time_split.settings

Global settings for the splitting logic.

Classes

auto_expand_limits()	Configuration for the 'auto' ExpandLimits logic.
log_split_progress()	Global settings for the log_split_progress()-function.
misc()	Miscellaneous settings that don't fit in any other namespace.
plot()	Global settings for the time_split.plot()-function.

class auto_expand_limits

Bases: object

Configuration for the ‘auto’ ExpandLimits logic.

This class determines how (min, max)-tuples are expanded when expanded limits are enabled.

SANITY_CHECK: bool = True

If True, use original limits if expanded limits do not pass sanity checks.

classmethod set_level(level: Literal['hour', 'day'], *, start_at: str | Timedelta | timedelta | timedelta64, round_to: str | Timedelta | timedelta | timedelta64, tolerance: str | Timedelta | timedelta | timedelta64) → None

Set a Level used by the auto-expand_limits logic.

Inputs are not verified until an actual split is made.

Parameters:

• level – The level to set; ‘hour’ or ‘day’.

• start_at – Span size at which this level starts.

• round_to – Frequency to round the range limits to.

• tolerance – Maximum amount by which to alter limits.

Raises:

AttributeError – For unknown level names.

class Level(start_at: str | Timedelta | timedelta | timedelta64, round_to: str | Timedelta | timedelta | timedelta64, tolerance: str | Timedelta | timedelta | timedelta64)

Bases: NamedTuple

Level type used by auto_expand_limits.

start_at: str | Timedelta | timedelta | timedelta64

Span size at which this level starts.

round_to: str | Timedelta | timedelta | timedelta64

Frequency to round the range limits to.

tolerance: str | Timedelta | timedelta | timedelta64

Maximum amount by which to alter limits.

hour: Level = ('6 hours', 'hour', '15 min')

Conditions under which bounds are rounded to the nearest hour.

Default setting:

Round to hour if the total range is at least 6 hours, but do not move the bounds more than 15 minutes.

day: Level = ('2 days', 'day', '3 hours')

Conditions under which bounds are rounded to the nearest day.

Default setting:

Round to day if the total range is at least 2 days, but do not move the bounds more than 3 hours.

class plot

Bases: object

Global settings for the time_split.plot()-function.

THOUSANDS_SEPARATOR: str = "'"

Separator to use when printing bar_labels.

THOUSANDS_SEPARATOR_CUTOFF: int = 10000

Minimum value before bar_labels include a THOUSANDS_SEPARATOR.

ROW_UNIT: str = 'rows'

Unit to append to the count when displaying number of rows on the bars.

DATA_LABEL: str = 'Data'

Label of the blue bar; data in the the [start, mid)-range.

FUTURE_DATA_LABEL: str = 'Future data'

Label of the red bar; data in the the [mid, end)-range.

DEFAULT_TIME_UNIT: str = 'hours'

Time unit to use by default when bar_labels=True and available=None or a limits tuple.

REMOVED_FOLD_STYLE: ClassVar[dict[str, Any]] = {'alpha': 0.35, 'height': 0.6}

Keyword arguments used to distinguish filtered folds when plotting with show_removed=True.

See matplotlib.pyplot.bar() for details.

class log_split_progress

Bases: object

Global settings for the log_split_progress()-function.

FOLD_FORMAT: str = "'{start.auto}' <= [schedule: '{mid.auto}' ({mid:%A})] < '{end.auto}'"

Pretty-printed fold-key for other messages.

• Only the start, mid, and end keys are available (see DatetimeSplitBounds). You may use <key>.auto to format as a date when time is zero (this is the default).

Sample output.

'2021-12-30' <= [schedule: '2022-01-04' (Tuesday)] < '2022-01-04 18:00:00'

SECONDS_FORMATTER: Callable[[float], str] | None = None

A callable (seconds) -> formatted_seconds. Use rics.strings.format_seconds() if None.

Both seconds and formatted_seconds will be available to the END_MESSAGE message.

START_MESSAGE: str = 'Begin fold {n}/{n_splits}: {fold}.'

Message indicating that the current fold has been yielded to the user.

Has access to all keys from the previous section, as well as:

• The fold key (see FOLD_FORMAT), and

• The n key, which is the 1-based position of the fold in splits, and

• The n_folds key, which is just len(splits).

Sample output.

 Begin fold 5/7: '2021-12-30' <= [schedule: '2022-01-04' (Tuesday)] < '2022-01-04 18:00:00'.

END_MESSAGE: str = "Finished fold {n}/{n_splits}: [schedule: '{mid.auto}' ({mid:%A})] after {formatted_seconds}."

Message indicating that the user is done with the current fold.

Has access to all keys from the previous sections, as well as:

• The seconds key, which is the (fractional) time the user spent in the fold, and

• The formatted_seconds key, obtained using the SECONDS_FORMATTER.

The value of seconds is obtained using time.perf_counter().

Sample output.

 Finished fold 5/7: [schedule: '2022-01-04' (Tuesday)] after 5m 21s.

AUTO_DATE_FORMAT = '%Y-%m-%d'

Short-form timestamp format_spec used by <key>.auto.

AUTO_DATETIME_FORMAT = '%Y-%m-%d %H:%M:%S'

Long-form timestamp format_spec used by <key>.auto.

FORMAT_METRICS: Callable[[str, Any], str] | None = None

A callable (end_message, metrics) -> str. Use the default formatter if None.

class misc

Bases: object

Miscellaneous settings that don’t fit in any other namespace.

snap_to_end: bool = True

Controls alignment of timedelta-based splits.

Note

Active only when both schedule and after are TimedeltaTypes.

If True, ensure that the Future data of the final fold ends at the rightmost edge of the available data range. Otherwise, alignment is determined by pandas.date_range().

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

A callable (start, mid, end) -> bool applied to all generated folds.

round_limits: bool = False

If True, use rounding instead of floor and ceil for limits expansion.

Normally, limits such as

>>> ('2019-04-10 23:55:00', '2019-05-12 01:00:00')

will never be rounded to

>>> ('2019-04-11', '2019-05-12')

since that wouldn’t be an expansion of the limits. Setting round_limits=True will allow the expand_limits() function round these bounds inward to the nearest dates.