Stack Rows

Aggregation

Concatenate rows from two or more inputs into a single output. In union mode the output schema is the union of all input columns and missing fields are filled with null; in strict mode every input must already share the same columns. Use it to combine monthly exports, append archived rows to live data, or merge per-region files before further processing.

Local processing

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

How it works

Stack Rows is a multi-input batch transform: it accepts an arbitrary number of input streams and emits one combined output. Inputs are read fully into memory and concatenated in input order (input 1’s rows first, then input 2’s rows, etc.); within each input, original row order is preserved.

In union mode, the executor scans every row across every input to build the union of column names, then normalizes each row to that full column set with missing keys set to null. This means the output rows always have identical shape, even if the source files had different columns.

In union_strict mode, no normalization happens — rows are concatenated as-is. The schema check happens at design time via getOutputSchema; if you connect inputs with mismatched columns the flow editor surfaces an error before execution.

Input: Two or more tabular data connections. Output: All rows from all inputs, in input order, with the union (or strict) column set.

Options

Option	Type	Default	Description
mode	"union" | "union_strict"	"union"	"union" fills missing columns with null. "union_strict" rejects schema mismatches at design time.

Examples

Combine two monthly exports with the same columns

You have January and February sales exports with identical columns and want a single combined dataset.

Input 1 (January):

order_id	customer	amount
1001	Acme Corp	250
1002	Beta Inc	750

Input 2 (February):

order_id	customer	amount
1003	Gamma LLC	1100
1004	Delta Co	500

Configuration: mode: union (or union_strict — both produce identical output when schemas match).

After:

order_id	customer	amount
1001	Acme Corp	250
1002	Beta Inc	750
1003	Gamma LLC	1100
1004	Delta Co	500

Stack inputs with different columns (union mode fills nulls)

The 2025 export added a region column that wasn’t in the 2024 export. You want both years stacked.

Input 1 (2024):

order_id	customer	amount
901	Acme Corp	200

Input 2 (2025):

order_id	customer	amount	region
1001	Beta Inc	800	EU

Configuration: mode: union.

After:

order_id	customer	amount	region
901	Acme Corp	200
1001	Beta Inc	800	EU

The first row’s region is null because the 2024 input did not have that column.

Tips and Edge Cases

• Union mode rebuilds every row. The executor walks every row across every input twice — once to discover column names, once to normalize — so memory and CPU scale with total row count, not just unique-column count. For very large stacks where you know the schemas already match, prefer union_strict to skip the second pass. See apps/web/src/transforms/stack-rows/logic.ts:45-65.
• Missing columns are null, not empty string. Filter Rows’ is empty operator matches both, but downstream Change Type rules on null skip the parser and write null; on "" the string target writes "". The distinction matters when exporting to CSV where the two render differently. See apps/web/src/transforms/stack-rows/logic.ts:53-58.
• Strict mode is enforced at design time, not at run time. union_strict does not error if rows have unexpected keys — it concatenates them as-is and downstream sees a ragged schema. The schema-mismatch error comes from getOutputSchema in the flow editor. See apps/web/src/transforms/stack-rows/logic.ts:41-43 and apps/web/src/transforms/stack-rows/logic.ts:78-87.

Related Transforms

• Lookup — join rows by a key column instead of stacking them vertically.
• Group By — collapse stacked rows into per-key aggregates after combining.
• Filter Rows — remove unwanted rows from a stacked dataset.