File size: 5,654 Bytes
3b80460
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
"""Exact-formula positional line-overlap certificates, evaluated in float64.

For two segments, Q is the measure of unoriented lines meeting both, with
line measure dtheta*dt for theta in [0,pi).  The endpoint formula is exact
mathematically; its implementation and all returned numbers use floating
point arithmetic, not outward-rounded interval arithmetic.

The certificates are NECESSARY conditions for an opaque barrier of the
unit-radius disk.  Passing them does not establish opacity.  Positive-length
collinear overlaps must be merged first, to avoid counting length twice.
"""
import math
import numpy as np


def _cross(u,v):
    return float(u[0]*v[1]-u[1]*v[0])


def _point(x):
    p=np.asarray(x,dtype=float)
    if p.shape!=(2,) or not np.isfinite(p).all():
        raise ValueError("Endpoints must be finite two-dimensional points")
    return p


def _distance(p,q):
    return math.hypot(float(p[0]-q[0]),float(p[1]-q[1]))


def _four_point(a,b,c,d):
    return abs(math.fsum([_distance(a,c),_distance(b,d),
                          -_distance(a,d),-_distance(b,c)]))


def pair_overlap(first,second,relative_tolerance=1e-11):
    """Evaluate Q by a finite sum of absolute four-endpoint distance terms.

    Split both segments at the intersection of their supporting lines when
    that intersection is interior.  On each resulting parameter rectangle,
    the mixed derivative of distance has constant sign.

    Collinear positive-length overlaps are rejected.  Numerically ambiguous
    near-collinear positive overlaps within relative_tolerance are rejected
    as well.  A shared endpoint and isolated zero-length pieces are allowed.
    """
    a,b=map(_point,first)
    c,d=map(_point,second)
    e,f=b-a,d-c
    le,lf=_distance(a,b),_distance(c,d)
    if le==0 or lf==0:
        return 0.0
    scale=max(le,lf,_distance(a,c),_distance(a,d),
              _distance(b,c),_distance(b,d))
    tol=relative_tolerance*scale
    determinant=_cross(e,f)
    near_parallel=abs(determinant)<=relative_tolerance*le*lf
    near_collinear=(near_parallel and
                    abs(_cross(c-a,e))/le<=tol and
                    abs(_cross(d-a,e))/le<=tol)
    if near_collinear:
        unit=e/le
        pc,pd=float(np.dot(c-a,unit)),float(np.dot(d-a,unit))
        overlap=min(le,max(pc,pd))-max(0.,min(pc,pd))
        if overlap>tol:
            raise ValueError("Collinear or numerically near-collinear positive-length overlap: merge these pieces first")
        return 0.0
    if determinant==0:
        return _four_point(a,b,c,d)
    t=_cross(c-a,f)/determinant
    u=_cross(c-a,e)/determinant
    # The average reduces roundoff when an intersection is interior to both.
    p=.5*((a+t*e)+(c+u*f))
    first_pieces=[(a,b)]
    second_pieces=[(c,d)]
    if relative_tolerance<t<1-relative_tolerance:
        first_pieces=[(a,p),(p,b)]
    if relative_tolerance<u<1-relative_tolerance:
        second_pieces=[(c,p),(p,d)]
    return math.fsum(_four_point(x,y,z,w)
                     for x,y in first_pieces for z,w in second_pieces)


def positional_certificate(segments,certified_max_hits=None):
    """Return pair energy and maximum-spanning-forest necessary bounds.

    If the input is an actual opaque barrier with total length L, then
      L >= pi + .5*maximum_spanning_forest_value.
    If a separate proof establishes N(line)<=M for almost every line,
      L >= pi + pair_energy/M.
    'certified_max_hits' records that EXTERNAL assumption.  This function
    never certifies a line-intersection maximum by sampling or otherwise.
    The universal automatic maximum is the number of nondegenerate pieces.
    """
    pieces=[tuple(map(_point,s)) for s in segments]
    n=len(pieces)
    if n==0:
        raise ValueError("At least one segment is required")
    lengths=[_distance(a,b) for a,b in pieces]
    n_active=sum(x>0 for x in lengths)
    pairs=[]
    for i in range(n):
        for j in range(i+1,n):
            pairs.append((i,j,pair_overlap(pieces[i],pieces[j])))
    parent=list(range(n))
    def find(x):
        while parent[x]!=x:
            parent[x]=parent[parent[x]]
            x=parent[x]
        return x
    forest=[]
    for i,j,q in sorted(pairs,key=lambda item:item[2],reverse=True):
        ri,rj=find(i),find(j)
        if ri!=rj:
            parent[ri]=rj
            forest.append((i,j,q))
    energy=math.fsum(q for i,j,q in pairs)
    forest_value=math.fsum(q for i,j,q in forest)
    result={"segment_count":n,"nondegenerate_segment_count":n_active,
            "total_segment_length":math.fsum(lengths),
            "pair_overlaps":[{"i":i,"j":j,"Q":q} for i,j,q in pairs],
            "pair_energy":energy,
            "maximum_spanning_forest_edges":[{"i":i,"j":j,"Q":q} for i,j,q in forest],
            "maximum_spanning_forest_value":forest_value,
            "forest_necessary_lower_bound":math.pi+.5*forest_value,
            "automatic_hit_bound":n_active,
            "automatic_pair_necessary_lower_bound":math.pi+energy/n_active if n_active else None,
            "opacity_established":False,
            "arithmetic":"float64 evaluation of exact geometric formulas"}
    if certified_max_hits is not None:
        if not isinstance(certified_max_hits,int) or not 1<=certified_max_hits<=n_active:
            raise ValueError("Externally certified M must be an integer between 1 and the number of nondegenerate pieces")
        result.update(externally_assumed_max_hits=certified_max_hits,
                      conditional_pair_necessary_lower_bound=math.pi+energy/certified_max_hits,
                      max_hits_assumption_verified_by_this_code=False)
    return result