The Dataset.stats derives the number of rows in a table in two ways, either just reading the table's property dc:extent - if specified - or actually counting the rows (see https://github.com/cldf/pycldf/blob/874275ad6895c17a2c1375450edbb915ff001f28/src/pycldf/dataset.py#L520). This can be confusing, e.g. when a StructureDataset and a Wordlist are created via pylexibank and share the LanguageTable. Then the Wordlist will have dc:extent pruned to the languages actually referenced in forms, while the StructureDataset will not.
Thus, this counting behaviour should be specified by the caller explicitly, and be available as option in the cldf stats command.
The
Dataset.statsderives the number of rows in a table in two ways, either just reading the table's propertydc:extent- if specified - or actually counting the rows (see https://github.com/cldf/pycldf/blob/874275ad6895c17a2c1375450edbb915ff001f28/src/pycldf/dataset.py#L520). This can be confusing, e.g. when aStructureDatasetand aWordlistare created viapylexibankand share theLanguageTable. Then theWordlistwill havedc:extentpruned to the languages actually referenced in forms, while theStructureDatasetwill not.Thus, this counting behaviour should be specified by the caller explicitly, and be available as option in the
cldf statscommand.