Split Tags to Rows — Solution & ExplanationProblems / Split Tags to Rows / EditorialApproach

string_split(tags, ',') turns each tags string into a list, and unnest(...) expands the list into one row per element — the PySpark equivalent is F.explode(F.split('tags', ',')).

The raw fragments keep the stray whitespace from the source data (' portable'), so apply trim() after splitting. Trimming before the split would only clean the ends of the whole string, not each fragment.

Group by the trimmed tag and count(*) — each product lists a tag at most once, so counting rows counts products.

Order by product_count DESC, tag for a stable ranking.

Key InsightA comma-separated column is a list masquerading as a string — un-nesting it back into rows (unnest/explode) is the wide-to-long reshape that makes counting trivial. The classic bug here is trim placement: trim must run on each fragment after the split, otherwise 'portable' and ' portable' are counted as two different tags.Hints

string_split(tags, ',') turns the string into a list; unnest(...) (or F.explode in PySpark) expands the list to one row per element.

Split first, trim second: the stray spaces live after the commas, so trimming the whole tags string before splitting fixes nothing.

Group by the trimmed tag with count(*) — without trim(), 'portable' and ' portable' land in different groups and the counts come out wrong.

Solve Split Tags to Rows yourself →