habit.combiners: compose feature trees

Built-in feature block combiners and their registry.

Combiners are the internal nodes of a feature composition tree: they merge the column blocks their sibling nodes produced and never touch images, subjects, or the filesystem. The same combiner implementations serve every granularity (voxel, supervoxel, habitat), because the tree wrappers hand them plain DataFrame blocks with positionally aligned rows.

User guide: Habitat domain API. Component names: Habitat Spec component catalog.

Concat, ratio, kinetic, and expression trees over extractors (concat(raw("T1"), voxel_radiomics("T2"))).

Classes

Combiner	Merge the feature blocks produced by sibling nodes of a feature tree.
ConcatCombiner	Merge sibling blocks by placing their columns side by side.
WeightedConcatCombiner	Concatenate sibling blocks after scaling each by a child-specific weight.
AverageCombiner	Average sibling blocks element-wise, optionally with child weights.
RatioCombiner	Element-wise ratio of two sibling blocks: first / (second + eps).
DifferenceCombiner	Element-wise difference of two sibling blocks: first - second.
ExpressionCombiner	Features defined by restricted arithmetic over sibling block columns.
KineticCombiner	Per-unit enhancement slopes across a dynamic contrast series.
CombinerRegistry	Name-to-implementation registry for feature block combiners.

Functions

block_sources	Resolve the source label of each child block, in child order.
check_blocks	Require at least one block and a shared row count across siblings.
concat_blocks	Concatenate child blocks column-wise with a duplicate-column guard.