Remove duplicate rows in one column

[Léa]

How can I remove duplicated rows in one column ?

[jereze]

There is currently no way to do that in a visual preparation recipe* (because a visual recipe more or less works row by row, and it cannot work on a full column, as it is designed for big data).

It's possible to do so in a visual GROUP recipe: click “Show mass actions”, select all columns, click “use as grouping keys”. If the csv is very big, I suggest synchronizing to a SQL DB first.

You can also do so in coding recipes:

• In a Python recipe, you can use the Pandas function (see example below) drop_duplicates()


• In a R recipe, you have several alternative (duplicated(), dplyr, ..): read here


• In a SQL recipe, I would use a a group by with min or max, or window function with partition by key and keep the first row.

* There is actually one way to do it in a visual preparation recipe, with a custom Python function, but that will not work all the time (if the recipe is multi-threaded), so I would not recommend this trick:

I hope that helps,

Jeremy

[Mattsco]

Hi,



In a python recipe I would do:

# -*- coding: utf-8 -*-
import dataiku
import pandas as pd

# Recipe inputs
df = dataiku.Dataset("input_dataset").get_dataframe()

df.drop_duplicates(subset=["my_key_colum"], inplace=True)
# or
# df.drop_duplicates(inplace=True)
# to use all columns to compare for duplicates

# Recipe outputs
out = dataiku.Dataset("output_dataset").write_with_schema(df)

Matt