Unpivot

Transform

Reshape wide tables into long format. Selected columns are melted into two new columns: one holding the original column name, one holding its value. Each input row produces one output row per unpivoted column. Use it to convert spreadsheet-style monthly columns into a single date column, or to feed wide pivot tables into transforms that expect tidy data.

Local processing

This transform runs entirely in your browser. No data is sent to external servers.

How it works

Unpivot is a full-dataset batch transform: it buffers the entire input before producing output, because it needs the row shape to determine which columns are index columns (everything not unpivoted) versus value columns (the ones being melted).

For an input row with N unpivoted columns, the transform emits N output rows. Each output row carries:

• All non-unpivoted columns from the source row, unchanged (the “index” columns)
• The variable column, set to the name of one of the unpivoted columns
• The value column, set to the cell value at that column (or null if missing)

The order of output rows is: input row 1’s first unpivoted column, then input row 1’s second, …, then input row 2’s first, and so on. The order of variables within a group follows the order of columnsToUnpivot in the configuration, not the column order of the source.

If columnsToUnpivot is empty, rows pass through unchanged.

Input: One tabular data connection. Output: A long-format table with all index columns plus the variable and value columns.

Options

Option	Type	Default	Description
columnsToUnpivot	string[]	[]	Names of columns to melt. Empty list is a passthrough.
variableColumnName	string	"variable"	Name of the output column that holds the original column name. Required.
valueColumnName	string	"value"	Name of the output column that holds the cell value. Required.

Examples

Melt monthly columns into a long format

You have quarterly sales as wide columns (q1, q2, q3, q4) and want them as one quarter / sales pair per row.

Before:

customer	q1	q2	q3	q4
Acme Corp	1200	1500	1800	2100
Beta Inc	800	900	1100	1300

Configuration: columns to unpivot: ["q1", "q2", "q3", "q4"], variable column: quarter, value column: sales.

After:

customer	quarter	sales
Acme Corp	q1	1200
Acme Corp	q2	1500
Acme Corp	q3	1800
Acme Corp	q4	2100
Beta Inc	q1	800
Beta Inc	q2	900
Beta Inc	q3	1100
Beta Inc	q4	1300

Multiple index columns are preserved per output row

You want to melt scores while keeping both student and class as identifiers.

Before:

student	class	math	english	science
Alice	5A	92	88	95
Bob	5B	78	84	80

Configuration: columns to unpivot: ["math", "english", "science"], variable column: subject, value column: score.

After:

student	class	subject	score
Alice	5A	math	92
Alice	5A	english	88
Alice	5A	science	95
Bob	5B	math	78
Bob	5B	english	84
Bob	5B	science	80

Tips and Edge Cases

• Index columns are inferred from the first row, not from the schema. The transform takes Object.keys(rows[0]) and treats every key not in columnsToUnpivot as an index column. If row 1 is missing a key that later rows have, that key is dropped from the output. To avoid surprises on ragged inputs, run Stack Rows in union mode upstream so every row has the same shape. See apps/web/src/transforms/unpivot/logic.ts:72-74.
• Missing values become null, not empty string. When an unpivoted column is absent from a row (or its value is undefined), the output value cell is null. Empty string in the source is preserved as "". See apps/web/src/transforms/unpivot/logic.ts:84.
• One input row produces N output rows in a deterministic order. The output groups together one input row’s unpivoted columns before moving on to the next input row. Sort downstream if you need a different order — Unpivot does not interleave rows by variable.

Related Transforms

• Pivot — the inverse: roll long-format data back into wide columns.
• Group By — aggregate the unpivoted long format (e.g. sum sales per quarter across all customers).
• Filter Rows — drop specific variables (e.g. keep only q4 rows) after unpivoting.