Loops & Iteration

1. What Is It?

A loop is a construct that executes a block of code repeatedly, usually until a condition becomes false or until there are no more items to process. Without loops you'd have to unroll every repetition by hand — impossible for anything but tiny, fixed-size tasks.

The real idea behind loops isn't the for keyword — it's iteration: a disciplined walk through a sequence of values, with a well-defined start, end, and per-step behavior. Every off-by-one bug, infinite loop, and "why is this so slow?" question comes from misunderstanding one of those three.

2. How It Works

There are three fundamental loop shapes, and all languages provide some mix of them:

1. Counted loop — iterate a known number of times (for i in range(n)).
2. Condition loop — iterate while a condition holds (while cond:).
3. Iterator loop — walk every element of a collection (for x in xs:).

# Python
# counted
for i in range(5):
    print(i)                    # 0..4

# condition
n = 100
while n > 1:
    n //= 2                     # 100, 50, 25, 12, 6, 3, 1 — exits when n == 1

# iterator
for name in ["ada", "lin", "hopper"]:
    print(name)

Under the hood: iterators

The range-based / for-each loop is sugar for an iterator protocol. The language asks the collection for an iterator; at each step it asks "next element, please" until the iterator signals it's done.

# Python's for loop is equivalent to:
it = iter(xs)
while True:
    try:
        x = next(it)
    except StopIteration:
        break
    # ... use x

Java's for-each over an Iterable compiles to calls on Iterator.hasNext() / Iterator.next() (over an array, it compiles to an indexed loop). C++ range-for expands to a loop driven by begin() / end() iterators.

This matters because you can make your own types iterable (implement the protocol) and they slot into for seamlessly. It's also why modifying a collection mid-iteration often crashes: the iterator's invariants are broken.

break, continue, return

• break — exit the innermost loop immediately.
• continue — skip the rest of this iteration, proceed to the next.
• return (inside a function) — exits the loop and the function at once.

Python adds for ... else: — the else runs only if the loop finished without break. Unusual but useful for search loops.

3. What You Actually Need to Know

• Off-by-one errors come from fuzzy bounds. Standard convention is half-open: [start, end) — inclusive start, exclusive end. That's why range(n) gives 0..n-1 and for (i = 0; i < n; i++) is the canonical C-style loop.
• Infinite loops happen when the condition never changes. while (x > 0) inside a body that never decrements x will hang forever. If a loop freezes your program, add a print of the controlling variable.
• Modifying a collection while iterating is usually broken. Java throws ConcurrentModificationException. Python may skip elements or raise RuntimeError: dictionary changed size during iteration. C++ iterators become invalidated — undefined behavior. Build a new list instead, or iterate over a copy.
• Use the right shape. Don't fake a for-each with an index loop (for i in range(len(xs)): x = xs[i]) — it's slower, uglier, and bug-prone. Use for x in xs:. If you also need the index, use enumerate(xs) (Python) or for (int i = 0; i < xs.size(); i++) with intent.
• Prefer iterator protocols / generators for large data. Loading a 10GB file and iterating a list of all its lines will run out of memory; iterating a file handle reads line by line. Python generators (yield), Java Stream, C++ ranges all express "lazy" iteration.
• Debugging clues:
  • Loop runs one too few / one too many times → off-by-one; check < vs. <=.
  • Loop never terminates → update step missing, or condition can never become false.
  • ConcurrentModificationException / RuntimeError: dictionary changed size → you mutated a collection mid-iteration.
  • Wrong value for i after a C-style for → loop body rebound i; don't.

4. Language Differences

The C++ unsigned-loop trap is worth memorizing. for (size_t i = v.size() - 1; i >= 0; --i) is an infinite loop — i wraps around to a huge positive number instead of going negative.

5. Tradeoffs & Decisions

• Index-based vs. for-each. Use for-each by default — clearer, harder to get wrong, usually the same performance. Use index-based only when you need the index for logic (pairing, distance, reverse traversal).
• Eager vs. lazy iteration. Building a full list ([f(x) for x in data]) is fine for small data and easy to reason about. For pipelines over big data, lazy iteration (generators, streams) avoids peak memory and enables early termination.
• Recursion vs. iteration. Recursion can express tree/graph traversal more naturally. Iteration avoids stack overflow and is usually faster. Python has no tail-call optimization; a deep recursive loop will blow the stack.
• Parallel iteration. Java streams (.parallelStream()), C++17 parallel algorithms, and explicit thread pools let you split iterations across cores. Worth it only when each iteration is substantial work and the data can be partitioned cleanly.

If you see a loop mutating a shared collection it's iterating over, it almost always indicates a bug or an architectural issue — build a new collection instead. You'd choose a lazy generator when memory matters more than simplicity.

6. Interview Cheat Sheet

• The three loop shapes are counted (known count), condition (until a predicate changes), and iterator (walk a collection).
• For-each loops are sugar over an iterator protocol — the loop asks the collection for elements until it says "done."
• Off-by-one errors come from confusing inclusive vs. exclusive bounds; the half-open [start, end) convention is standard.
• Never modify a collection while iterating it. Build a new one or take a snapshot.
• break exits the innermost loop; continue skips to the next iteration.
• Python generators / Java streams / C++ ranges enable lazy iteration — process one element at a time without materializing the full sequence.

Follow-ups:

• "What does for x in xs actually do?" — Calls iter(xs) once, then next() repeatedly until StopIteration.
• "Why might iterating over a dict and mutating it crash?" — The iterator holds state referring to the dict's internal structure; mutating it breaks that state.
• "When should you use while over for?" — When you don't know the iteration count in advance and the stopping condition depends on computed state (e.g., convergence loops).