snappy 3.3__cp310-cp310-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl

snappy/geometric_structure/geodesic/geodesic_start_point_info.py

@@ -0,0 +1,435 @@
+from . import constants
+from . import exceptions
+
+from .line import R13LineWithMatrix
+from .fixed_points import r13_fixed_line_of_psl2c_matrix
+from .multiplicity import compute_and_verify_multiplicity
+from .graph_trace_helper import find_lifted_tetrahedra_containing_point
+
+from .. import word_to_psl2c_matrix
+
+from ...tiling.lifted_tetrahedron import LiftedTetrahedron
+from ...hyperboloid import r13_dot, o13_inverse # type: ignore
+from ...hyperboloid.distances import distance_r13_points
+from ...hyperboloid.line import R13Line
+from ...snap.t3mlite import simplex # type: ignore
+from ...snap.t3mlite import Tetrahedron, Vertex, Mcomplex # type: ignore
+from ...exceptions import InsufficientPrecisionError # type: ignore
+from ...matrix import make_identity_matrix # type: ignore
+
+from typing import Tuple, Sequence, Optional, Any
+
+__all__ = ['compute_geodsic_info', 'GeodesicStartPointInfo', 'sample_line']
+
+def sample_line(line : R13Line):
+    """
+    Pick a point on a line in the hyperboloid model.
+    Returns an unnormalised time-like vector computed
+    as the weighted average of the two light-like
+    endpoints of the line.
+
+    The ratio of the weights is some fixed number picked at random so
+    that we avoid picking a point that lies, e.g., on an edge of the
+    triangulation (which happens for some geodesics in some
+    triangulated hyperbolic manifolds when picking equal weights for
+    the fixed points computed by r13_fixed_points_of_psl2c_matrix).
+
+    Note that we want to avoid picking a point that is far away from
+    the fundamental polyhedron. By the choices we made, this is not
+    the case: the fundamental polyhedron is contains the origin in the
+    hyperboloid model. r13_fixed_points_of_psl2c_matrix returns
+    light-like vectors of the form (1, ...) so the average corresponds
+    to taking the mid-point in the Klein model and is thus the point
+    on the line closest to the origin. Furthermore, the bias is close
+    enough to 1 (the log is ~0.22, so we move the point by ~0.11
+    units in hyperbolic space).
+    """
+
+    RF = line.points[0][0].parent()
+    bias = RF(constants.start_point_bias)
+
+    return line.points[0] + bias * line.points[1]
+
+# @dataclass
+
+
+class GeodesicStartPointInfo:
+    """
+    Information needed to trace a closed geodesic through a triangulation
+    given as snappy.snap.t3mlite.Mcomplex with geometric structure added
+    by add_r13_geometry.
+
+    The basic information consists of a line in the hyperboloid model
+    that is a lift of the closed geodesic and a start and end point on or
+    close to that line such that the line segment from the start to the
+    end point maps to a simple closed curve in the manifold isotopic to
+    the closed geodesic.
+
+    If a client has instantiated this class with the basic information,
+    it can call find_tet_or_core_curve. The method find_tet_or_core_curve
+    will either:
+
+    1. Detect that the closed geodesic is actually a core curve of a
+       filled cusp and set core_curve_cusp and core_curve_multiplicity
+       accordingly. This means that instead tracing the geodesic
+       through the triangulation, the client has to unfill the
+       corresponding cusp instead.
+    2. Apply a Decktransformation to the line and points such that
+       start point is either in the interior of a tetrahedron (in the
+       fundamental domain) or in the union of two (lifted) tetrahedra
+       (in the universal cover which is the hyperboloid model). That
+       is, if the start point is on a face of the triangulation, it
+       will return the two adjacent tetrahedra. If the start point is
+       in the interior of a tetrahedron, the client can attempt to
+       trace the geodesic through the triangulation. The client can
+       use the given (lifted) tetrahedra to develop a tube about the
+       geodesic to compute its injectivity radius.
+
+    There is an additional field index that can be used by clients for
+    book-keeping purposes, for example, to store the index of the cusp
+    obtained by drilling this geodesic.
+
+    The start and end point are unnormalised time-like vectors. Note
+    that normalisation is not required for many applications (such as
+    computing the intersection of the line segment from the start to
+    the end point with a plane) and will enlarge the intervals when
+    performing verified computations.
+    """
+
+    def __init__(self,
+                 # The triangulation
+                 mcomplex : Mcomplex,
+
+                 # Word were are drilling
+                 word : str,
+                 
+                 # Trace of corresponding PSL(2,C)-matrix.
+                 trace : Any,
+
+                 # A point on or near the line corresponding to the closed geodesic.
+
+                 # It is a light-like R13-vector. Using Any as type annotation because
+                 # this might be a SimpleVector or SageMath type.
+                 unnormalised_start_point : Any,
+
+                 # Optional: image of the point under the matrix corresponding to
+                 # the closed geodesic and fixing the given line (set-wise).
+                 unnormalised_end_point : Optional[Any] = None,
+
+                 # Line corresponding to the closed geodesic with matrix.
+                 # Must be given if we want to detect whether this geodesic
+                 # is a core curve.
+                 line : Optional[R13LineWithMatrix] = None,
+
+                 # Output of find_tet_or_core_curve: if not None, the start point
+                 # is guaranteed to be in this tetrahedron (as part of the
+                 # fundamental domain).
+                 tet : Optional[Tetrahedron] = None,
+
+                 # Output of find_tet_or_core_curve: if non-empty, the start point
+                 # is guaranteed to be in the union of the lifted tetrahedra.
+                 # A lifted tetrahedron is encoded as a pair of a tetrahedron
+                 # (in the fundamental domain) and an O(1,3)-matrix and is the
+                 # image of this tetrahedron under the matrix.
+                 # domain and a O(1,3)-matrix that needs to be applied
+
+                 lifted_tetrahedra : Sequence[LiftedTetrahedron] = (),
+
+                 # Output of find_tet_or_core_curve: if not None, the geodesic
+                 # corresponds to the core curve for this cusp.
+                 core_curve_cusp : Optional[Vertex] = None,
+
+                 # Output of find_tet_or_core_corve: sign (+1/-1) indicating whether
+                 # the given geodesic and the core curve run parallel or
+                 # anti-parallel.
+                 core_curve_multiplicity : Optional[int] = None,
+
+                 # Field filled by client to indicate which index the cusp resulting
+                 # from drilling this geodesic is supposed to have.
+                 index : Optional[int] = None):
+
+        self.mcomplex = mcomplex
+        self.word = word
+        self.trace = trace
+        self.unnormalised_start_point = unnormalised_start_point
+        self.unnormalised_end_point = unnormalised_end_point
+        self.line = line
+        self.tet = tet
+        self.lifted_tetrahedra = lifted_tetrahedra
+        self.core_curve_cusp = core_curve_cusp
+        self.core_curve_multiplicity = core_curve_multiplicity
+        self.index = index
+
+    def find_tet_or_core_curve(self) -> None:
+        """
+        Apply Deck-transformations to the start and end point and hyperbolic
+        line until we either detected that the given geodesic corresponds to
+        a core curve (only if line is not None) or we have captured the start
+        point in one or two tetrahedra (in case the start close is on or very
+        close to a face).
+        This method also computes the distance of the geodesic to the core
+        curves (only if line is not None) and raises an exception if we could
+        not ensure that this distance is positive.
+        """
+
+        self._graph_trace()
+
+    def _graph_trace(self):
+        self.tet = None
+        self.lifted_tetrahedra = ()
+        self.core_curve_cusp = None
+        self.core_curve_multiplicity = None
+
+        # Walks from tetrahedron to tetrahedron (transforming the start point
+        # and other data) trying to get the start_point closer and closer to
+        # one of the tetrahedra.
+        # Stops when the start_point appears to be in a tetrahedron or on the
+        # face of a tetrahedron or when (the lift of) the geodesic was really
+        # close to (a lift of) a core curve.
+
+        if self.mcomplex.verified:
+            epsilon = 0
+            def key(face_and_signed_distance):
+                return face_and_signed_distance[1].center()
+        else:
+            epsilon = _compute_epsilon(self.mcomplex.RF)
+            def key(face_and_signed_distance):
+                return face_and_signed_distance[1]
+
+        # Face through which tetrahedron was entered - to avoid we are
+        # going back through the face we just came from. Initialized to
+        # simplex.T for the first iteration.
+        tet = self.mcomplex.baseTet
+        entry_cell = simplex.T
+
+        for i in range(constants.graph_trace_max_steps):
+            # See whether the geodesic is close to a core curve.
+            # v is the vertex of the simplex for that core curve.
+            # Or None
+            v = self._find_cusp_if_core_curve(tet, entry_cell, epsilon)
+
+            if v is not None:
+                # Verify that the the geodesic is really the core curve and
+                # determine whether the geodesic and core curve or parallel
+                # or anti-parallel.
+                self.core_curve_multiplicity = self._multiplicity_of_core_curve(
+                    tet, v)
+                self.core_curve_cusp = tet.Class[v]
+                return
+
+            # Compute the signed distance of the start point to the
+            # planes supporting a face for each face of the tetrahedron.
+            #
+            # The maximum value tells us which face to cross next.
+            #
+            # Note that we only care about the maximum value, so we can used
+            # the unnormalised start point (that is a time-like vector that
+            # is not necessarily a unit vector). Since we compare the value
+            # for different faces though, we need to use the normalised
+            # plane equations.
+            #
+            faces_and_signed_distances = [
+                (face, r13_dot(self.unnormalised_start_point, tet.R13_planes[face]))
+                for face in simplex.TwoSubsimplices ]
+
+            # If we cannot confirm that the start point is outside the current
+            # tetrahedron, stop.
+            #
+            # Note the subtle difference here between using
+            # signed_distance >  epsilon to determine whether to stop
+            # signed_distance < -epsilon to determine whether the start point
+            # is inside.
+            #
+            if not any( signed_distance > epsilon
+                        for face, signed_distance
+                        in faces_and_signed_distances ):
+
+                self.lifted_tetrahedra = find_lifted_tetrahedra_containing_point(
+                    LiftedTetrahedron(
+                        tet, make_identity_matrix(ring=self.mcomplex.RF, n=4)),
+                    faces_and_signed_distances,
+                    self.unnormalised_start_point,
+                    epsilon)
+
+                if len(self.lifted_tetrahedra) == 1:
+                    # We verified that there is a unique tetrahedron containing
+                    # the start point.
+                    # Signal to the client that we can start tracing the geodesic
+                    # throguh the triangulation from this tetrahedron:
+                    self.tet = self.lifted_tetrahedra[0].tet
+
+                return
+
+            # Find face for which the signed distance is largest.
+            #
+            # For intervals, we compare by using the center of each interval.
+            #
+            # This is fine in that we use the signed distances here to
+            # heuristically guide us the right tetrahedron containing the
+            # start point - and have code to explicitly check at the end that
+            # the start point is really contained in the resulting tetrahedron
+            # or pair of neighboring tetrahedra.
+            # Also note that if two largest signed distance are really close
+            # to each other, then making either choice will probably eventually
+            # get us to the tetrahedron containing the start point.
+            #
+            face, worst_distance = max(
+                [ face_and_signed_distance
+                  for face_and_signed_distance in faces_and_signed_distances
+                  if face_and_signed_distance[0] != entry_cell],
+                key=key)
+
+            self._transform(tet.O13_matrices[face])
+            entry_cell = tet.Gluing[face].image(face)
+            tet = tet.Neighbor[face]
+
+        raise exceptions.UnfinishedGraphTraceGeodesicError(
+            constants.graph_trace_max_steps)
+
+    def _transform(self, m): # m is Decktransformation O(1,3)-matrix
+        """
+        Transform the data by matrix.
+        """
+
+        self.unnormalised_start_point = m * self.unnormalised_start_point
+        if self.unnormalised_end_point:
+            self.unnormalised_end_point = m * self.unnormalised_end_point
+        if self.line:
+            self.line = self.line.transformed(m)
+
+    def _find_cusp_if_core_curve(
+            self,
+            tet : Tetrahedron,
+            entry_cell : int,
+            epsilon) -> Optional[int]:
+        """
+        Check that the lift of the geodesic is close to the lifts of the core
+        curves at the vertices of the tetrahedron adjacent to entry_cell
+        where entry_cell is either in simplex.TwoSubsimplices or simplex.T.
+
+        If close, returns the vertex of the tetrahedron (in
+        simplex.ZeroSubsimplices), else None.
+        """
+
+        # Bail if we do not know the line that is the lift of the geodesic.
+        #
+        # This happens when perturbing the start and end point:
+        # in a first pass, it is determined which geodesics are curve
+        # curves. The corresponding cusps are unfilled and the geodesics
+        # dropped. Thus, future passes no longer need to check which
+        # geodesics are core curves - and the code perturbing
+        # the start and end point is dropping the line.
+        if not self.line:
+            return None
+
+        # For each vertex of the entry_cell
+        for v in simplex.ZeroSubsimplices:
+            if not simplex.is_subset(v, entry_cell):
+                continue
+            # Determine whether that vertex of the tetrahedron
+            # corresponds to a filled cusp and get (the corresponding
+            # lift of) the core curve.
+            core_curve = tet.core_curves.get(v)
+            if not core_curve:
+                continue
+            # Compute the inner products between the endpoints
+            # of the lifted core curve and geodesic. These endpoints
+            # are light-like. Thus, if they are co-linear (corresponding
+            # to the same point), the inner product will be zero.
+            p = [[ -r13_dot(pt0, pt1)
+                   for pt0 in self.line.r13_line.points ]
+                 for pt1 in tet.core_curves[v].r13_line.points ]
+
+            # We do not know which end of the line corresponding
+            # to the geodesic corresponds to which end of the line
+            # corresponding to the core curve. So check both cases.
+            #
+            # Note that there are two reasons we do not know this:
+            # we do not know whether the core curve and geodesic are
+            # parallel or anti-parallel. And
+            # r13_fixed_points_of_psl2c_matrix makes no guarantee
+            # on whether the attracting or repelling fixed point is
+            # returned first.
+            if not (p[0][0] > epsilon or p[1][1] > epsilon):
+                return v
+            if not (p[0][1] > epsilon or p[1][0] > epsilon):
+                return v
+
+        return None
+
+    def _multiplicity_of_core_curve(self,
+                                    tet : Tetrahedron,
+                                    vertex : int) -> int:
+        """
+        Verify that geodesic is indeed a multiple of the core curve (including
+        sign).
+        """
+
+        if self.line is None:
+            raise Exception(
+                "There is a bug in the code: it is trying to verify that "
+                "geodesic is a core curve without being given a line.")
+
+        try:
+            return compute_and_verify_multiplicity(
+                tet.core_curves[vertex],
+                self.line,
+                self.mcomplex)
+        except InsufficientPrecisionError:
+            raise InsufficientPrecisionError(
+                "Geodesic is very close to a core curve but could not verify "
+                "it is the core curve. Increasing the precision will probably "
+                "fix this.")
+
+def compute_geodesic_start_point_info(mcomplex : Mcomplex,
+                          word) -> GeodesicStartPointInfo:
+    """
+    Compute basic information about a geodesic given a word.
+
+    add_r13_geometry must have been called on the Mcomplex.
+    """
+
+    m = word_to_psl2c_matrix(mcomplex, word)
+    _verify_not_parabolic(m, mcomplex, word)
+    # Line fixed by matrix
+    line : R13LineWithMatrix = r13_fixed_line_of_psl2c_matrix(m)
+
+    # Pick a point on the line
+    start_point = sample_line(line.r13_line)
+
+    g = GeodesicStartPointInfo(
+        mcomplex=mcomplex,
+        word=word,
+        trace=m.trace(),
+        unnormalised_start_point=start_point,
+        unnormalised_end_point=line.o13_matrix * start_point,
+        line=line)
+
+    # Determines whether geodesic corresponds to a core curve.
+    # Applies Decktransformations so that start point lies within
+    # the interior of one tetrahedron in the fundamental domain or
+    # within the union of two tetrahedra neighboring in the hyperboloid
+    # model.
+    #
+    # See GeodesicStartPointInfo for details.
+    g.find_tet_or_core_curve()
+
+    return g
+
+def _verify_not_parabolic(m, mcomplex, word):
+    """
+    Raise exception when user gives a word corresponding to a parabolic
+    matrix.
+    """
+
+    if mcomplex.verified:
+        epsilon = 0
+    else:
+        epsilon = _compute_epsilon(mcomplex.RF)
+
+    tr = m.trace()
+    if not (abs(tr - 2) > epsilon and abs(tr + 2) > epsilon):
+        raise exceptions.WordAppearsToBeParabolic(word, tr)
+
+def _compute_epsilon(RF):
+    return RF(0.5) ** (RF.prec() // 2)

snappy/geometric_structure/geodesic/graph_trace_helper.py

@@ -0,0 +1,67 @@
+from ...tiling.lifted_tetrahedron import LiftedTetrahedron
+from ...snap.t3mlite import simplex # type: ignore
+from ...hyperboloid import r13_dot
+from ...exceptions import InsufficientPrecisionError # type: ignore
+
+def find_lifted_tetrahedra_containing_point(
+        lifted_tetrahedron : LiftedTetrahedron,
+        faces_and_signed_distances,
+        lifted_pt,
+        epsilon):
+
+    # Report faces for which we cannot confirm that the start point
+    # is to the inside of the plane supporting that face.
+    faces = [ face
+              for face, signed_distance
+              in faces_and_signed_distances
+              if not signed_distance < -epsilon ]
+
+    if len(faces) == 0:
+        # Signal to the client that we can start with this tetrahedron
+        # when developing a tube about the geodesic.
+        return [ lifted_tetrahedron ]
+
+    if len(faces) == 1:
+        # The start point is probably on the face of the tetrahedron,
+        # that is, we could verify it lies to the right side of the
+        # supporting planes for three faces but not one:
+        face, = faces
+
+        # Even though we cannot verify that the start point lies
+        # exactly on the face, we can verify that the start point
+        # lies in the interior of the union of two neighboring
+        # tetrahedra. That is, the union is a hexahedron and
+        # it suffices to check that the start point lies to the
+        # inside of the six faces of the tetrahedron.
+        #
+        # _graph_trace already checked the three faces of the given
+        # tetrahedron. But it is left to check this for the neighboring
+        # tetrahedron.
+
+        tet = lifted_tetrahedron.tet
+        m = lifted_tetrahedron.o13_matrix
+
+        # Find the other tetrahedron of the neighboring tetrahedra.
+        other_tet = tet.Neighbor[face]
+        other_pt = tet.O13_matrices[face] * lifted_pt
+        other_face = tet.Gluing[face].image(face)
+
+        for f in simplex.TwoSubsimplices:
+            if f == other_face:
+                continue
+            if not r13_dot(other_pt, other_tet.R13_planes[f]) < epsilon:
+                raise InsufficientPrecisionError(
+                    "Failed to find lift of geodesic and prove that "
+                    "it intersects tetrahedra of the fundamental domain. "
+                    "Increasing the precision will probably fix this "
+                    "problem.")
+
+        return [ lifted_tetrahedron,
+                 LiftedTetrahedron(
+                     other_tet,
+                     m * other_tet.O13_matrices[other_face]) ]
+
+    raise InsufficientPrecisionError(
+        "Start point chosen on geodesic too close to 1-skeleton of "
+        "triangulation to verify it is not on the 1-skeleton. "
+        "Increasing the precision will probably fix this problem.")

snappy/geometric_structure/geodesic/line.py

@@ -0,0 +1,30 @@
+from ...hyperboloid import o13_inverse # type: ignore
+from ...hyperboloid.line import R13Line
+
+__all__ = [ 'R13LineWithMatrix' ]
+
+class R13LineWithMatrix:
+    """
+    A line in the hyperboloid model together with a O(1,3)-matrix moving
+    the line forward by the given complex length (with real positive part)
+    (the matrix is fixing the line set-wise).
+    """
+    def __init__(self,
+                 r13_line : R13Line,
+                 o13_matrix,
+                 complex_length):
+        self.r13_line = r13_line
+        self.o13_matrix = o13_matrix
+        self.complex_length = complex_length
+
+    def transformed(self, m):
+        """
+        Returns image of line with matrix under given O13-matrix m.
+
+        That is, the matrix will be conjugated by m so that the new
+        matrix will fix the image of the line (set-wise).
+        """
+        return R13LineWithMatrix(
+            self.r13_line.transformed(m),
+            m * self.o13_matrix * o13_inverse(m),
+            self.complex_length)

snappy/geometric_structure/geodesic/multiplicity.py

@@ -0,0 +1,127 @@
+from .line import R13LineWithMatrix
+from ...hyperboloid.distances import distance_r13_points
+from ...hyperboloid import o13_inverse
+from ...exceptions import InsufficientPrecisionError
+
+def compute_and_verify_multiplicity(line : R13LineWithMatrix,
+                                    line_power : R13LineWithMatrix,
+                                    mcomplex):
+    """
+    Assume that line and line_power are created from matrices from the geometric
+    representation.
+
+    Verify that the matrix of line_power is a signed multiple of the matrix
+    of line and return the multiple.
+    """
+
+    # First use real length to compute multiple up to sign.
+    multiplicity = _compute_absolute_multiplicity(
+        line.complex_length.real(),
+        line_power.complex_length.real(),
+        mcomplex.verified)
+
+    sign = _determine_and_verify_sign(
+        multiplicity,
+        line.o13_matrix,
+        line_power.o13_matrix,
+        mcomplex)
+    
+    return sign * multiplicity
+
+_epsilon = 0.001
+_max_multiple = 500
+
+def _compute_absolute_multiplicity(
+        length,
+        length_multiple,
+        verified):
+
+    r = length_multiple / length
+
+    if verified:
+        is_int, r_int = r.is_int()
+        if not is_int:
+            raise InsufficientPrecisionError(
+                "When verifying that a geodesic is a multiple of another "
+                "geodesic, the interval for the multiplicity does not contain "
+                "a unique integer.")
+    else:
+        r_int = r.round()
+        if abs(r_int - r) > _epsilon:
+            raise InsufficientPrecisionError(
+                "When verifying that a geodesic is a multiple of another "
+                "geodesic, the floating point approximation for the "
+                "multiplicity is too far off an integer.")
+        r_int = int(r_int)
+
+    if not r_int > 0:
+        if verified:
+            raise RuntimeError(
+                "When verifying that a geodesic is a multiple of another "
+                "geodesic, we got zero for multiplicity. This is a bug.")
+        else:
+            raise InsufficientPrecisionError(
+                "When verifying that a geodesic is a multiple of another "
+                "geodesic, we got zero for multiplicity. Increasing the "
+                "precision might help.")
+    if not r_int < _max_multiple:
+        raise RuntimeError(
+            "When verifying that a geodesic is a multiple of another "
+            "geodesic, we got a multiple (%d) higher than what we support "
+            "(%d)" % (r_int, _max_multiple))
+
+    return r_int
+
+def _determine_and_verify_sign(multiplicity, m, m_power, mcomplex):
+    """
+    Given matrices m and m_power coming from the geometric representation,
+    verify that either m^multiplicity = m_power^sign for either sign = +1
+    or sign = -1. Return sign.
+    """
+
+    base_pt = mcomplex.R13_baseTetInCenter
+
+    # Compute image of base point under m^multiplicity.
+    pt = base_pt
+    for i in range(multiplicity):
+        pt = m * pt
+
+    # Check whether it is equal to image under m_power
+    if _are_images_of_basepoints_equal(
+            pt,             m_power  * base_pt, mcomplex):
+        return +1
+
+    # Check whether it is equal to image under m_power^-1
+    if _are_images_of_basepoints_equal(
+            pt, o13_inverse(m_power) * base_pt, mcomplex):
+        return -1
+
+    raise RuntimeError(
+        "Given geodesic is not a multiple of other given geodesic.")
+
+def _are_images_of_basepoints_equal(pt0, pt1, mcomplex):
+    """
+    Given two images of the base point under Deck transformations,
+    check whether they are the same (and thus correspond to the same
+    Deck transformation).
+    """
+
+    # We use that the the base point is the incenter of the base
+    # tetrahedron. Thus, we know that if the minimum distance for
+    # them to be distinct is the in-radius of the base tetrahedron.
+    #
+    # We add in a factor of 1/2 for safety.
+
+    d = distance_r13_points(pt0, pt1)
+    if d < mcomplex.baseTetInRadius / 2:
+        return True
+    if d > mcomplex.baseTetInRadius / 2:
+        return False
+
+    raise InsufficientPrecisionError(
+        "When determining whether a geodesic is a multiple of another "
+        "geodesic, we could not verify that the images of the base point "
+        "under the corresponding matrices are the same or not.\n"
+        "Distance between basepoints: %r\n"
+        "Cut-off: %r" % (d, mcomplex.baseTetInRadius / 2))
+