A dataframe is a bad abstraction. Trading away robustness for simplicity… | by Niels Cautaerts | Medium
Discovered: Aug 19, 2024 14:50 A dataframe is a bad abstraction. Trading away robustness for simplicity… ¦ by Niels Cautaerts ¦ Medium <– I am sure it is but honestly what are the alternatives :-) ? –> QUOTE: The problem is that knowing that an object is a dataframe does not help you reason about the correctness of code in the slightest. Yet almost any operation you perform on a dataframe requires knowledge about the underlying data being stored. Otherwise you encounter runtime errors or you compute nonsense. Required knowledge may include basic things like the schema of a table (available columns and data types), but also more advanced rules like constraints that must hold over multiple columns (e.g. “the sum of these three columns must always equal 1”).