Explore populated values

value_patterns complements availability analysis with evidence about populated values.

patterns = fw.value_patterns(df, by=["exam"], max_pairs=100, max_patterns=10)
print(patterns.to_frame("summaries"))
print(patterns.to_frame("findings"))

String summaries count ASCII letter/digit formats, three-character prefixes and lengths. Indexed columns such as image_1 and image_2 form a name-based family; an additional evidence label records identical availability when observed. A name family is not an assertion of equivalent values.

Numeric summaries include finite ranges, nonfinite counts, smallest observed spacing and whether values lie on that observed grid. Paired numeric fields can produce constant offset or ratio findings; ratios exclude zero denominators. These checks use at least two finite rows and record excluded rows and numeric tolerances.

Context constancy asks how many populated context groups contain only one populated value for a field. Use dependency discovery for full exact/approximate mapping and exception-group evidence. Foundation levels, pairs and joint_counts provide distributions, common joint combinations, context-specific associations and empirical or declared-domain absence.