(logging_file_path)
def get_processor_logs(logging_file_path: str) -> pd.DataFrame:
"""Get execution (time) info for each processor of a ``SeriesPipeline``.
Parameters
----------
logging_file_path: str
The file path where the logged messages are stored. This is the file path that
is passed to the ``SeriesPipeline`` its ``process`` method.
Returns
-------
pd.DataFrame
A DataFrame containing each processor its series names, output names, and
(%) duration.
"""
df = _parse_logging_execution_to_df(logging_file_path)
df["duration"] = pd.to_timedelta(df["duration"], unit="s")
return dfGet execution (time) info for each processor of a SeriesPipeline.
logging_file_path : str
SeriesPipeline its process method.
pd.DataFrame
class SeriesProcessor (function, series_names, **kwargs)
class SeriesProcessor(FrozenClass):
"""Class that executes a specific operation on the passed series_dict.
Parameters
----------
function : Callable
The function that processes the series (given in the `series_names`).
The prototype of the function should match: \n
function(*series: pd.Series, **kwargs)
-> Union[np.ndarray, pd.Series, pd.DataFrame, List[pd.Series]]
.. note::
A function that processes a ``np.ndarray`` instead of ``pd.Series``
should work just fine.
series_names : Union[str, Tuple[str, ...], List[str], List[Tuple[str, ...]]]
The names of the series on which the processing function should be applied.
This argument should match the `function` its input; \n
* If `series_names` is a (list of) string (or tuple of a single string),
than `function` should require just one series as input.
* If `series_names` is a (list of) tuple of strings, than `function` should
require `len(tuple)` series as input **and in exactly the same order**.
A list means multiple series (combinations) to process; \n
* If `series_names` is a string or a tuple of strings, than `function` will
be called only once for the series of this argument.
* If `series_names` is a list of either strings or tuple of strings, than
`function` will be called for each entry of this list.
.. note::
when passing a list as `series_names`, all items in this list should
have the same type, i.e, either \n
* all a `str`
* or, all a `tuple` _with same length_. \n
**kwargs: dict, optional
Keyword arguments which will be also passed to the `function`
Notes
-----
If the output of `function` is a `np.ndarray` or a `pd.Series` without a name,
than (items of) the given `series_names` must have length 1, i.e., the function
requires just 1 series! That series its name and index are used to transform
(i.e., **replace**) that series.
If you want to transform (i.e., **replace**) the input series with the
processor, than `function` should return either: \n
* a `np.ndarray` (see above).
* a `pd.Series` with no name or with the same name as the input series.
* a `pd.DataFrame` with (one) column name equal to the input series its name.
* a list of `pd.Series` in which (exact) one series has the same name as the
input series.
Series (& columns) with other (column) names will be added to the series dict.
"""
def __init__( # type: ignore[no-untyped-def]
self,
function: Callable,
series_names: Union[str, Tuple[str, ...], List[str], List[Tuple[str, ...]]],
**kwargs,
):
series_names = [to_tuple(names) for names in to_list(series_names)]
# Assert that function inputs (series) all have the same length
assert all(
len(series_names[0]) == len(series_name_tuple)
for series_name_tuple in series_names
)
self.series_names: List[Tuple[str, ...]] = series_names
self.function = function
self.name = self.function.__name__
self.kwargs = kwargs
self._freeze()
def get_required_series(self) -> List[str]:
"""Return all required series names for this processor.
Return the list of series names that are required in order to execute the
processing function.
Returns
-------
List[str]
List of all the required series names.
"""
return list(set(flatten(name for name in self.series_names)))
def __call__(self, series_dict: Dict[str, pd.Series]) -> Dict[str, pd.Series]:
"""**Call**culates the processed series.
Parameters
----------
series_dict : Dict[str, pd.Series]
A dict of `pd.Series` containing the data that need to be processed.
The key should always be the accompanying series its name.
Returns
-------
Dict[str, pd.Series]
The processed `series_dict`.
Raises
------
KeyError
Raised when a key is not present in the `series_dict` but required for the
processing.
TypeError
Raised when the output of the `SeriesProcessor` is not of the correct type.
Notes
-----
* The `series_dict` is an internal representation of the time-series data .
This internal representation is constructed in the `process` method of the
`SeriesPipeline`.
* If you want to test or debug your `SeriesProcessor` object, just encapsulate
your instance of this class in a `SeriesPipeline`. The latter allows more
versatile input for its `process` method.
"""
t_start = time.perf_counter()
# Only selecting the series that are needed for this processing step
# requested_dict = {}
# try:
# for sig in self.get_required_series():
# requested_dict[sig] = series_dict[sig]
# except KeyError as key:
# # Re raise error as we can't continue
# raise KeyError(
# "Key %s is not present in the input dict and needed for processor %s"
# % (key, self.name)
# )
# Variable that will contain the final output of this method
processed_output: Dict[str, pd.Series] = {}
def get_series_list(keys: Tuple[str, ...]) -> List[pd.Series]:
"""Get an ordered series list view for the given keys."""
return [series_dict[key] for key in keys]
def get_series_dict(keys: Tuple[str, ...]) -> Dict[str, pd.Series]:
"""Get a series dict view for the given keys."""
return {key: series_dict[key] for key in keys}
for series_name_tuple in self.series_names:
func_output = self.function(
*get_series_list(series_name_tuple), **self.kwargs
)
func_output = _handle_seriesprocessor_func_output(
func_output,
get_series_dict(series_name_tuple),
self.name,
)
# Check that the output of the function call produces unique columns / keys
assert (
len(set(processed_output.keys()).intersection(func_output.keys())) == 0
)
processed_output.update(func_output)
elapsed = time.perf_counter() - t_start
logger.info(
f"Finished function [{self.name}] on {self.series_names} with output "
f"{list(processed_output.keys())} in [{elapsed} seconds]!"
)
return processed_output
def __repr__(self) -> str:
"""Return formal representation of object."""
repr_str = self.name + (" " + str(self.kwargs))
repr_str += " : " + " ".join([str(s) for s in self.series_names])
return repr_str
def __str__(self) -> str:
"""Return informal representation of object."""
return self.__repr__()Class that executes a specific operation on the passed series_dict.
Parametersfunction : Callable
The function that processes the series (given in the series_names). The prototype of the function should match:
function(*series: pd.Series, **kwargs)
-> Union[np.ndarray, pd.Series, pd.DataFrame, List[pd.Series]]
Note
A function that processes a np.ndarray instead of pd.Series should work just fine.
series_names : Union[str, Tuple[str, …], List[str], List[Tuple[str, …]]]
The names of the series on which the processing function should be applied.
This argument should match the function its input;
series_names is a (list of) string (or tuple of a single string), than function should require just one series as input.series_names is a (list of) tuple of strings, than function should require len(tuple) series as input and in exactly the same order.A list means multiple series (combinations) to process;
series_names is a string or a tuple of strings, than function will be called only once for the series of this argument.series_names is a list of either strings or tuple of strings, than function will be called for each entry of this list.Note
when passing a list as series_names, all items in this list should have the same type, i.e, either
strtuple with same length.**kwargs : dict, optional
function
If the output of function is a np.ndarray or a pd.Series without a name, than (items of) the given series_names must have length 1, i.e., the function requires just 1 series! That series its name and index are used to transform (i.e., replace) that series.
If you want to transform (i.e., replace) the input series with the processor, than function should return either:
np.ndarray (see above).pd.Series with no name or with the same name as the input series.pd.DataFrame with (one) column name equal to the input series its name.pd.Series in which (exact) one series has the same name as the input series.Series (& columns) with other (column) names will be added to the series dict.
Ancestors def get_required_series(self)
def get_required_series(self) -> List[str]:
"""Return all required series names for this processor.
Return the list of series names that are required in order to execute the
processing function.
Returns
-------
List[str]
List of all the required series names.
"""
return list(set(flatten(name for name in self.series_names)))Return all required series names for this processor.
Return the list of series names that are required in order to execute the processing function.
ReturnsList[str]
def __call__(self, series_dict)
def __call__(self, series_dict: Dict[str, pd.Series]) -> Dict[str, pd.Series]:
"""**Call**culates the processed series.
Parameters
----------
series_dict : Dict[str, pd.Series]
A dict of `pd.Series` containing the data that need to be processed.
The key should always be the accompanying series its name.
Returns
-------
Dict[str, pd.Series]
The processed `series_dict`.
Raises
------
KeyError
Raised when a key is not present in the `series_dict` but required for the
processing.
TypeError
Raised when the output of the `SeriesProcessor` is not of the correct type.
Notes
-----
* The `series_dict` is an internal representation of the time-series data .
This internal representation is constructed in the `process` method of the
`SeriesPipeline`.
* If you want to test or debug your `SeriesProcessor` object, just encapsulate
your instance of this class in a `SeriesPipeline`. The latter allows more
versatile input for its `process` method.
"""
t_start = time.perf_counter()
# Only selecting the series that are needed for this processing step
# requested_dict = {}
# try:
# for sig in self.get_required_series():
# requested_dict[sig] = series_dict[sig]
# except KeyError as key:
# # Re raise error as we can't continue
# raise KeyError(
# "Key %s is not present in the input dict and needed for processor %s"
# % (key, self.name)
# )
# Variable that will contain the final output of this method
processed_output: Dict[str, pd.Series] = {}
def get_series_list(keys: Tuple[str, ...]) -> List[pd.Series]:
"""Get an ordered series list view for the given keys."""
return [series_dict[key] for key in keys]
def get_series_dict(keys: Tuple[str, ...]) -> Dict[str, pd.Series]:
"""Get a series dict view for the given keys."""
return {key: series_dict[key] for key in keys}
for series_name_tuple in self.series_names:
func_output = self.function(
*get_series_list(series_name_tuple), **self.kwargs
)
func_output = _handle_seriesprocessor_func_output(
func_output,
get_series_dict(series_name_tuple),
self.name,
)
# Check that the output of the function call produces unique columns / keys
assert (
len(set(processed_output.keys()).intersection(func_output.keys())) == 0
)
processed_output.update(func_output)
elapsed = time.perf_counter() - t_start
logger.info(
f"Finished function [{self.name}] on {self.series_names} with output "
f"{list(processed_output.keys())} in [{elapsed} seconds]!"
)
return processed_outputCallculates the processed series.
Parametersseries_dict : Dict[str, pd.Series]
pd.Series containing the data that need to be processed. The key should always be the accompanying series its name.
Dict[str, pd.Series]
series_dict.
KeyError
series_dict but required for the processing.
TypeError
SeriesProcessor is not of the correct type.
series_dict is an internal representation of the time-series data . This internal representation is constructed in the process method of the SeriesPipeline.SeriesProcessor object, just encapsulate your instance of this class in a SeriesPipeline. The latter allows more versatile input for its process method. class SeriesPipeline (processors=None)
class SeriesPipeline:
"""Pipeline for applying ``SeriesProcessor`` objects sequentially.
Parameters
----------
processors : List[Union[SeriesProcessor, SeriesPipeline]], optional
List of ``SeriesProcessor`` or ``SeriesPipeline`` instances that will be applied
sequentially to the internal series dict, by default None.
**The processing steps will be executed in the same order as passed in this
list.**
"""
def __init__(
self, processors: Optional[List[Union[SeriesProcessor, SeriesPipeline]]] = None
):
self.processing_steps: List[SeriesProcessor] = [] # TODO: dit private of niet?
if processors is not None:
assert isinstance(processors, list)
self.processing_steps = list(
flatten(
[
p.processing_steps if isinstance(p, SeriesPipeline) else [p]
for p in processors
]
)
)
def get_required_series(self) -> List[str]:
"""Return all required series names for this pipeline.
Return the list of series names that are required in order to execute all the
``SeriesProcessor`` objects of this processing pipeline.
Returns
-------
List[str]
List of all the required series names.
"""
return list(
set(flatten(step.get_required_series() for step in self.processing_steps))
)
def append(self, processor: Union[SeriesProcessor, SeriesPipeline]) -> None:
"""Append a ``SeriesProcessor`` at the end of the pipeline.
Parameters
----------
processor : Union[SeriesProcessor, SeriesPipeline]
The ``SeriesProcessor`` or ``SeriesPipeline`` that will be added to the
end of the pipeline.
"""
if isinstance(processor, SeriesProcessor):
self.processing_steps.append(processor)
elif isinstance(processor, SeriesPipeline):
self.processing_steps.extend(processor.processing_steps)
else:
raise TypeError(
"Can only append SeriesProcessor or SeriesPipeline, "
+ f"not {type(processor)}"
)
def insert(
self, idx: int, processor: Union[SeriesProcessor, SeriesPipeline]
) -> None:
"""Insert a ``SeriesProcessor`` at the given index in the pipeline.
Parameters
----------
idx : int
The index where the given processor should be inserted in the pipeline.
Index 0 will insert the given processor at the front of the pipeline,
and index ``len(pipeline)`` is equivalent to appending the processor.
processor : Union[SeriesProcessor, SeriesPipeline]
The ``SeriesProcessor`` or ``SeriesPipeline`` that will be inserted.<br>
.. note::
If the given processor is a ``SeriesPipeline``, all its processors will
be inserted sequentially, starting from the given index.
"""
if isinstance(processor, SeriesProcessor):
self.processing_steps.insert(idx, processor)
elif isinstance(processor, SeriesPipeline):
for i, ps in enumerate(processor.processing_steps):
self.insert(idx + i, ps)
else:
raise TypeError(
"Can only insert a SeriesProcessor or SeriesPipeline, "
+ f"not {type(processor)}"
)
def process(
self,
data: Union[pd.Series, pd.DataFrame, List[Union[pd.Series, pd.DataFrame]]],
return_df: Optional[bool] = False,
return_all_series: Optional[bool] = True,
drop_keys: Optional[List[str]] = None,
copy: Optional[bool] = False,
logging_file_path: Optional[Union[str, Path]] = None,
) -> Union[List[pd.Series], pd.DataFrame]:
"""Execute all ``SeriesProcessor`` objects in pipeline sequentially.
Apply all the processing steps on passed Series list or DataFrame and return the
preprocessed Series list or DataFrame.
Parameters
----------
data : Union[pd.Series, pd.DataFrame, List[Union[pd.Series, pd.DataFrame]]]
Dataframe or Series or list thereof, with all the required data for the
processing steps. \n
**Remark**: each Series / DataFrame must have a ``pd.DatetimeIndex``.
**Remark**: we assume that each name / column is unique.
return_df : bool, optional
Whether the output needs to be a series list or a DataFrame, by default
False.
If True the output series will be combined to a DataFrame with an outer
merge.
return_all_series : bool, optional
Whether the output needs to return all the series, by default True.
* If True the output will contain all series that were passed to this
method.
* If False the output will contain just the required series (see
``get_required_series``).
drop_keys : List[str], optional
Which keys should be dropped when returning the output, by default None.
copy : bool, optional
Whether the series in ``data`` should be copied, by default False.
logging_file_path : Union[str, Path], optional
The file path where the logged messages are stored, by default None.
If ``None``, then no logging ``FileHandler`` will be used and the logging
messages are only pushed to stdout. Otherwise, a logging ``FileHandler`` will
write the logged messages to the given file path.
Returns
-------
Union[List[pd.Series], pd.DataFrame]
The preprocessed series.
Notes
-----
* If a ``logging_file_path`` is provided, the execution (time) info can be
retrieved by calling ``logger.get_processor_logs(logging_file_path)``. <br>
Be aware that the ``logging_file_path`` gets cleared before the logger pushes
logged messages. Hence, one should use a separate logging file for each
constructed processing and feature instance with this library.
* If a series processor its function output is a ``np.ndarray``, the input series
dict (required dict for that function) must contain just 1 series! That series
its name and index are used to return a series dict. When a user does not want
a numpy array to replace its input series, it is his / her responsibility to
create a new ``pd.Series`` (or ``pd.DataFrame``) of that numpy array with a
different (column) name.
* If ``func_output`` is a ``pd.Series``, keep in mind that the input series gets
transformed (i.e., replaced) in the pipeline with the ``func_output`` when the
series name is equal.
Raises
------
_ProcessingError
Error raised when a processing step fails.
"""
# Delete other logging handlers
delete_logging_handlers(logger)
# Add logging handler (if path provided)
if logging_file_path:
f_handler = add_logging_handler(logger, logging_file_path)
# Convert the data to a series_dict
series_dict: Dict[str, pd.Series] = {}
for s in to_series_list(data):
# Assert the assumptions we make!
if len(s):
assert isinstance(s.index, pd.DatetimeIndex)
# TODO: also check monotonic increasing?
if s.name in self.get_required_series():
series_dict[str(s.name)] = s.copy() if copy else s
elif return_all_series:
# If all the series have to be returned
series_dict[str(s.name)] = s.copy() if copy else s
output_keys: Set[str] = set() # Maintain set of output series
for processor in self.processing_steps:
try:
processed_dict = processor(series_dict)
output_keys.update(processed_dict.keys())
series_dict.update(processed_dict)
except Exception as e:
# Close the file handler (this avoids PermissionError: [WinError 32])
if logging_file_path:
f_handler.close()
logger.removeHandler(f_handler)
raise _ProcessingError(
"Error while processing function {}:\n {}".format(
processor.name, str(e)
)
) from e
# Close the file handler (this avoids PermissionError: [WinError 32])
if logging_file_path:
f_handler.close()
logger.removeHandler(f_handler)
if not return_all_series:
# Return just the output series
output_dict = {key: series_dict[str(key)] for key in output_keys}
series_dict = output_dict
if drop_keys is not None:
# Drop the keys that should not be included in the output
output_dict = {
key: series_dict[key]
for key in set(series_dict.keys()).difference(drop_keys)
}
series_dict = output_dict
if return_df:
# We merge the series dict into a DataFrame
return series_dict_to_df(series_dict)
else:
return [s for s in series_dict.values()]
def serialize(self, file_path: Union[str, Path]) -> None:
"""Serialize this ``SeriesPipeline`` instance.
Notes
------
As we use [Dill](https://github.com/uqfoundation/dill){:target="_blank"} to
serialize, we can also serialize (decorator)functions which are defined in the
local scope, like lambdas.
Parameters
----------
file_path : Union[str, Path]
The path where the ``SeriesProcessor`` will be serialized.
"""
with open(file_path, "wb") as f:
dill.dump(self, f, recurse=True)
def __repr__(self) -> str:
"""Return formal representation of object."""
return "[\n" + "".join([f"\t{str(p)}\n" for p in self.processing_steps]) + "]"
def __str__(self) -> str:
"""Return informal representation of object."""
return self.__repr__() def get_required_series(self)
def get_required_series(self) -> List[str]:
"""Return all required series names for this pipeline.
Return the list of series names that are required in order to execute all the
``SeriesProcessor`` objects of this processing pipeline.
Returns
-------
List[str]
List of all the required series names.
"""
return list(
set(flatten(step.get_required_series() for step in self.processing_steps))
)Return all required series names for this pipeline.
Return the list of series names that are required in order to execute all the SeriesProcessor objects of this processing pipeline.
List[str]
def append(self, processor)
def append(self, processor: Union[SeriesProcessor, SeriesPipeline]) -> None:
"""Append a ``SeriesProcessor`` at the end of the pipeline.
Parameters
----------
processor : Union[SeriesProcessor, SeriesPipeline]
The ``SeriesProcessor`` or ``SeriesPipeline`` that will be added to the
end of the pipeline.
"""
if isinstance(processor, SeriesProcessor):
self.processing_steps.append(processor)
elif isinstance(processor, SeriesPipeline):
self.processing_steps.extend(processor.processing_steps)
else:
raise TypeError(
"Can only append SeriesProcessor or SeriesPipeline, "
+ f"not {type(processor)}"
) def insert(self, idx, processor)
def insert(
self, idx: int, processor: Union[SeriesProcessor, SeriesPipeline]
) -> None:
"""Insert a ``SeriesProcessor`` at the given index in the pipeline.
Parameters
----------
idx : int
The index where the given processor should be inserted in the pipeline.
Index 0 will insert the given processor at the front of the pipeline,
and index ``len(pipeline)`` is equivalent to appending the processor.
processor : Union[SeriesProcessor, SeriesPipeline]
The ``SeriesProcessor`` or ``SeriesPipeline`` that will be inserted.<br>
.. note::
If the given processor is a ``SeriesPipeline``, all its processors will
be inserted sequentially, starting from the given index.
"""
if isinstance(processor, SeriesProcessor):
self.processing_steps.insert(idx, processor)
elif isinstance(processor, SeriesPipeline):
for i, ps in enumerate(processor.processing_steps):
self.insert(idx + i, ps)
else:
raise TypeError(
"Can only insert a SeriesProcessor or SeriesPipeline, "
+ f"not {type(processor)}"
)Insert a SeriesProcessor at the given index in the pipeline.
idx : int
len(pipeline) is equivalent to appending the processor.
processor : Union[SeriesProcessor, SeriesPipeline]
SeriesProcessor or SeriesPipeline that will be inserted.
Note
If the given processor is a
SeriesPipeline
, all its processors will be inserted sequentially, starting from the given index.
def process(self, data, return_df=False, return_all_series=True, drop_keys=None, copy=False, logging_file_path=None)
def process(
self,
data: Union[pd.Series, pd.DataFrame, List[Union[pd.Series, pd.DataFrame]]],
return_df: Optional[bool] = False,
return_all_series: Optional[bool] = True,
drop_keys: Optional[List[str]] = None,
copy: Optional[bool] = False,
logging_file_path: Optional[Union[str, Path]] = None,
) -> Union[List[pd.Series], pd.DataFrame]:
"""Execute all ``SeriesProcessor`` objects in pipeline sequentially.
Apply all the processing steps on passed Series list or DataFrame and return the
preprocessed Series list or DataFrame.
Parameters
----------
data : Union[pd.Series, pd.DataFrame, List[Union[pd.Series, pd.DataFrame]]]
Dataframe or Series or list thereof, with all the required data for the
processing steps. \n
**Remark**: each Series / DataFrame must have a ``pd.DatetimeIndex``.
**Remark**: we assume that each name / column is unique.
return_df : bool, optional
Whether the output needs to be a series list or a DataFrame, by default
False.
If True the output series will be combined to a DataFrame with an outer
merge.
return_all_series : bool, optional
Whether the output needs to return all the series, by default True.
* If True the output will contain all series that were passed to this
method.
* If False the output will contain just the required series (see
``get_required_series``).
drop_keys : List[str], optional
Which keys should be dropped when returning the output, by default None.
copy : bool, optional
Whether the series in ``data`` should be copied, by default False.
logging_file_path : Union[str, Path], optional
The file path where the logged messages are stored, by default None.
If ``None``, then no logging ``FileHandler`` will be used and the logging
messages are only pushed to stdout. Otherwise, a logging ``FileHandler`` will
write the logged messages to the given file path.
Returns
-------
Union[List[pd.Series], pd.DataFrame]
The preprocessed series.
Notes
-----
* If a ``logging_file_path`` is provided, the execution (time) info can be
retrieved by calling ``logger.get_processor_logs(logging_file_path)``. <br>
Be aware that the ``logging_file_path`` gets cleared before the logger pushes
logged messages. Hence, one should use a separate logging file for each
constructed processing and feature instance with this library.
* If a series processor its function output is a ``np.ndarray``, the input series
dict (required dict for that function) must contain just 1 series! That series
its name and index are used to return a series dict. When a user does not want
a numpy array to replace its input series, it is his / her responsibility to
create a new ``pd.Series`` (or ``pd.DataFrame``) of that numpy array with a
different (column) name.
* If ``func_output`` is a ``pd.Series``, keep in mind that the input series gets
transformed (i.e., replaced) in the pipeline with the ``func_output`` when the
series name is equal.
Raises
------
_ProcessingError
Error raised when a processing step fails.
"""
# Delete other logging handlers
delete_logging_handlers(logger)
# Add logging handler (if path provided)
if logging_file_path:
f_handler = add_logging_handler(logger, logging_file_path)
# Convert the data to a series_dict
series_dict: Dict[str, pd.Series] = {}
for s in to_series_list(data):
# Assert the assumptions we make!
if len(s):
assert isinstance(s.index, pd.DatetimeIndex)
# TODO: also check monotonic increasing?
if s.name in self.get_required_series():
series_dict[str(s.name)] = s.copy() if copy else s
elif return_all_series:
# If all the series have to be returned
series_dict[str(s.name)] = s.copy() if copy else s
output_keys: Set[str] = set() # Maintain set of output series
for processor in self.processing_steps:
try:
processed_dict = processor(series_dict)
output_keys.update(processed_dict.keys())
series_dict.update(processed_dict)
except Exception as e:
# Close the file handler (this avoids PermissionError: [WinError 32])
if logging_file_path:
f_handler.close()
logger.removeHandler(f_handler)
raise _ProcessingError(
"Error while processing function {}:\n {}".format(
processor.name, str(e)
)
) from e
# Close the file handler (this avoids PermissionError: [WinError 32])
if logging_file_path:
f_handler.close()
logger.removeHandler(f_handler)
if not return_all_series:
# Return just the output series
output_dict = {key: series_dict[str(key)] for key in output_keys}
series_dict = output_dict
if drop_keys is not None:
# Drop the keys that should not be included in the output
output_dict = {
key: series_dict[key]
for key in set(series_dict.keys()).difference(drop_keys)
}
series_dict = output_dict
if return_df:
# We merge the series dict into a DataFrame
return series_dict_to_df(series_dict)
else:
return [s for s in series_dict.values()]Execute all SeriesProcessor objects in pipeline sequentially.
Apply all the processing steps on passed Series list or DataFrame and return the preprocessed Series list or DataFrame.
Parametersdata : Union[pd.Series, pd.DataFrame, List[Union[pd.Series, pd.DataFrame]]]
Dataframe or Series or list thereof, with all the required data for the processing steps.
Remark: each Series / DataFrame must have a pd.DatetimeIndex. Remark: we assume that each name / column is unique.
return_df : bool, optional
return_all_series : bool, optional
get_required_series).
drop_keys : List[str], optional
copy : bool, optional
data should be copied, by default False.
logging_file_path : Union[str, Path], optional
None, then no logging FileHandler will be used and the logging messages are only pushed to stdout. Otherwise, a logging FileHandler will write the logged messages to the given file path.
Union[List[pd.Series], pd.DataFrame]
logging_file_path is provided, the execution (time) info can be retrieved by calling get_processor_logs()(logging_file_path). logging_file_path gets cleared before the logger pushes logged messages. Hence, one should use a separate logging file for each constructed processing and feature instance with this library.np.ndarray, the input series dict (required dict for that function) must contain just 1 series! That series its name and index are used to return a series dict. When a user does not want a numpy array to replace its input series, it is his / her responsibility to create a new pd.Series (or pd.DataFrame) of that numpy array with a different (column) name.func_output is a pd.Series, keep in mind that the input series gets transformed (i.e., replaced) in the pipeline with the func_output when the series name is equal._ProcessingError
def serialize(self, file_path)
def serialize(self, file_path: Union[str, Path]) -> None:
"""Serialize this ``SeriesPipeline`` instance.
Notes
------
As we use [Dill](https://github.com/uqfoundation/dill){:target="_blank"} to
serialize, we can also serialize (decorator)functions which are defined in the
local scope, like lambdas.
Parameters
----------
file_path : Union[str, Path]
The path where the ``SeriesProcessor`` will be serialized.
"""
with open(file_path, "wb") as f:
dill.dump(self, f, recurse=True)Serialize this SeriesPipeline instance.
As we use Dill to serialize, we can also serialize (decorator)functions which are defined in the local scope, like lambdas.
Parametersfile_path : Union[str, Path]
SeriesProcessor will be serialized.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4