Practice Problems

Let the review queue choose the next problem, or select one when you know the skill you want to isolate. One rep should produce one useful correction, not a larger solve count.

Start in the editor

Start a comments-mode rep from Copilot Chat:

/comments

No labels or prefixes are required. To select a problem, add its topic and name, such as /comments graphs dijkstra. The direct terminal equivalent is:

just practice-start comments graphs dijkstra

If named vocabulary helps you think, /reacto, /clarp, and /umpire start the same loop with optional labels.

To study a named problem before a rep, open read-only snapshots of its committed solution and reference tests:

just practice-study topic problem

Study creates no rep or log. When you are ready, run the exact IMPLEMENT or TESTS_FIRST transition it emits. Either one starts a fresh isolated candidate pair.

Write your comments in the file, save, then use /continue or just practice-next for one next instruction. Then implement and add cases in your test file.

just practice-test
just practice-repl
just practice-finish "one fix"

Core 43

The core set covers common patterns across sessions. just catalog prints all editor-practice targets and marks each Core or Extra. Add natural names to search; ambiguous words such as anagram and prime keep every valid choice.

Topic	Problems
Arrays	two_sum, group_anagrams, product_except_self, top_k_frequent
Two pointers	three_sum, trapping_rain_water
Sliding window	min_window_substring, longest_substring_no_repeat
Stacks and queues	valid_parentheses, daily_temperatures
Searching	binary_search, search_rotated_array
Linked lists	reverse_linked_list, lru_cache
Trees	validate_bst, level_order_traversal, trie
Graphs	number_of_islands, topological_sort, course_schedule, dijkstra, a_star_search, bellman_ford, minimum_spanning_tree
Dynamic programming	coin_change, edit_distance, knapsack, longest_increasing_subseq, longest_common_subseq
Heaps	kth_largest, merge_k_sorted_lists
Backtracking	subsets, combination_sum, n_queens
Greedy	merge_intervals, jump_game
Strings	valid_palindrome, longest_palindromic_substring
Recursion	generate_parentheses, flatten_nested_list
Bit manipulation	single_number
Sorting	quickselect
Math	sieve_of_eratosthenes

Choose by the signal you want to improve: restating the problem, selecting an example, recognizing a pattern, implementing cleanly, testing, or explaining complexity. The decision tree helps with pattern selection. Inspect the algorithm library after the rep reaches a natural stopping point.

More practice

The repository includes additional problems in the same topic directories. Any entry from just catalog can start with just practice-start. Use the learning paths when you want a curated order, or extended problems for code reading and open-ended decomposition.

Untimed conversation and timed board-style reps can use the same catalog. The editor workspace is the default surface, not the only valid one.

Extended problems

Use these after ordinary editor reps when you want to practice reading unfamiliar code, finding operational risks, or decomposing an open-ended system. They supplement the problems above; they do not replace the comments, code, and tests loop.

Code reading

The Python files are intentionally buggy or inefficient. For each exercise:

1. Read signatures and docstrings.
2. Trace one execution path.
3. Test empty, large, malformed, and concurrent inputs where relevant.
4. Estimate the time and space cost of the hot path.
5. Propose two or three fixes, ordered by impact.

Exercise	File	Focus
Caching Service	src/practice/code_reading/ex01_caching_service.py	eviction, thread safety
Flight Data Pipeline	src/practice/code_reading/ex02_flight_data_pipeline.py	streaming, batching, failures
Rate Limiter	src/practice/code_reading/ex03_rate_limiter.py	token buckets, windows, concurrency

Check for unbounded memory, nested work on the hot path, missing validation, shared mutable state, blocking I/O, and calls inside loops. Explain the impact before proposing a change.

Problem decomposition

The open-ended prompts train scope, interfaces, tradeoffs, and test strategy.

Exercise	File	Focus
Vehicle Tracking	src/practice/decomposition/ex02_vehicle_tracking.md	geospatial indexing, live updates

State assumptions, split the system into parts, identify the riskiest seam, and describe how you would verify it. An untimed conversational rep works well for a first pass; use an observed mock when concise narration is the target.

For a longer scheduled block, run just practice-day 12 and follow the printed stop conditions. Use the daily conductor only when you intend to run the full block.