It is possible to define custom Splitfile commands that follow the normal Splitfile execution semantics (cache invalidation and idempotency), allowing you to write your own transformation or data import stages.
The basic idea is that a custom command extends
implements at least one method,
execute() that runs the command and
returns the command context hash which is combined with the hash of the
previous image to create the hash of the new image. Additionally, the
custom command can also implement
calc_hash() if it's able to
calculate the command context hash without actually running the command.
For example, for a simple data importer that loads a file from HTTP,
calc_hash() could return the hash of the file's last modified
execute() would perform the actual computation-heavy