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”).

Leave a comment on github