When creating a canonical list, can non-string values be combined in various ways?

[webmaven]

Examples:

1. If a field is a list of strings (eg. tags), the canonical field value could be either a de-duped union of the lists, or the intersection.
2. Numeric values could produce the mean (or median).
3. Numeric values could produce a range (highest and lowest values seen)
4. Numeric ranges (eg. 60-70) could produce a range where the high and low values are medians or means
5. Geo coordinates could produce the centroid.

Etc.

For my use case I am most interested in the first four.

[fgregg]

Our current canonicalize method does not implement those things, but could.

[fgregg]

I've brought canonicalize back into core dedupe so repopening.