298aac33cc
Float values were often reported with extra precision from float32 -> 64 conversion. Add repr_float_32 to give cleaner output.