# Single File > However deep the nesting, every voice arrives in its own turn. Canonical URL: Domain: Python · Difficulty: easy · Seniority: L3 ## Problem A content `feed` arrives as a list whose entries are grouped into sub-lists nested to arbitrary depth; collapse it into a single flat list of leaf values in their original left-to-right order. A leaf is any non-list value, so a string stays whole rather than being split into its characters, and empty sub-lists contribute nothing. ## Worked solution and explanation ### What you are really being asked Strip the feed costume off and this is depth-first leaf collection: walk the structure, and every value that is not itself a list is a leaf you keep, in the order you meet it. The recursion is the easy part. The thing that separates a clean pass from a quietly broken one is the type test. The tempting move is to ask 'can I iterate this?' and recurse when the answer is yes. That instinct detonates on the first string: 'tech' is perfectly iterable, so you descend into it and hand back its characters. Branch on `isinstance(item, list)` specifically and the trap never opens. --- ### Where it bites **isinstance(item, list)** Recurses only into real sub-lists. A string like 'tech' is not a list, so it lands in the result untouched. This is the behavior you want. **isinstance(item, Iterable)** Treats every iterable as nestable. 'tech' is iterable, so you recurse into it and get back 't', 'e', 'c', 'h'. The feed silently shreds every string leaf. --- ### Building it #### Step 1: Walk one level at a time Loop over the entries at the current depth. You do not need to know how deep the feed goes; recursion handles depth for you, so this loop only ever reasons about one level at a time. #### Step 2: Branch on list, never on iterable For each entry, ask only one question: is it a list? If yes, it is more structure to descend into. If no, it is a leaf (string included) and goes straight into the result. Widening this test to any-iterable is exactly how strings get shredded. #### Step 3: Splice children in place Recurse on a sub-list and extend the result with whatever comes back. Because you extend in iteration order, the depth-first left-to-right ordering falls out for free, with no sorting or post-processing needed. **Recursive depth-first flattening** ```python def flatten(feed): result = [] for item in feed: if isinstance(item, list): result.extend(flatten(item)) else: result.append(item) return result ``` > **Common pitfall** > > Reaching for `isinstance(item, Iterable)` or a try/except around `iter(item)`. Strings, bytes, and dict keys are all iterable, so any of these quietly recurses into a string and returns its characters. Test for `list` (or list and tuple if tuples count as structure) and nothing else. > **Time and space** > > Time is O(n) for n total leaves: each leaf is appended exactly once. Space is O(n) for the output plus O(d) for the call stack, where d is the deepest nesting. For pathological depth (tens of thousands of levels) you would hit Python's recursion limit and switch to an explicit stack. > **Interviewers watch for** > > Two tells. First, does the empty feed just work? It should: the loop never runs and you return []. No special case needed, and adding one signals you did not trust your own base case. Second, do you mention the recursion-depth ceiling before being asked? --- ## Common follow-up questions - How would you flatten iteratively using a stack? _(Tests replacing recursion with an explicit stack for O(1) call-stack usage.)_ - What if the input can contain dicts or tuples? _(Tests broadening the type check or using a more general iterable detection.)_ - What if you needed to preserve nesting depth as metadata? _(Tests returning (value, depth) pairs instead of raw values.)_ ## Related - [All practice problems](https://datadriven.io/problems) - [Mock interview mode](https://datadriven.io/interview/single_file) - [Python Interview Questions](https://datadriven.io/python-interview-questions) - [Data Engineering Interview Prep Guide](https://datadriven.io/data-engineer-interview-prep) - [Daily Challenge](https://datadriven.io/daily) --- Source: DataDriven (https://datadriven.io). 100% free data engineering interview prep. Live code execution against Postgres 16, Python 3.11, and Spark sandboxes. No paywall, no premium tier, no signup gate.