tslumen.profile.base module¶

Base functionality for the profile package, including a dataclass for representing results and a decorator for allowing profiling functions to be employed correctly, as well as automatically capturing execution information.

class tslumen.profile.base.BundledProfiler(config: Optional[Dict[str, Dict[str, Any]]] = None, scheduler: Optional[tslumen.scheduling.Scheduler] = None)[source]¶

Bases: object

Base class for creating bundled profilers.

Contains functionality to manage configurations, orchestrate the execution of the profiling functions and collate the results. It’s up for the subclasses to decide on which profiling functions to include, done so by assigning a list with said profiling functions to the class variable _profilers.

Parameters

config (Optional[dict]) – Profiler functions’ configurations. Whatever options are provided here will override the defaults. Expects a 2-level dictionary with the first level indexed by the name of the profiling functions and the second level the function’s parameters. Assumed configs (i.e. defaults + overrides) are stored in the class variable config.
scheduler (Optional[Scheduler]) – For executing the profiling functions. Instantiates the default one if not provided.

classmethod get_config_defaults(as_dict: bool = True) → Dict[str, Any][source]¶

Returns: Bundled configurations (defaults), profiler name => configuration.
Return type: dict

classmethod get_profilers(target: str = 'any') → Dict[str, tslumen.profile.base.ProfilingFunction][source]¶

Parameters: target – {‘series’, ‘frame’, ‘any’}, default ‘any’ Filter by function target (single series, data frame, or any/no filter).
Returns: Bundled profilers, profiler name => profiler function.
Return type: dict

profile(df: pandas.core.frame.DataFrame) → tslumen.profile.base.BundledResult [source]¶

Executes the profiling functions on the supplied DataFrame.

Parameters: df (pd.DataFrame) – TimeSeries data to be profiled.
Returns: Result of the profiling.
Return type: ProfileResult

class tslumen.profile.base.BundledResult[source]¶

Bases: tslumen.profile.base.ProfileResult

result: tslumen.profile.base.BundledResultDetails¶

class tslumen.profile.base.BundledResultDetails[source]¶

Bases: tslumen.profile.base._DCDict

config: Dict[str, Dict[str, Any]]¶

exec_details: pandas.core.frame.DataFrame¶

frame: Dict[str, Any]¶

series: Dict[str, Dict[str, Any]]¶

exception tslumen.profile.base.ProfileException[source]¶: Bases: BaseException

class tslumen.profile.base.ProfileResult[source]¶

Bases: tslumen.profile.base._DCDict

For recording the results of a profiling operation.

end: datetime.datetime¶

exception: Optional[tslumen.profile.base.ProfileException]¶

name: str¶

result: Optional[Any]¶

scope: str¶

start: datetime.datetime¶

success: bool¶

target: str¶

warnings: List[warnings.WarningMessage]¶

class tslumen.profile.base.ProfilingFunction(fn: Callable)[source]¶

Bases: object

Decorator for turning a function into a profiler. Inspects the function’s signature and builds a dataclass representing its configurations (for integrating with Hydra). Adds two attributes is_pseries and is_pframe to distinguish whether it targets Series or DataFrames. This is inferred based on the type annotation of the first parameter. Captures the start and end timestamps, the return value, as well as any warnings or exceptions raised during the execution.

Parameters: fn (Callable) – Profiler function to decorate.

config: Any¶

fn: Callable¶

is_pframe: bool = False¶

is_pseries: bool = False¶

sig: inspect.Signature¶

tslumen.profile.base.valid_timeseries(df: pandas.core.frame.DataFrame) → pandas.core.frame.DataFrame[source]¶

Checks if the DataFrame meets the necessary criteria to be profiled: be a Pandas DataFrame, have values (must be numeric), have a DateTimeIndex, have a recognizable frequency/period.

Parameters: df – Pandas DataFrame.
Returns: A copy of the input containing only numeric series and a sorted index.
Return type: pandas.DataFrame