time-split 1.0.5__tar.gz → 1.1.1__tar.gz

{time_split-1.0.5 → time_split-1.1.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: time-split
-Version: 1.0.5
+Version: 1.1.1
 Summary: Time-based k-fold validation splits for heterogeneous data.
 License: MIT
 License-File: LICENSE.md

{time_split-1.0.5 → time_split-1.1.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "time-split"
-version = "1.0.5"
+version = "1.1.1"
 description = "Time-based k-fold validation splits for heterogeneous data."
 authors = [
     { name = "Richard Sundqvist", email = "richard.sundqvist@live.se" },

{time_split-1.0.5 → time_split-1.1.1}/src/time_split/__init__.py

@@ -9,4 +9,4 @@ __all__ = [
     "split",
 ]
 
-__version__ = "1.0.5"
+__version__ = "1.1.1"

{time_split-1.0.5 → time_split-1.1.1}/src/time_split/_backend/_schedule.py

@@ -1,12 +1,14 @@
-from typing import NamedTuple, get_args
+from typing import Literal, NamedTuple, get_args
 
 from pandas import DatetimeIndex, NaT, Timedelta, Timestamp, date_range
 
+from ..settings import misc as settings
 from ..types import DatetimeIterable, ExpandLimits, Schedule, TimedeltaTypes
 from ._limits import LimitsTuple
 from ._process_available import ProcessAvailableResult, process_available
 
 NO_LIMITS: LimitsTuple = NaT, NaT
+ScheduleType = Literal["cron", "explicit", "timedelta"]
 
 
 class MaterializedSchedule(NamedTuple):
@@ -14,6 +16,7 @@ class MaterializedSchedule(NamedTuple):
 
     schedule: DatetimeIndex
     available_metadata: ProcessAvailableResult
+    schedule_type: ScheduleType
 
 
 def materialize_schedule(
@@ -25,6 +28,7 @@ def materialize_schedule(
             return MaterializedSchedule(
                 DatetimeIndex(schedule),
                 available_metadata=ProcessAvailableResult(None, NO_LIMITS, NO_LIMITS),
+                schedule_type="explicit",
             )
         except TypeError as e:
             raise ValueError("Schedule must be explicit when not bounded by an available range.") from e
@@ -32,21 +36,26 @@ def materialize_schedule(
     available_metadata = process_available(available, expand_limits=expand_limits)
     min_dt, max_dt = available_metadata.expanded_limits
 
+    schedule_type: ScheduleType
     if isinstance(schedule, str) and _cron_like(schedule):
         schedule = _handle_cron(schedule, min_dt, max_dt)
+        schedule_type = "cron"
     elif isinstance(schedule, get_args(TimedeltaTypes)):
         schedule = _from_timedelta(schedule, available_metadata.expanded_limits)
+        schedule_type = "timedelta"
     else:
         if not isinstance(schedule, DatetimeIndex):
             schedule = DatetimeIndex(schedule)
 
         schedule = schedule.tz_localize(min_dt.tz) if schedule.tz is None else schedule.tz_convert(min_dt.tz)
         schedule = schedule[(min_dt <= schedule) & (schedule <= max_dt)]
-    return MaterializedSchedule(schedule, available_metadata=available_metadata)
+        schedule_type = "explicit"
+
+    return MaterializedSchedule(schedule, available_metadata, schedule_type)
 
 
 def _cron_like(schedule: str) -> bool:
-    # CroniterBadCronError: Exactly 5 or 6 columns has to be specified for iterator expression.
+    # CroniterBadCronError: Exactly 5 or 6 columns have to be specified for iterator expression.
     return len(schedule.split()) > 4 or schedule[0] == "@"  # noqa: PLR2004
 
 
@@ -54,7 +63,11 @@ def _from_timedelta(schedule: TimedeltaTypes, limits: LimitsTuple) -> DatetimeIn
     timedelta = Timedelta(schedule)
     if timedelta <= Timedelta(0):
         raise ValueError(f"unbounded {schedule=} must be greater than zero.")
-    return date_range(*limits, freq=timedelta, inclusive="both")
+
+    if settings.snap_to_end:
+        return date_range(limits[1], limits[0], freq=-timedelta, inclusive="both")[::-1]
+    else:
+        return date_range(limits[0], limits[1], freq=timedelta, inclusive="both")
 
 
 def _handle_cron(expr: str, min_dt: Timestamp, max_dt: Timestamp) -> DatetimeIndex:

{time_split-1.0.5 → time_split-1.1.1}/src/time_split/_backend/_splitter.py

@@ -53,20 +53,21 @@ class DatetimeIndexSplitter:
         if (
             settings.snap_to_end
             and self.after != "all"
-            and isinstance(self.schedule, types)
-            and isinstance(self.after, types)
+            and ms.schedule_type == "timedelta"
+            and isinstance(self.after, (types, int))
         ):
             ms = self._snap_to_end(ms)
 
         return ms
 
     def _snap_to_end(self, ms: MaterializedSchedule) -> MaterializedSchedule:
-        schedule_frequency = ms.schedule.freq
-        if schedule_frequency is None:
-            return ms
+        data_end = ms.available_metadata.expanded_limits[1]
+        schedule_end = ms.schedule[-1]
 
-        from_end = ms.available_metadata.expanded_limits[1] - (ms.schedule[-1] + Timedelta(self.after))
-        from_end = from_end.floor(schedule_frequency.base)  # TODO(settings): should this depend on misc.round_limits?
+        from_end = data_end - schedule_end
+        if not isinstance(self.after, int):
+            after = Timedelta(self.after)
+            from_end -= after
 
         return ms._replace(schedule=ms.schedule + from_end)
 

{time_split-1.0.5 → time_split-1.1.1}/src/time_split/app/_create_explorer_link.py

@@ -99,8 +99,8 @@ def _create_query(
         query = {"data": data, **kwargs}
     else:
         splitter = DatetimeIndexSplitter(**kwargs)
-        _, ms = splitter.get_plot_data(data)
-        query["data"] = _get_data_param(*ms.available_metadata.limits)
+        start, end = splitter._materialize_schedule(data).available_metadata.limits
+        query["data"] = _get_data_param(start, end)
 
         if skip_default:
             schedule = kwargs["schedule"]