Differential Conflation makes use of the Unifying conflation framework within Hootenanny, but attempts a much simpler operation than "full" conflation. The idea is to allow a user to take two datasets, A and B, and generate a changeset from them such that the changeset includes all of the data in set B that does not overlap (conflict, need merging, etc.) with set A. The changeset can then be applied directly to dataset A, without generating any review items.
To re-iterate, Hootenanny will consider any conflatable feature type during differential conflation. Types that cannot be conflated by Hootenanny (that is, types for which hootenanny does not have a matcher) will be removed from the differential output. This is to prevent possible duplication of features when differential conflation is used in an automated fashion to bulk-process data.
In a nutshell, the basic differential algorithm works like this: both inputs are loaded into a single map, with elements marked "unknown1" and "unknown2", respectively. Matches are generated using all specified matchers, then all of the elements involved in the matches are removed from the map. Next, all of the remaining elements from the first input ("unknown1") are removed. At this point, the only elements remaining in the map will be elements from the second input that did not match any elements in the first input.
So now we have all of the "new" geometry from the second input isolated as output, but what if there
were also new or updated tags on some of the elements in the second input? How can we capture those
new tags? If the conflate.differential.include.tags option is enabled, we simply go back through the
matches derived in the previous step, and compare the tags of the elements involved in the match. If the
element from the second input contains new tag keys or tag values different from the first input element,
the tags are merged using an overwriting algorithm. All key/value pairs are kept from the first
input, all key/value pairs from the second input are appended, and any keys that conflict are
assigned the value from the second input, as it is assumed that the second input contains updated
information. The merged tag set is applied to the original element geometry, and it is written out
as a changeset "modify" directive.
By default, when using Differential Conflation against linear features other than rivers with the
Unifying Roads Algorithm, it calculates the differential against partial linear feature matches
rather than whole matches. This generally generates a more granular differential. Differentials
generated with river data, however, do not seem to benefit from partial linear match removal the
majority of the time. There may, however, be situations when a less granular differential is
desired, and the default behavior for linear non-river features may be turned off by enabling the
differential.remove.linear.partial.matches.as.whole configuration option (enable "Remove features
involved in linear partial matches completely" option in the UI). To control river partial matching
with Differential Conflation, use the differential.remove.river.partial.matches.as.whole ("Remove
rivers involved in river partial matches completely" option in the UI).
Differential Conflation can further be customized by modifying the
conf/core/DifferentialConflation.conf configuration options file.