PR got comments from someone who successfully tested it, but so far no code review, yet.
I wonder whether me implementing this as three packages is the right approach:
package prevo-tools provides
the executable prevodb to pre-process the dictionary data
the executable prevo to query the dictionary (requiring the pre-processed data)
They’re build from the same source code repo and have a relation similar to the one between updatedb and locate.
package prevo-data provides the pre-processed dictionary data (using prevodb from prevo-tools at build-time)
package prevo links the other two packages together, so that the prevo command will use the data from prevo-data by default. (Can be overridden by a command line option.)