> ## Documentation Index > Fetch the complete documentation index at: https://systematica.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # Custom Splitter > systematica.utils.custom_splitter ## `CustomSplitter` ```python theme={null} CustomSplitter( wrapper: vectorbtpro.base.wrapping.ArrayWrapper, index: pandas.core.indexes.base.Index, splits_arr: numpy.ndarray, **kwargs, ) ``` Custom splitter. ### Ancestors * `vectorbtpro.generic.splitting.base.Splitter` * `vectorbtpro.generic.analyzable.Analyzable` * `vectorbtpro.base.wrapping.Wrapping` * `vectorbtpro.utils.config.Configured` * `vectorbtpro.utils.config.HasSettings` * `vectorbtpro.utils.caching.Cacheable` * `vectorbtpro.utils.checks.Comparable` * `vectorbtpro.utils.pickling.Pickleable` * `vectorbtpro.utils.formatting.Prettified` * `vectorbtpro.utils.chaining.Chainable` * `vectorbtpro.base.wrapping.HasWrapper` * `vectorbtpro.base.indexing.ExtPandasIndexer` * `vectorbtpro.base.indexing.PandasIndexer` * `vectorbtpro.base.indexing.IndexingBase` * `vectorbtpro.utils.params.ItemParamable` * `vectorbtpro.utils.params.Itemable` * `vectorbtpro.utils.params.Paramable` * `vectorbtpro.base.indexes.IndexApplier` * `vectorbtpro.utils.attr_.AttrResolverMixin` * `vectorbtpro.generic.stats_builder.StatsBuilderMixin` * `vectorbtpro.generic.plots_builder.PlotsBuilderMixin` * `vectorbtpro.utils.base.Base` ### Static methods #### `from_custom_rolling` ```python theme={null} from_custom_rolling( index: pandas.core.indexes.base.Index, length: int, offset: int = 0, offset_anchor: str = 'prev_end', offset_anchor_set: int = 0, offset_space: str = 'prev', backwards: str | bool = False, split: int = None, split_range_kwargs: Dict[str, Any] = None, range_bounds_kwargs: Dict[str, Any] = None, template_context: Dict[str, Any] = None, freq: str = None, **kwargs, ) ‑> ~SplitterT ``` Create a custom `Splitter` instance from a rolling range of fixed length. Superset `Splitter.from_rolling` by adding the first (if backwards is sorted) or last remaining split. Creates a rolling sequence of slices based on the provided index. The first slice uses a length defined by `length`, and each subsequent slice begins after an offset from the previous slice's right boundary. Uses `Splitter.from_splits` to generate an array of splits and corresponding labels, and then construct the `Splitter` instance. See `Splitter.from_rolling` for further informations. **Parameters**: | Name | Type | Default | Description | | --------------------- | ------------------ | ---------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `index` | `tp.IndexLike` | `--` | The index to use for the splits. Must be a datetime-like index. | | `length` | `int` | `--` | The length of each rolling slice. Can be an `int`, `Float`, or `timedelta`. | | `offset` | `int` | `0` | The offset to apply to the start of each slice after the first one. Can be an integer, float, or timedelta. | | `offset_anchor` | `str` | `prev_end` | The anchor point for the offset. Can be `prev_end` or `start`. | | `offset_anchor_set` | `int` | `0` | The index of the anchor point to use for the offset. If `None`, uses the last anchor point. | | `offset_space` | `str` | `prev` | The space in which the offset is applied. Can be `prev` or `all`. | | `backwards` | `bool` | `False` | If `True`, the rolling slices are created in reverse order. If 'sorted', the first slice is sorted in reverse order. If `False`, the rolling slices are created in forward order. | | `split` | `tp.SplitLike` | `None` | If provided, the rolling slices will be split further according to this parameter. Can be a single split or a list of splits. | | `split_range_kwargs` | `tp.KwargsLike` | `None` | Additional keyword arguments to pass to the `split_range` method. | | `range_bounds_kwargs` | `tp.KwargsLike` | `None` | Additional keyword arguments to pass to the `get_range_bounds` method. | | `template_context` | `tp.KwargsLike` | `None` | Context for template rendering, if applicable. | | `freq` | `tp.FrequencyLike` | `None` | Frequency of the index. If `None`, it will be inferred from the index. | | `kwargs` | `tp.Kwargs` | `--` | Additional keyword arguments to pass to the `Splitter.from_splits` method. | **Raises**: | Type | Description | | ------------ | ------------------------------------------------------- | | `ValueError` | If `backwards` is a string that is not `sorted`. | | `ValueError` | If the `length` is negative when `backwards` is `True`. | **Returns**: | Type | Description | | ----------- | ------------------------------------------------------------------------- | | `SplitterT` | An instance of `Splitter` containing the rolling slices and their labels. | #### `from_custom_expanding` ```python theme={null} from_custom_expanding( index: pandas.core.indexes.base.Index, min_length: int, offset: int, split: int = None, split_range_kwargs: Dict[str, Any] = None, range_bounds_kwargs: Dict[str, Any] = None, template_context: Dict[str, Any] = None, freq: str = None, **kwargs, ) ‑> ~SplitterT ``` Create a `Splitter` instance from an expanding range. Creates an expanding sequence of slices based on the provided index. The first slice uses a minimum length defined by min\_length, and each subsequent slice begins after an offset from the previous slice's right boundary. Uses `Splitter.from_expanding` by adding the last remaining split. See `Splitter.from_expanding` for further informations. **Parameters**: | Name | Type | Default | Description | | --------------------- | ------------------ | ------- | ------------------------------------------------------------------------------------------------------------------------------- | | `index` | `tp.IndexLike` | `--` | The index to use for the splits. Must be a datetime-like index. | | `min_length` | `int` | `--` | The minimum length of each expanding slice. Can be an `int`, `float`, or `timedelta`. | | `offset` | `int` | `--` | The offset to apply to the start of each slice after the first one. Can be an `int`, `Float`, or `timedelta`. | | `split` | `tp.SplitLike` | `None` | If provided, the expanding slices will be split further according to this parameter. Can be a single split or a list of splits. | | `split_range_kwargs` | `tp.KwargsLike` | `None` | Additional keyword arguments to pass to the `split_range` method. | | `range_bounds_kwargs` | `tp.KwargsLike` | `None` | Additional keyword arguments to pass to the `get_range_bounds` method. | | `template_context` | `tp.KwargsLike` | `None` | Context for template rendering, if applicable. | | `freq` | `tp.FrequencyLike` | `None` | Frequency of the index. If `None`, it will be inferred from the index. | | `kwargs` | `tp.Kwargs` | `--` | Additional keyword arguments to pass to the `Splitter.from_splits` method. | **Raises**: | Type | Description | | ------------ | ----------------------------------------- | | `ValueError` | If the `min_length` is negative. | | `ValueError` | If the `offset` is negative. | | `ValueError` | If the `split` is not a valid split type. | **Returns**: | Type | Description | | ----------- | --------------------------------------------------------------------------- | | `SplitterT` | An instance of `Splitter` containing the expanding slices and their labels. |