Merge pull request #393 from Ensembl/bugfix/circular_chromosome_spanning

ENSCORESW-3132: Limited implementation of Intervals with start > end …

modules/Bio/EnsEMBL/Utils/Interval.pm

@@ -35,7 +35,7 @@ Bio::EnsEMBL::Utils::Interval
 =head1 SYNOPSIS
 
   # let's get an interval spanning 9e5 bp and associated it with some data
-  my $i2 = Bio::EnsEMBL::Utils::Interval->new(1e5, 1e6, { 'key1' => 'value1', 'key2' => 'value2' });
+  my $i1 = Bio::EnsEMBL::Utils::Interval->new(1e5, 1e6, { 'key1' => 'value1', 'key2' => 'value2' });
 
   # and another one which overlaps with the previous,
   # but with scalar associated data
@@ -50,12 +50,18 @@ Bio::EnsEMBL::Utils::Interval
     print "I1 and I2 do not overlap\n";
   }
 
+  # If an interval is defined with a start > end, then it is assumed
+  # to be spanning the origin on a circular chromosome
+  my $i3 = Bio::EnsEMBL::Utilities::Interval->new(1e5, 1e2);
+  warn "Interval spans the origin" if $i3->spans_origin;
+
   etc.
 
 =head1 DESCRIPTION
 
 A class representing an interval defined on a genomic region. Instances of this
-class can store arbitrarily defined data.
+class can store arbitrarily defined data. If created with start > end, then it
+is assumed that this interval is on a circular chromosome spanning the origin.
 
 =head1 METHODS
 
@@ -78,9 +84,12 @@ use Bio::EnsEMBL::Utils::Exception qw(throw);
   Arg [3]     : (optional) $data
                 The data associated with the interval, can be anything
   Example     : my $i = Bio::EnsEMBL::Utils::Interval(1e2, 2e2, { 'key' => 'value' });
+                my $i2 = Bio::EnsEMBL::Utilities::Interval(1e5, 1e2);
+                $i->spans_origin # returns 0
+                $i2->spans_origin # returns 1
   Description : Constructor. Creates a new instance
   Returntype  : Bio::EnsEMBL::Utils::Interval
-  Exceptions  : none
+  Exceptions  : Throws an exception if start and end are not defined.
   Caller      : general
 
 =cut
@@ -92,9 +101,17 @@ sub new {
   my ($start, $end, $data) = @_;
   throw 'Must specify interval boundaries [start, end]'
     unless defined $start and defined $end;
-  throw 'start must be <= end' if $start > $end;
+
+  my $spans_origin = 0;
+  if ($start > $end) {
+      $spans_origin = 1;
+  }
   
-  my $self = bless({ start => $start, end => $end, data => $data }, $class);
+  my $self = bless({ start => $start,
+                     end => $end,
+                     data => $data ,
+                     spans_origin => $spans_origin},
+                   $class);
   return $self;
 }
 
@@ -146,6 +163,23 @@ sub data {
   return $self->{data};
 }
 
+=head2 spans_origin
+
+  Arg []      : none
+  Description : Returns whether this interval was created spanning zero
+                (more particularly: if the interval was instantiated with start > end)
+  Returntype  : boolean
+  Exceptions  : none
+  Caller      : general
+
+=cut
+
+sub spans_origin {
+  my $self = shift;
+
+  return $self->{spans_origin};
+}
+
 =head2 is_empty
 
   Arg []      : none
@@ -159,7 +193,11 @@ sub data {
 sub is_empty {
   my $self = shift;
 
-  return $self->start >= $self->end;
+  if ($self->spans_origin) {
+    return ($self->end >= $self->start);
+  } else {
+    return ($self->start >= $self->end);
+  }
 }
 
 =head2 is_point
@@ -194,7 +232,11 @@ sub contains {
   return 0 if $self->is_empty or not defined $point;
   throw 'point must be a number' unless looks_like_number($point);
   
-  return ($point >= $self->start and $point <= $self->end);
+  if ($self->spans_origin) {
+    return ($point >= $self->start or $point <= $self->end);
+  } else {
+    return ($point >= $self->start and $point <= $self->end);
+  }
 }
 
 =head2 intersects
@@ -210,16 +252,26 @@ sub contains {
 sub intersects {
   my ($self, $interval) = @_;
   assert_ref($interval, 'Bio::EnsEMBL::Utils::Interval');
-    
-  return ($self->start <= $interval->end and $interval->start <= $self->end);
+
+  if ($self->spans_origin and $interval->spans_origin) {
+    return 1;
+  } elsif ($self->spans_origin or $interval->spans_origin) {
+    return ($interval->end >= $self->start or $interval->start <= $self->end);
+  } else {
+    return ($self->start <= $interval->end and $interval->start <= $self->end);
+  }
 }
 
 =head2 is_right_of
 
   Arg [1]     : An instance of Bio::EnsEMBL::Utils::Interval or a scalar
-  Description : Checks if this current interval is entirely to the right of a point. 
+  Description : Checks if this current interval is entirely to the right of a point
+                or Interval.
                 More formally, the method will return true, if for every point x from 
-                the current interval the inequality x > point holds.
+                the current interval the inequality x > point holds, where point
+                is either a single scalar, or point is the end of another Interval.
+                If spans_origin is true for either this Interval or an Interval
+                passed in, then this method returns false.
   Returntype  : boolean
   Exceptions  : none
   Caller      : general
@@ -232,18 +284,26 @@ sub is_right_of {
   return 0 unless defined $other;
 
   if ( looks_like_number($other) ) {
-    return $self->start > $other;
+    return $self->spans_origin ?
+           throw "is_right_of not defined for an interval that spans the origin" :
+           $self->start > $other;
+  } elsif ($self->spans_origin or $other->spans_origin) {
+    throw "is_right_of not defined for an interval that spans the origin";
+  } else {
+    return $self->start > $other->end;
   }
-
-  return $self->start > $other->end;
 }
 
 =head2 is_left_of
 
   Arg [1]     : An instance of Bio::EnsEMBL::Utils::Interval or a scalar
-  Description : Checks if this current interval is entirely to the left of a point. 
+  Description : Checks if this current interval is entirely to the left of a point
+                or Interval.
                 More formally, the method will return true, if for every point x from 
-                the current interval the inequality x < point holds.
+                the current interval the inequality x < point holds, where point
+                is either a single scalar, or point is the start of another Interval.
+                If spans_origin is true for either this Interval or an Interval
+                passed in, then this method returns false
   Returntype  : boolean
   Exceptions  : none
   Caller      : general
@@ -256,10 +316,14 @@ sub is_left_of {
   return 0 unless defined $other;
 
   if ( looks_like_number($other) ) {
-    return $self->end < $other;
+    return $self->spans_origin ?
+           throw "is_left_of not defined for an interval that spans the origin" :
+           $self->end < $other;
+  } elsif ($self->spans_origin or $other->spans_origin) {
+    throw "is_left_of not defined for an interval that spans the origin";
+  } else {
+    return $self->end < $other->start;
   }
-
-  return $self->end < $other->start;
 }
 
 1;

modules/Bio/EnsEMBL/Utils/Tree/Interval/Immutable.pm

@@ -62,6 +62,9 @@ be added to or removed from the tree during its life cycle.
 
 Implementation heavily inspired by https://github.com/tylerkahn/intervaltree-python
 
+This implementation does not support Intervals having a start > end - i.e.
+intervals spanning the origin of a circular chromosome.
+
 =head1 METHODS
 
 =cut
@@ -297,6 +300,9 @@ sub _divide_intervals {
   my ($s_center, $s_left, $s_right) = ([], [], []);
   
   foreach my $interval (@{$intervals}) {
+    if ($interval->spans_origin) {
+      throw "Cannot build a tree containing an interval that spans the origin";
+    }
     if ($interval->end < $x_center) {
       push @{$s_left}, $interval;
     } elsif ($interval->start > $x_center) {

modules/Bio/EnsEMBL/Utils/Tree/Interval/Mutable/PP.pm

@@ -48,6 +48,7 @@ use Carp;
 
 use Bio::EnsEMBL::Utils::Tree::Interval::Mutable::Node;
 use Bio::EnsEMBL::Utils::Interval;
+use Bio::EnsEMBL::Utils::Exception qw(throw);
 
 =head2 new
 
@@ -108,7 +109,7 @@ sub size {
   Example     : $tree->insert(Bio::EnsEMBL::Utils::Interval->new(10, 20, 'data'));
   Description : Insert an interval in the tree
   Returntype  : scalar (1), upon success
-  Exceptions  : none
+  Exceptions  : thrown if Interval spans origin (has start > end)
   Caller      : general
 
 =cut
@@ -116,6 +117,9 @@ sub size {
 sub insert {
   my ($self, $i) = @_;
   
+  if ($i->spans_origin) {
+    throw "Cannot insert an interval that spans the origin into a mutable tree";
+  }
   # base case: empty tree, assign new node to root
   unless (defined $self->root) {
     $self->root(Bio::EnsEMBL::Utils::Tree::Interval::Mutable::Node->new($self, $i));

modules/t/interval_tree_immutable.t

@@ -25,8 +25,6 @@ use_ok 'Bio::EnsEMBL::Utils::Interval';
 
 throws_ok { Bio::EnsEMBL::Utils::Interval->new() } qr/specify.+?boundaries/, 'Throws with no arguments';
 throws_ok { Bio::EnsEMBL::Utils::Interval->new(1) } qr/specify.+?boundaries/, 'Throws with an undefined argument';
-throws_ok { Bio::EnsEMBL::Utils::Interval->new(10, 1) } qr/start.+?end/, 'Throws with invalid arguments';
-throws_ok { Bio::EnsEMBL::Utils::Interval->new(100, 10) } qr/start.+?end/, 'Throws with invalid arguments';
 
 # degenerate (point) case
 my $i = Bio::EnsEMBL::Utils::Interval->new(10, 10);
@@ -36,27 +34,82 @@ ok($i->is_point, 'interval is point');
 
 # a normal interval, start < end
 $i = Bio::EnsEMBL::Utils::Interval->new(100, 200);
+# an interval spanning the origin, start > end
+my $i_span = Bio::EnsEMBL::Utils::Interval->new(200,100);
+
 isa_ok($i, 'Bio::EnsEMBL::Utils::Interval');
+isa_ok($i_span, 'Bio::EnsEMBL::Utils::Interval');
+
+is($i->spans_origin, 0, 'spans_origin false for non-spanning interval');
+is($i_span->spans_origin, 1, 'spans_origin true for spanning interval');
+
 is($i->start, 100, 'start position');
 is($i->end, 200, 'end position');
+is($i_span->start, 200, 'spanning start position');
+is($i_span->end, 100, 'spanning end position');
+
 ok(!$i->is_empty, 'interval not empty');
 ok(!$i->is_point, 'interval\'s not a point');
+ok(!$i_span->is_empty, 'interval not empty');
+ok(!$i_span->is_point, 'interval\'s not a point');
+
 ok($i->contains(100) && $i->contains(200) && $i->contains(150), 'interval contains points');
 ok(!$i->contains(99) && !$i->contains(201), 'interval does not contain points');
+ok($i_span->contains(100) && $i_span->contains(200) && $i_span->contains(250), 'spanning interval contains points');
+ok(!$i_span->contains(101) && !$i_span->contains(199), 'spanning interval does not contain points');
+
 # check is_right_of/is_left_of with point/interval
 ok(!$i->is_right_of && !$i->is_left_of, 'interval is not left/right of nothing');
+ok(!$i_span->is_right_of && !$i_span->is_left_of, 'spanning interval is not left/right of nothing');
+
 ok($i->is_right_of(99), 'interval right of point');
 ok(!$i->is_right_of(100) && !$i->is_right_of(150) && !$i->is_right_of(201), 'interval not right of point');
 ok($i->is_left_of(201), 'interval left of point');
 ok(!$i->is_left_of(99) && !$i->is_left_of(150) && !$i->is_left_of(200), 'interval not left of point');
+throws_ok { $i_span->is_right_of(150) }
+  qr/is_right_of not defined for an interval that spans the origin/,
+  'exception calling is_right_of with a spanning interval and a point';
+throws_ok { $i_span->is_left_of(150) }
+  qr/is_left_of not defined for an interval that spans the origin/,
+  'exception calling is_left_of with a spanning interval and a point';
+
 my $j = Bio::EnsEMBL::Utils::Interval->new(50, 99);
 my $k = Bio::EnsEMBL::Utils::Interval->new(50, 150);
 my $l = Bio::EnsEMBL::Utils::Interval->new(201, 250);
+my $m = Bio::EnsEMBL::Utils::Interval->new(101, 199);
+my $n_span = Bio::EnsEMBL::Utils::Interval->new(201,100);
+
+# non-spanning with non-spanning query
 ok($i->is_right_of($j), 'interval right of another');
 ok(!$i->is_right_of($k) && !$i->is_right_of($l), 'interval not right of others');
 ok($i->is_left_of($l), 'interval left of another');
 ok(!$i->is_left_of($j) && !$i->is_left_of($k), 'interval not left of others');
 
+# non-spanning with spanning query
+throws_ok { $i->is_right_of($n_span) }
+  qr/is_right_of not defined for an interval that spans the origin/,
+  'exception calling is_right_of with a spanning interval';
+throws_ok { $i->is_left_of($n_span) }
+  qr/is_left_of not defined for an interval that spans the origin/,
+  'exception calling is_left_of with a spanning interval';
+
+# spanning with non-spanning query
+throws_ok { $i_span->is_right_of($m) }
+  qr/is_right_of not defined for an interval that spans the origin/,
+  'exception calling is_right_of with a spanning interval';
+throws_ok { $i_span->is_left_of($m) }
+  qr/is_left_of not defined for an interval that spans the origin/,
+  'exception calling is_left_of with a spanning interval';
+
+# spanning with spanning query
+throws_ok { $i_span->is_right_of($n_span) }
+  qr/is_right_of not defined for an interval that spans the origin/,
+  'exception calling is_right_of with a spanning interval';
+throws_ok { $i_span->is_left_of($n_span) }
+  qr/is_left_of not defined for an interval that spans the origin/,
+  'exception calling is_left_of with a spanning interval';
+
+
 # check interval data
 $j = Bio::EnsEMBL::Utils::Interval->new(100, 200, [100, 200]);
 is_deeply($j->data, [100, 200], 'interval data');
@@ -64,17 +117,31 @@ is_deeply($j->data, [100, 200], 'interval data');
 # check intersection with other intervals
 $k = Bio::EnsEMBL::Utils::Interval->new(50, 150);
 ok($i->intersects($k), 'intervals intersect');
+ok($i_span->intersects($k), 'spanning interval and interval intersect');
 $k = Bio::EnsEMBL::Utils::Interval->new(150, 250);
 ok($i->intersects($k), 'intervals intersect');
 $k = Bio::EnsEMBL::Utils::Interval->new(50, 99);
 ok(!$i->intersects($k), 'intervals do not intersect');
 $k = Bio::EnsEMBL::Utils::Interval->new(201, 250);
 ok(!$i->intersects($k), 'intervals do not intersect');
+$k = Bio::EnsEMBL::Utils::Interval->new(101,199);
+ok(!$i_span->intersects($k), 'spanning interval and interval do not intersect');
+ok($i_span->intersects($n_span), 'two spanning intervals intersect');
+ok($i->intersects($n_span), 'interval and spanning interval intersect');
+my $o_span = Bio::EnsEMBL::Utils::Interval->new(201,99);
+ok(!$i->intersects($o_span), 'interval and spanning interval do not intersect');
 
 use_ok 'Bio::EnsEMBL::Utils::Tree::Interval::Immutable::Node';
 
 use_ok 'Bio::EnsEMBL::Utils::Tree::Interval::Immutable';
 
+my $intervals_with_span = [ Bio::EnsEMBL::Utils::Interval->new(20, 30),
+                            Bio::EnsEMBL::Utils::Interval->new(30, 20)];
+
+throws_ok { my $impossible_tree = Bio::EnsEMBL::Utils::Tree::Interval::Immutable->new($intervals_with_span) }
+  qr/Cannot build a tree containing an interval that spans the origin/,
+  'exception when building an interval tree with an interval that spans the origin';
+
 my $intervals = [ Bio::EnsEMBL::Utils::Interval->new(121626874, 122092717),
 		  Bio::EnsEMBL::Utils::Interval->new(121637917, 121658918),
 		  Bio::EnsEMBL::Utils::Interval->new(122096077, 124088369) ];

modules/t/interval_tree_mutable.t

@@ -122,6 +122,11 @@ is(scalar @{$search_result}, 2, 'Number of search results');
 is($search_result->[0]->data, 'data1', 'Search result');
 is($search_result->[1]->data, 'data2', 'Search result');
 
+$tree = Bio::EnsEMBL::Utils::Tree::Interval::Mutable->new();
+throws_ok { $tree->insert(make_interval(200, 100, 'spanning_interval')) }
+  qr/Cannot insert an interval that spans the origin into a mutable tree/,
+  'exception when trying to insert an interval that spans the origin';
+
 $tree = Bio::EnsEMBL::Utils::Tree::Interval::Mutable->new();
 map { $tree->insert($_) } @{$intervals};
 is($tree->size(), scalar @{$intervals}, 'Tree size');