Generate sequence, restart at 1 each time a given column changes

UserBird

Hi,

If I have data like this:

A 1
A 5
A 7
B 1
B 10
B 20



I want to generate a third column with a sequence:

A 1 1
A 5 2
A 7 3
B 1 1
B 10 2
B 20 3

How can I do it in DSS (other than R, Python or SQL)?

Clément_Stenac

Hi Simon,



Unfortunately, this is not a feature that is builtin in Dataiku (yet), ie not something that you can do without coding.

This is definitely something we are considering currently, and it might become available soon.



Especially if your dataset is large and/or unsorted, the best way to do that would indeed be to use SQL partitioning. A SQL query recipe with something like:

SELECT category, numberdata, RANK() OVER (PARTITION BY category ORDER BY numberdata ASC);

would do the trick.



If your dataset is already ordered (for example, it's a file, only one file), you can also use the visual data preparation with a custom Python processor. Something like:

current_category = None
current_rank = 0

# Modify the process function to fit your needs
def process(row):
  global current_category, current_rank

  if current_category is None or row["category"] != current_category:
      # New category seen
      current_rank = 1
      current_category =row["category"]
  else:
      current_rank += 1

  row["rank"] = current_rank

Hope this helps,

Simon

Thanks. I would add to this using sequence from R (using datatables as an example)
DT.dt[, seq_var := sequence(.N), by = "A"])

To SQL I would also add that I should more probably use DENSE_RANK or even ROW_NUMBER if the SQL supports its.