time_split.types

Types related to splitting data.

Module Attributes

DatetimeTypes	Types that may be cast to pandas.Timestamp.
DatetimeIterable	Iterable that may be cast to pandas.DatetimeIndex.
TimedeltaTypes	Types that may be cast to pandas.Timedelta.
Schedule	User schedule type.
Span	User span type.
ExpandLimits	Limits flexibility spec for floor/ceil.
Filter	A callable (start, mid, end) -> bool used to filter folds.
DatetimeSplits	A list of bounds.
LogSplitProgressLoggerArg	A logger or string.
MetricsType	Metrics argument type.
GetMetrics	A callable (Timestamp) -> MetricsType.
FormatMetrics	A callable (end_message, metrics) -> str.

Classes

DatetimeIndexSplitterKwargs	Keyword arguments for DatetimeIndexSplitter.
DatetimeSplitBounds(start, mid, end)	A 3-tuple which denotes two adjacent datetime ranges.
DatetimeSplitCounts(data, future_data)	Relative importance of data and future_data.
LogSplitProgress()	The sequence-like return type of log_split_progress().
LogSplitProgressKwargs	Keyword arguments for by log_split_progress().
SplitProgressExtras	Named extras used for messages logged by log_split_progress().

DatetimeTypes: TypeAlias = str | pandas._libs.tslibs.timestamps.Timestamp | datetime.datetime | datetime.date | numpy.datetime64

Types that may be cast to pandas.Timestamp.

DatetimeIterable

Iterable that may be cast to pandas.DatetimeIndex.

alias of Iterable[str | Timestamp | datetime | date | datetime64]

TimedeltaTypes: TypeAlias = str | pandas._libs.tslibs.timedeltas.Timedelta | datetime.timedelta | numpy.timedelta64

Types that may be cast to pandas.Timedelta.

Schedule: TypeAlias = pandas.core.indexes.datetimes.DatetimeIndex | collections.abc.Iterable[str | pandas._libs.tslibs.timestamps.Timestamp | datetime.datetime | datetime.date | numpy.datetime64] | str | pandas._libs.tslibs.timedeltas.Timedelta | datetime.timedelta | numpy.timedelta64

User schedule type.

Span = int | typing.Literal['all', 'empty'] | str | pandas._libs.tslibs.timedeltas.Timedelta | datetime.timedelta | numpy.timedelta64

User span type. Used to determine limits from the timestamps given by a Schedule.

ExpandLimits = bool | typing.Literal['auto'] | str

Limits flexibility spec for floor/ceil. Pass False to disable.

Filter

A callable (start, mid, end) -> bool used to filter folds.

alias of Callable[[Timestamp, Timestamp, Timestamp], bool]

class DatetimeSplitBounds(start: Timestamp, mid: Timestamp, end: Timestamp)

Bases: NamedTuple

A 3-tuple which denotes two adjacent datetime ranges.

start: Timestamp

Left (inclusive) limit of the data range.

mid: Timestamp

Schedule timestamp; simulated training_date.

Right (exclusive) limit of the data range, left (inclusive) limit of the future_data range.

When using integration functions, These are available as DatetimeSplit.data and DatetimeSplit.future_data, respectively.

end: Timestamp

Right (exclusive) limit of the future_data range.

DatetimeSplits

A list of bounds.

alias of list[DatetimeSplitBounds]

class DatetimeSplitCounts(data: int, future_data: int)

Bases: NamedTuple

Relative importance of data and future_data.

data: int

Alias for field number 0

future_data: int

Alias for field number 1

class DatetimeIndexSplitterKwargs

Bases: TypedDict

Keyword arguments for DatetimeIndexSplitter.

The DatetimeIndexSplitter is a backend implementation. In most cases, it should not be used directly. See time_split.split() or one of the related functions for user-facing APIs.

schedule: Required[DatetimeIndex | Iterable[str | Timestamp | datetime | date | datetime64] | str | Timedelta | timedelta | timedelta64]

before: int | Literal['all', 'empty'] | str | Timedelta | timedelta | timedelta64

after: int | Literal['all', 'empty'] | str | Timedelta | timedelta | timedelta64

step: int

n_splits: int

expand_limits: bool | Literal['auto'] | str

filter: Callable[[Timestamp, Timestamp, Timestamp], bool] | str | None

LogSplitProgressLoggerArg = logging.Logger | logging.LoggerAdapter[typing.Any] | str

A logger or string.

class MetricsType

Metrics argument type.

alias of TypeVar(‘MetricsType’)

GetMetrics

A callable (Timestamp) -> MetricsType.

alias of Callable[[Timestamp], MetricsType]

FormatMetrics

A callable (end_message, metrics) -> str.

alias of Callable[[str, MetricsType], str]

class LogSplitProgressKwargs

Bases: TypedDict, Generic[MetricsType]

Keyword arguments for by log_split_progress().

logger: Logger | LoggerAdapter[Any] | str

start_level: int

end_level: int

extra: dict[str, Any] | None

get_metrics: Callable[[Timestamp], MetricsType]

class SplitProgressExtras

Bases: TypedDict, Generic[MetricsType]

Named extras used for messages logged by log_split_progress().

n: int

Current split number (starting at 1).

n_splits: int

Total split count.

start: str

An ISO-formatted timestamp (see DatetimeSplitBounds.start).

mid: str

An ISO-formatted timestamp (see DatetimeSplitBounds.mid).

end: str

An ISO-formatted timestamp (see DatetimeSplitBounds.end).

seconds: NotRequired[float]

User time for the fold. Available only for the fold-end message.

metrics: NotRequired[MetricsType]

Optional fold metrics. Typically appended to the fold-end message.

class LogSplitProgress

Bases: ABC, Sequence[DatetimeSplitBounds]

The sequence-like return type of log_split_progress().

A sequence [(start, mid, end), ...] which logs progress when iterating.

splits: Sequence[DatetimeSplitBounds]

The underlying splits.

logger: Logger | LoggerAdapter

Logger instance that emits messages.