Per issue #19, the persistent homology computation steps now modify
data set list-columns in place, and these and the disaggregation steps
assign roles as conventional {recipes} steps. Previously, users could
ignore the presence of persistence diagram list-columns because they
were assigned the new role "persistent homology". They must
now be manually removed or their roles updated. Alternatively, and
preferably, the data set list-columns can have their roles updated to
something other than "predictor", as illustrated in the
README. Users now also lose access to the data set list-column unless
they explicitly duplicate it before computing persistent homology.
The minor version incorporates a patch to debug the vectorization tests, which had failed to run due to a prefix in the names of the generated test files. Once corrected, the tent template functions step failed one test due to missing parameters; this has been corrected.
Default scale sequences derived from persistence diagrams now use extrema calculated after infinite values are dropped. This accommodates an upgrade to {ripserr} that returns features with deaths at infinity.