M
Mark Nelson [MS]
Visio's SpatialNeighbors method is useful for finding shapes that are nearby
other shapes or that overlap other shapes. It can be used to
programmatically answer the question "What did I drop my shape on top of?".
Unfortunately, the SpatialNeighbors method can be extremely slow under
certain circumstances. In real world scenarios, searching might take
anywhere from a second to over a minute. That's unacceptable to some
developers who need to make frequent calls to the method. Two factors can
significantly alter the performance of the method.
First, searching group shapes is much slower than searching non-group
shapes. This has to do with the way that SpatialNeighbors defines
containment. Shape A is contained within group Shape B if A is contained
within the union of all the sub-shapes of B. Thus a lot of work is done to
determine that union before any relationship determination is made. It may
not be possible to eliminate groups, but it is always helpful to performance
to minimize the number of sub-shapes as much as possible. Also make sure to
avoid nested groups. Opt instead for of a single group shape with many
sub-shapes. Reducing the number of groups or shapes within groups can
double or triple the performance of SpatialNeighbors.
That's a decent improvement, but the second factor is even more important
for performance. With each call to SpatialNeighbors, the developer must
specify the Relation types they wish to detect. The four relation types are
ContainedIn, Contain, Overlap and Touching. Any combination of types can be
searched for. The problem is that Visio checks for each of these
relationships individually and must redo all the analysis work described
above every time. This leads to astronomic execution times. The workaround
here is to not specify the relation type at all. When passing in 0 for the
Relation parameter, Visio makes several internal optimizations because it no
longer cares about the particular way in which two shapes are related. For
developers who don't care, passing in 0 is a couple orders of magnitude
faster (100x) than passing in all the types. For those that really do need
to find the shapes matching a particular relationship, call SpatialNeighbors
passing in 0 to get a collection of candidate shapes. Then iterate through
the collection and use the SpatialRelation property to get the relationship.
This technique uses the really fast searching code to weed out all the
shapes that are completely unrelated. The expensive code to determine
actual relationships is then used with a much smaller set of shapes.
other shapes or that overlap other shapes. It can be used to
programmatically answer the question "What did I drop my shape on top of?".
Unfortunately, the SpatialNeighbors method can be extremely slow under
certain circumstances. In real world scenarios, searching might take
anywhere from a second to over a minute. That's unacceptable to some
developers who need to make frequent calls to the method. Two factors can
significantly alter the performance of the method.
First, searching group shapes is much slower than searching non-group
shapes. This has to do with the way that SpatialNeighbors defines
containment. Shape A is contained within group Shape B if A is contained
within the union of all the sub-shapes of B. Thus a lot of work is done to
determine that union before any relationship determination is made. It may
not be possible to eliminate groups, but it is always helpful to performance
to minimize the number of sub-shapes as much as possible. Also make sure to
avoid nested groups. Opt instead for of a single group shape with many
sub-shapes. Reducing the number of groups or shapes within groups can
double or triple the performance of SpatialNeighbors.
That's a decent improvement, but the second factor is even more important
for performance. With each call to SpatialNeighbors, the developer must
specify the Relation types they wish to detect. The four relation types are
ContainedIn, Contain, Overlap and Touching. Any combination of types can be
searched for. The problem is that Visio checks for each of these
relationships individually and must redo all the analysis work described
above every time. This leads to astronomic execution times. The workaround
here is to not specify the relation type at all. When passing in 0 for the
Relation parameter, Visio makes several internal optimizations because it no
longer cares about the particular way in which two shapes are related. For
developers who don't care, passing in 0 is a couple orders of magnitude
faster (100x) than passing in all the types. For those that really do need
to find the shapes matching a particular relationship, call SpatialNeighbors
passing in 0 to get a collection of candidate shapes. Then iterate through
the collection and use the SpatialRelation property to get the relationship.
This technique uses the really fast searching code to weed out all the
shapes that are completely unrelated. The expensive code to determine
actual relationships is then used with a much smaller set of shapes.