add Library functions for finding instances and extracting hierarchy

added child_graph, parent_graph, child_order, find_refs_local and find_refs_global

masque/library.py

@@ -22,12 +22,13 @@ import copy
 from pprint import pformat
 from collections import defaultdict
 from abc import ABCMeta, abstractmethod
+from graphlib import TopologicalSorter
 
 import numpy
 from numpy.typing import ArrayLike, NDArray
 
 from .error import LibraryError, PatternError
-from .utils import rotation_matrix_2d, layer_t
+from .utils import layer_t, apply_transforms
 from .shapes import Shape, Polygon
 from .label import Label
 from .abstract import Abstract
@@ -474,16 +475,13 @@ class ILibraryView(Mapping[str, 'Pattern'], metaclass=ABCMeta):
                 raise LibraryError(f'.dfs() called on pattern with circular reference to "{target}"')
 
             for ref in pattern.refs[target]:
+                ref_transforms: list[bool] | NDArray[numpy.float64]
                 if transform is not False:
-                    sign = numpy.ones(2)
-                    if transform[3]:
-                        sign[1] = -1
-                    xy = numpy.dot(rotation_matrix_2d(transform[2]), ref.offset * sign)
-                    ref_transform = transform + (xy[0], xy[1], ref.rotation, ref.mirrored)
-                    ref_transform[3] %= 2
+                    ref_transforms = apply_transforms(transform, ref.as_transforms())
                 else:
-                    ref_transform = False
+                    ref_transforms = [False]
 
+                for ref_transform in ref_transforms:
                     self.dfs(
                         pattern=self[target],
                         visit_before=visit_before,
@@ -510,6 +508,143 @@ class ILibraryView(Mapping[str, 'Pattern'], metaclass=ABCMeta):
 
         return self
 
+    def child_graph(self) -> dict[str, set[str | None]]:
+        """
+          Return a mapping from pattern name to a set of all child patterns
+        (patterns it references).
+
+        Returns:
+            Mapping from pattern name to a set of all pattern names it references.
+        """
+        graph = {name: set(pat.refs.keys()) for name, pat in self.items()}
+        return graph
+
+    def parent_graph(self) -> dict[str, set[str]]:
+        """
+          Return a mapping from pattern name to a set of all parent patterns
+        (patterns which reference it).
+
+        Returns:
+            Mapping from pattern name to a set of all patterns which reference it.
+        """
+        igraph: dict[str, set[str]] = {name: set() for name in self}
+        for name, pat in self.items():
+            for child, reflist in pat.refs.items():
+                if reflist and child is not None:
+                    igraph[child].add(name)
+        return igraph
+
+    def child_order(self) -> list[str]:
+        """
+          Return a topologically sorted list of all contained pattern names.
+        Child (referenced) patterns will appear before their parents.
+
+        Return:
+            Topologically sorted list of pattern names.
+        """
+        return list(TopologicalSorter(self.child_graph()).static_order())
+
+    def find_refs_local(
+            self,
+            name: str,
+            parent_graph: dict[str, set[str]] | None = None,
+            ) -> dict[str, list[NDArray[numpy.float64]]]:
+        """
+          Find the location and orientation of all refs pointing to `name`.
+        Refs with a `repetition` are resolved into multiple instances (locations).
+
+        Args:
+            name: Name of the referenced pattern.
+            parent_graph: Mapping from pattern name to the set of patterns which
+                reference it. Default (`None`) calls `self.parent_graph()`.
+                The provided graph may be for a superset of `self` (i.e. it may
+                contain additional patterns which are not present in self; they
+                will be ignored).
+
+        Returns:
+            Mapping of {parent_name: transform_list}, where transform_list
+                is an Nx4 ndarray with rows
+                `(x_offset, y_offset, rotation_ccw_rad, mirror_across_x)`.
+        """
+        instances = defaultdict(list)
+        if parent_graph is None:
+            parent_graph = self.parent_graph()
+        for parent in parent_graph[name]:
+            if parent not in self:          # parent_graph may be a for a superset of self
+                continue
+            for ref in self[parent].refs[name]:
+                instances[parent].append(ref.as_transforms())
+
+        return instances
+
+    def find_refs_global(
+            self,
+            name: str,
+            order: list[str] | None = None,
+            parent_graph: dict[str, set[str]] | None = None,
+            ) -> dict[tuple[str, ...], NDArray[numpy.float64]]:
+        """
+          Find the absolute (top-level) location and orientation of all refs (including
+        repetitions) pointing to `name`.
+
+        Args:
+            name: Name of the referenced pattern.
+            order: List of pattern names in which children are guaranteed
+                to appear before their parents (i.e. topologically sorted).
+                Default (`None`) calls `self.child_order()`.
+            parent_graph: Passed to `find_refs_local`.
+                Mapping from pattern name to the set of patterns which
+                reference it. Default (`None`) calls `self.parent_graph()`.
+                The provided graph may be for a superset of `self` (i.e. it may
+                contain additional patterns which are not present in self; they
+                will be ignored).
+
+        Returns:
+            Mapping of `{hierarchy: transform_list}`, where `hierarchy` is a tuple of the form
+                `(toplevel_pattern, lvl1_pattern, ..., name)` and `transform_list` is an Nx4 ndarray
+                with rows `(x_offset, y_offset, rotation_ccw_rad, mirror_across_x)`.
+        """
+        if name not in self:
+            return {}
+        if order is None:
+            order = self.child_order()
+        if parent_graph is None:
+            parent_graph = self.parent_graph()
+
+        self_keys = set(self.keys())
+
+        transforms: dict[str, list[tuple[
+                tuple[str, ...],
+                NDArray[numpy.float64]
+                ]]]
+        transforms = defaultdict(list)
+        for parent, vals in self.find_refs_local(name, parent_graph=parent_graph).items():
+            transforms[parent] = [((name,), numpy.concatenate(vals))]
+
+        for next_name in order:
+            if next_name not in transforms:
+                continue
+            if not parent_graph[next_name] & self_keys:
+                continue
+
+            outers = self.find_refs_local(next_name, parent_graph=parent_graph)
+            inners = transforms.pop(next_name)
+            for parent, outer in outers.items():
+                for path, inner in inners:
+                    combined = apply_transforms(numpy.concatenate(outer), inner)
+                    transforms[parent].append((
+                        (next_name,) + path,
+                        combined,
+                        ))
+        result = {}
+        for parent, targets in transforms.items():
+            for path, instances in targets:
+                full_path = (parent,) + path
+                assert full_path not in result
+                result[full_path] = instances
+        return result
+
+
 
 class ILibrary(ILibraryView, MutableMapping[str, 'Pattern'], metaclass=ABCMeta):
     """

masque/ref.py

@@ -183,6 +183,16 @@ class Ref(
             self.rotation += pi
         return self
 
+    def as_transforms(self) -> NDArray[numpy.float64]:
+        xys = self.offset[None, :]
+        if self.repetition is not None:
+            xys = xys + self.repetition.displacements
+        transforms = numpy.empty((xys.shape[0], 4))
+        transforms[:, :2] = xys
+        transforms[:, 2] = self.rotation
+        transforms[:, 3] = self.mirrored
+        return transforms
+
     def get_bounds_single(
             self,
             pattern: 'Pattern',

masque/utils/__init__.py

@@ -24,6 +24,7 @@ from .transform import (
     rotation_matrix_2d as rotation_matrix_2d,
     normalize_mirror as normalize_mirror,
     rotate_offsets_around as rotate_offsets_around,
+    apply_transforms as apply_transforms,
     )
 from .comparisons import (
     annotation2key as annotation2key,

masque/utils/transform.py

@@ -5,7 +5,7 @@ from collections.abc import Sequence
 from functools import lru_cache
 
 import numpy
-from numpy.typing import NDArray
+from numpy.typing import NDArray, ArrayLike
 from numpy import pi
 
 
@@ -57,8 +57,62 @@ def rotate_offsets_around(
         ) -> NDArray[numpy.float64]:
     """
     Rotates offsets around a pivot point.
+
+    Args:
+        offsets: Nx2 array, rows are (x, y) offsets
+        pivot: (x, y) location to rotate around
+        angle: rotation angle in radians
+
+    Returns:
+        Nx2 ndarray of (x, y) position after the rotation is applied.
     """
     offsets -= pivot
     offsets[:] = (rotation_matrix_2d(angle) @ offsets.T).T
     offsets += pivot
     return offsets
+
+
+def apply_transforms(
+        outer: ArrayLike,
+        inner: ArrayLike,
+        tensor: bool = False,
+        ) -> NDArray[numpy.float64]:
+    """
+    Apply a set of transforms (`outer`) to a second set (`inner`).
+    This is used to find the "absolute" transform for nested `Ref`s.
+
+    The two transforms should be of shape Ox4 and Ix4.
+    Rows should be of the form `(x_offset, y_offset, rotation_ccw_rad, mirror_across_x)`.
+    The output will be of the form (O*I)x4 (if `tensor=False`) or OxIx4 (`tensor=True`).
+
+    Args:
+        outer: Transforms for the container refs. Shape Ox4.
+        inner: Transforms for the contained refs. Shape Ix4.
+        tensor: If `True`, an OxIx4 array is returned, with `result[oo, ii, :]` corresponding
+            to the `oo`th `outer` transform applied to the `ii`th inner transform.
+            If `False` (default), this is concatenated into `(O*I)x4` to allow simple
+            chaining into additional `apply_transforms()` calls.
+
+    Returns:
+        OxIx4 or (O*I)x4 array. Final dimension is
+            `(total_x, total_y, total_rotation_ccw_rad, net_mirrored_x)`.
+    """
+    outer = numpy.atleast_2d(outer).astype(float, copy=False)
+    inner = numpy.atleast_2d(inner).astype(float, copy=False)
+
+    # If mirrored, flip y's
+    xy_mir = numpy.tile(inner[:, :2], (outer.shape[0], 1, 1))   # dims are outer, inner, xyrm
+    xy_mir[outer[:, 3].astype(bool), :, 1] *= -1
+
+    rot_mats = [rotation_matrix_2d(angle) for angle in outer[:, 2]]
+    xy = numpy.einsum('ort,oit->oir', rot_mats, xy_mir)
+
+    tot = numpy.empty((outer.shape[0], inner.shape[0], 4))
+    tot[:, :, :2] = outer[:, None, :2] + xy
+    tot[:, :, 2:] = outer[:, None, 2:] + inner[None, :, 2:]     # sum rotations and mirrored
+    tot[:, :, 2] %= 2 * pi        # clamp rot
+    tot[:, :, 3] %= 2             # clamp mirrored
+
+    if tensor:
+        return tot
+    return numpy.concatenate(tot)