PART 5 · CODING ROUND PATTERNS

The pattern map: 18 patterns at a glance

FreeLesson 56 of 78

Most DE Python coding questions are one of these 18 shapes, or two of them stacked. Learn to hear the shape inside the question.

#	Pattern	The question sounds like	Core tool	Pg
P1	Frequency count	"how many of each", "most common error"	`Counter, dict.get(k, 0) + 1`	59
P2	Top-K	"top 3 customers", "k largest files"	`heapq.nlargest, size-k min-heap`	59
P3	Group by key	"totals per order", "items per user"	`defaultdict(list)`	60
P4	Dedupe, keep latest	"remove duplicates", "latest record per id"	`dict keyed by id + timestamp test`	60
P5	Sliding window	"rolling 3-day sum", "moving average"	`deque, running total`	61
P6	Merge sorted streams	"merge two sorted logs", "combine feeds"	`heapq.merge, two pointers`	61
P7	Flatten nested JSON	"flatten this payload", "explode items"	`recursion, json_normalize`	62
P8	Sessionize events	"split into sessions", "30 min idle"	`sort + gap test`	63
P9	Consecutive streaks	"longest login streak", "consecutive days"	`set + start-of-run test`	63
P10	Parse logs	"count errors per job from this log"	`re named groups + dict`	64
P11	Merge intervals	"overlapping bookings", "busy periods"	`sort by start, extend last`	65
P12	Hash join	"join these two files without pandas"	`dict on the small side`	65
P13	Chunked processing	"the file is 50 GB", "batch inserts"	`generator + islice`	66
P14	Retry with backoff	"the API returns 429", "flaky warehouse"	`exponential delay + jitter`	66
P15	Paginated extraction	"pull all pages from this API"	`cursor loop + generator`	67
P16	Memoize / LRU cache	"avoid repeat lookups", "build an LRU"	`lru_cache, OrderedDict`	67
P17	Validate and quarantine	"reject bad rows", "data contract"	`per-row rules, two outputs`	68
P18	Diff two snapshots	"what changed since yesterday", "CDC"	`set algebra on keys`	68

One skeleton covers half the list

THE ONE-PASS DICT SKELETONANY PYTHON

result = {}                              # or defaultdict(...) / Counter()
for record in stream:                    # one pass over the input: O(n)
    key = key_of(record)                 # what defines "the same thing"?
    result[key] = combine(result.get(key), record)   # count, sum, append, keep latest
answer = finish(result)                  # sort, take top k, format

Practice before you move on

For each of P1 to P6, write the phrase from a question that tells you which pattern it is.
Use the one pass dict skeleton to total the amounts per status in orders.csv.