Open robinhoutmeyers opened 6 years ago
Our GeoPackage creation software defaults all feature attributes to TEXT because we encode the attribute values as Strings. We are looking into a type parsing solution to improve the SQL feature tables.
This wasn't addressed in Sprint 2 but it could be an interesting issue to mention in the report. If a feature table is supposed to contain a REAL column (e.g. elevation) but is published as a TEXT column there is no way a general-purpose ETS could catch it as an error, since attribute column types are up to the user. SQLite will not prevent selecting from a text column using a real value, and doing so will likely select rows in error.
The properties defined in the feature tables inside the GeoPackage Vector Rigel data source are all of type TEXT - regardless whether they express a number or not. A more appropriate type aligned with the content would be nice; our client can automatically use this information to ease the creation of data queries in the UI.