Best afterLinear search

Filtering a list

Filtering — keeping only the items that pass some test — is the everyday move behind "find the data that matters." The safe way is to build a new list. The tempting way is to delete the unwanted items in place, and the state pane shows exactly why that backfires.

Keep only the numbers greater than 10, and leave the original list untouched.

Filtering a list — deleting as you go execution-derived · CPython

same data, two zoom levels

state after line 1 runs

nums

432218715

step 1 / 12

Click a line, drag the slider, or use the ← → keys.

What you are looking at

The list is [4, 3, 22, 18, 7, 15] and we want only the numbers over 10. The right answer is [22, 18, 15].

Buggy — deletes each small number from nums while looping over it. Removing 4 shifts 3 into a slot the loop has already passed, so 3 is skipped and survives. nums ends as [3, 22, 18, 15] — a small number leaked through, and the original list is destroyed.
Procedural — copies the passing numbers into a fresh list big and never touches nums. Nothing shifts under the loop, so nothing is skipped.
Compact — [n for n in nums if n > 10] builds the new list in one expression. Python still runs it as a loop — watch n take each value — andbig appears complete at the end.

Every state you see came from running the program under CPython's tracer at build time — see how GlassBox stays honest.