Copy-on-Write

noun

say it: “KOP-ee on RYT” /ˈkɑpi ɑn ˈraɪt/

A way of handling copies where a copy shares the original's data until one of them is changed, so copying is cheap and the original stays safe.

What is Copy-on-Write in pandas?

Copy-on-Write is how pandas 3 handles copies: a copy shares its data with the original until one of them changes, and only then is the data really copied. Copying is cheap, and changing a copy never changes the original. A chained assignment, such as df["temp_f"][0] = 61.8, changes nothing; set one value with .loc instead.

readings.loc[0, "temp_f"] = 61.8      # changes the table
# readings["temp_f"][0] = 61.8       # chained assignment: changes nothing in pandas 3

In a sentence

  • “Under Copy-on-Write, the chained assignment left the table unchanged.”
  • “Thanks to Copy-on-Write, df.copy() at the start of each step costs almost nothing.”

Level: Freelancer · More Python terms at this level

Updated