frictionless infers decimal-comma prices ('1,5') as geopoints

frictionless 5.21.0 type inference tries geopoint (default format 'lon, lat') second in DEFAULT_FIELD_CANDIDATES, ahead of integer, number and string. So a European CSV column like price;1,5;2,25 is described as type 'geopoint', and frictionless.extract returns [1.0, 5.0] and [2.0, 25.0] instead of the numbers. Values outside the coordinate ranges (0,99 or 181,5) fall back to string. When describing EU data, pass an explicit schema (or field_candidates/field_type in the Detector), or post-check the inferred types for geopoint before publishing a data package. Related, in agate 1.14 the default Number type reads '1,5' as Decimal('15') (en_US grouping).