Sample

Take a subset of rows from your data using various sampling strategies.

Palette name	Sample (category Compute)
Type name	Sample

Sockets

Socket	Direction	Description
input	Input	Data to sample from
output	Output	Sampled rows

Sampling Modes

Mode	Description	Blocking
First N	Take the first N rows	No (streaming)
Last N	Take the last N rows	Yes
Every N	Take every Nth row (with optional offset)	No (streaming)
Random	Take N random rows	Yes
Random Percent	Take a random percentage of rows	Yes
First Percent	Take the first N% of rows	Yes

Blocking Modes

Last N, Random, Random Percent, and First Percent modes require reading all data before producing output. This means the entire dataset must fit in memory.

First N and Every N are streaming-friendly and work efficiently with large datasets.

Configuration

Option	Default	Applies To	Description
Mode	First N	All	Sampling strategy to use
N	100	First N, Last N, Random, Every N	Number of rows (or interval for Every N)
Percent	10.0	Random Percent, First Percent	Percentage of rows (0-100)
Offset	0	Every N	Starting row offset
Seed	(none)	Random, Random Percent	Random seed for reproducible results

Examples

Take First 1000 Rows

1. Connect data to the Sample tool
2. Set Mode to First N
3. Set N to 1000

Random 10% Sample

1. Connect data to the Sample tool
2. Set Mode to Random Percent
3. Set Percent to 10
4. Optionally set a Seed for reproducibility

Every 5th Row

1. Connect data to the Sample tool
2. Set Mode to Every N
3. Set N to 5
4. Set Offset to 0 (start from the first row)

Notes

• The output schema is always identical to the input schema
• First N and Every N are the most memory-efficient modes since they don't require loading all data
• Use Seed with Random/Random Percent modes to get the same sample each time you run the workflow
• Every N with an offset of 2 and N of 3 takes rows 2, 5, 8, 11, etc.