Why are pandas ufuncs recommended over apply?

Pandas has an apply function that lets you run arbitrary functions across every value in a column. The catch is that apply is only marginally faster than a plain Python loop. That’s why pandas’ built-in ufuncs are the preferred choice for column preprocessing.
ufuncs are special functions built on top of numpy and implemented in C, which is why they’re so fast. Below we introduce several examples of ufuncs: .diff, .shift, .cumsum, .cumcount, .str commands (for strings), and .dt commands (for dates).