Pandas SettingWithCopyWarning — chained assignment pitfall

Pandas raises SettingWithCopyWarning when assigning values. The assignment sometimes works and sometimes silently fails. Data modifications don't persist as expected.

(1) The warning means you might be modifying a view instead of the original DataFrame. (2) WRONG: df[df['col'] > 5]['col2'] = 10 — this chains indexing, creating an intermediate view. (3) RIGHT: df.loc[df['col'] > 5, 'col2'] = 10 — single indexing operation on original. (4) When slicing: use .copy() if you want an independent DataFrame: subset = df[df['col'] > 5].copy(). (5) In pandas 3.0+: Copy-on-Write (CoW) mode changes this behavior — enable with pd.options.mode.copy_on_write = True. (6) Rule: use .loc for label-based setting, .iloc for position-based setting. Never chain [] operators for assignment.

Chained indexing (df[x][y] = v) creates an intermediate object that may be a view or a copy depending on the data types and memory layout. If it's a copy, the assignment is lost.