Challenges

Coding practice and systems · Build a Unix Shell

#	Challenge title	Domain	Difficulty
01	Build a Unix Shell PREQUELS 0/6 LEVELS 0/6	Command shell	Medium
PREQUELS (0 / 6)Focused coding problems · Solve in any order Start Prequels → Prequel 1Split a Line into Words· Strings, Tokenization Easy Prequel 2Find a Program's Path· Strings, Linear Search Easy Prequel 3Track Finished Tasks· Arrays, Strings Medium Prequel 4Split a Command into Stages· Arrays, Strings Medium Prequel 5Calculate Window Statistics· Sliding Window, Arrays Hard Prequel 6Top K Frequent Items· Arrays, Strings Hard LEVELS (0 / 6)Build a working system · One level at a time Start Level 1 → Level 1Builtin Command Execution (echo, pwd, cd, exit)· Strings, Parsing Easy Level 2PATH Resolution & Binary Execution (which, run, exec)· Strings, Linear Search Medium Level 3Process Lifecycle & Zombie Reaping (sigint, spawn-bg, reap)· Hash Maps, State Machines Medium Level 4Multi-Stage Pipelines & I/O Redirection (pipe, dup, >, >>)· Strings, Queues Hard Level 5Sliding-Window Latency Profiling (profile window)· Sliding Window, Arrays Medium Level 6Command History Frequency Ranking (hot k)· Hash Maps, Sorting Medium
02	Build an HTTP ServerLocked PREQUELS 0/6 LEVELS 0/6	Web server	Hard
03	Build a Git EngineLocked PREQUELS 0/6 LEVELS 0/6	Version control	Medium
04	Build a Key-Value StoreLocked PREQUELS 0/6 LEVELS 0/6	Key-value database	Hard
05	Build an Object StoreLocked PREQUELS 0/6 LEVELS 0/6	Object storage	Hard
06	Build an LRU CacheLocked PREQUELS 0/6 LEVELS 0/6	Memory cache	Medium
07	Build a Database IndexLocked PREQUELS 0/6 LEVELS 0/6	Database index	Hard
08	Build a Container RuntimeLocked PREQUELS 0/6 LEVELS 0/7	Container runtime	Hard
09	Build a Message QueueLocked PREQUELS 0/6 LEVELS 0/6	Message queue	Hard
10	Build a Log Analytics EngineLocked PREQUELS 0/6 LEVELS 0/6	Log processing	Medium
11	Build a Rate LimiterLocked PREQUELS 0/6 LEVELS 0/6	Rate limiting	Medium
12	Build a Load BalancerLocked PREQUELS 0/6 LEVELS 0/6	Load balancing	Hard
13	Build a Task SchedulerLocked PREQUELS 0/6 LEVELS 0/6	Task scheduling	Medium
14	Build a Raft Consensus ModelLocked PREQUELS 0/6 LEVELS 0/8	Distributed consensus	Hard
15	Build a Service RegistryLocked PREQUELS 0/6 LEVELS 0/7	Service discovery	Hard
16	Build a Distributed Object StoreLocked PREQUELS 0/6 LEVELS 0/6	Distributed storage	Hard
17	Build a Search EngineLocked PREQUELS 0/6 LEVELS 0/6	Search engine	Medium
18	Build a Vector DatabaseLocked PREQUELS 0/6 LEVELS 0/7	Vector database	Medium
19	Build an LLM Inference ModelLocked PREQUELS 0/6 LEVELS 0/6	AI inference	Medium
20	Build an MCP Tool RuntimeLocked PREQUELS 0/6 LEVELS 0/6	AI tool runtime	Medium

Showing 20 of 20 verified systems challenges

Python, C++, Rust, Go, Java Supported

Build a Unix Shell

5. Sliding-Window Latency Profiling (profile window)

Medium

A developer sees several recorded response times and asks: was the whole run slow, or was there one busy stretch? A total and average show the overall picture; a consecutive window helps locate a burst.

What you'll build

Add profile to summarize supplied timing samples. Report their count, total, whole-number average, and largest K-sample window. You analyze recorded numbers; you do not benchmark the shell.

Input: One or more profile K sample... command lines, optionally followed by exit. Samples are nonnegative integer microseconds; 1 <= K <= the number of samples.

Output: For each profile command, print SAMPLES:n TOTAL_US:t AVG_US:a PEAK_WINDOW_US:p, with single spaces between fields.

Example 1

Input:

profile 2 120 180 90 210 150

Output:

SAMPLES:5 TOTAL_US:750 AVG_US:150 PEAK_WINDOW_US:360

Explanation: The five samples total 750 and average 150. The two-sample sums are 300, 270, 300, and 360; the largest is 360.

Example 2

Input:

profile 4 50 60 70 80

Output:

SAMPLES:4 TOTAL_US:260 AVG_US:65 PEAK_WINDOW_US:260

Explanation: The window covers the entire four-sample list, so its sum equals the total: 260. The average is 65.

Example 3

Input:

profile 1 45 310 85 120

Output:

SAMPLES:4 TOTAL_US:560 AVG_US:140 PEAK_WINDOW_US:310

Explanation: A one-sample window asks for the largest individual sample: 310. The total is 560 and the average is 140.

Example 4

Input:

profile 3 100 100 100 100 100

Output:

SAMPLES:5 TOTAL_US:500 AVG_US:100 PEAK_WINDOW_US:300

Explanation: Every three-sample window totals 300. Tied windows still produce one peak value, not several reports.

Example 5

Input:

profile 2 10 15 12 400 450

Output:

SAMPLES:5 TOTAL_US:887 AVG_US:177 PEAK_WINDOW_US:850

Explanation: The last two samples form the largest window, 850. Whole-number division gives 887 divided by 5 as 177, not a decimal average.

Try a small change

Paste this into Custom Input and run your solution.

profile 2 10 20 30

If K changes to 1, which fields change?

Check your prediction

Expected output:

SAMPLES:3 TOTAL_US:60 AVG_US:20 PEAK_WINDOW_US:50

Only PEAK_WINDOW_US changes, to 30. The count, total, and average describe the whole list and do not depend on window size.

Constraints

n is the sample count, t is the sum of all samples, and a is t divided by n using whole-number division.
p is the largest sum of any K consecutive samples, not the largest individual sample unless K is 1.
Treat each profile line independently. Stop at exit. This summary does not time commands.

Hint

After summing the first window, can you update it by adding one value and subtracting one value?

Where This Skill Takes You

The same calculation appears in the window-statistics Prequel. Here you wrap it in a shell command and a labeled report. The timings are supplied data, not evidence of real shell performance.

What Comes Next?

Summarize a known command trace using counts and a deterministic tie rule.

Level 5 of 6

Code

5 executions per level

Loading ALGO Monaco Editor...

Input:Case 1: Sliding window finds the peak

Output:

SAMPLES:5 TOTAL_US:750 AVG_US:150 PEAK_WINDOW_US:360

Challenges

Coding practice and systems · Build a Unix Shell

#	Challenge title	Domain	Difficulty
01	Build a Unix Shell PREQUELS 0/6 LEVELS 0/6	Command shell	Medium
PREQUELS (0 / 6)Focused coding problems · Solve in any order Start Prequels → Prequel 1Split a Line into Words· Strings, Tokenization Easy Prequel 2Find a Program's Path· Strings, Linear Search Easy Prequel 3Track Finished Tasks· Arrays, Strings Medium Prequel 4Split a Command into Stages· Arrays, Strings Medium Prequel 5Calculate Window Statistics· Sliding Window, Arrays Hard Prequel 6Top K Frequent Items· Arrays, Strings Hard LEVELS (0 / 6)Build a working system · One level at a time Start Level 1 → Level 1Builtin Command Execution (echo, pwd, cd, exit)· Strings, Parsing Easy Level 2PATH Resolution & Binary Execution (which, run, exec)· Strings, Linear Search Medium Level 3Process Lifecycle & Zombie Reaping (sigint, spawn-bg, reap)· Hash Maps, State Machines Medium Level 4Multi-Stage Pipelines & I/O Redirection (pipe, dup, >, >>)· Strings, Queues Hard Level 5Sliding-Window Latency Profiling (profile window)· Sliding Window, Arrays Medium Level 6Command History Frequency Ranking (hot k)· Hash Maps, Sorting Medium
02	Build an HTTP ServerLocked PREQUELS 0/6 LEVELS 0/6	Web server	Hard
03	Build a Git EngineLocked PREQUELS 0/6 LEVELS 0/6	Version control	Medium
04	Build a Key-Value StoreLocked PREQUELS 0/6 LEVELS 0/6	Key-value database	Hard
05	Build an Object StoreLocked PREQUELS 0/6 LEVELS 0/6	Object storage	Hard
06	Build an LRU CacheLocked PREQUELS 0/6 LEVELS 0/6	Memory cache	Medium
07	Build a Database IndexLocked PREQUELS 0/6 LEVELS 0/6	Database index	Hard
08	Build a Container RuntimeLocked PREQUELS 0/6 LEVELS 0/7	Container runtime	Hard
09	Build a Message QueueLocked PREQUELS 0/6 LEVELS 0/6	Message queue	Hard
10	Build a Log Analytics EngineLocked PREQUELS 0/6 LEVELS 0/6	Log processing	Medium
11	Build a Rate LimiterLocked PREQUELS 0/6 LEVELS 0/6	Rate limiting	Medium
12	Build a Load BalancerLocked PREQUELS 0/6 LEVELS 0/6	Load balancing	Hard
13	Build a Task SchedulerLocked PREQUELS 0/6 LEVELS 0/6	Task scheduling	Medium
14	Build a Raft Consensus ModelLocked PREQUELS 0/6 LEVELS 0/8	Distributed consensus	Hard
15	Build a Service RegistryLocked PREQUELS 0/6 LEVELS 0/7	Service discovery	Hard
16	Build a Distributed Object StoreLocked PREQUELS 0/6 LEVELS 0/6	Distributed storage	Hard
17	Build a Search EngineLocked PREQUELS 0/6 LEVELS 0/6	Search engine	Medium
18	Build a Vector DatabaseLocked PREQUELS 0/6 LEVELS 0/7	Vector database	Medium
19	Build an LLM Inference ModelLocked PREQUELS 0/6 LEVELS 0/6	AI inference	Medium
20	Build an MCP Tool RuntimeLocked PREQUELS 0/6 LEVELS 0/6	AI tool runtime	Medium

Showing 20 of 20 verified systems challenges

Python, C++, Rust, Go, Java Supported

Build a Unix Shell

5. Sliding-Window Latency Profiling (profile window)

Medium

What you'll build

Add profile to summarize supplied timing samples. Report their count, total, whole-number average, and largest K-sample window. You analyze recorded numbers; you do not benchmark the shell.

Input: One or more profile K sample... command lines, optionally followed by exit. Samples are nonnegative integer microseconds; 1 <= K <= the number of samples.

Output: For each profile command, print SAMPLES:n TOTAL_US:t AVG_US:a PEAK_WINDOW_US:p, with single spaces between fields.

Example 1

Input:

profile 2 120 180 90 210 150

Output:

SAMPLES:5 TOTAL_US:750 AVG_US:150 PEAK_WINDOW_US:360

Explanation: The five samples total 750 and average 150. The two-sample sums are 300, 270, 300, and 360; the largest is 360.

Example 2

Input:

profile 4 50 60 70 80

Output:

SAMPLES:4 TOTAL_US:260 AVG_US:65 PEAK_WINDOW_US:260

Explanation: The window covers the entire four-sample list, so its sum equals the total: 260. The average is 65.

Example 3

Input:

profile 1 45 310 85 120

Output:

SAMPLES:4 TOTAL_US:560 AVG_US:140 PEAK_WINDOW_US:310

Explanation: A one-sample window asks for the largest individual sample: 310. The total is 560 and the average is 140.

Example 4

Input:

profile 3 100 100 100 100 100

Output:

SAMPLES:5 TOTAL_US:500 AVG_US:100 PEAK_WINDOW_US:300

Explanation: Every three-sample window totals 300. Tied windows still produce one peak value, not several reports.

Example 5

Input:

profile 2 10 15 12 400 450

Output:

SAMPLES:5 TOTAL_US:887 AVG_US:177 PEAK_WINDOW_US:850

Explanation: The last two samples form the largest window, 850. Whole-number division gives 887 divided by 5 as 177, not a decimal average.

Try a small change

Paste this into Custom Input and run your solution.

profile 2 10 20 30

If K changes to 1, which fields change?

Check your prediction

Expected output:

SAMPLES:3 TOTAL_US:60 AVG_US:20 PEAK_WINDOW_US:50

Only PEAK_WINDOW_US changes, to 30. The count, total, and average describe the whole list and do not depend on window size.

Constraints

n is the sample count, t is the sum of all samples, and a is t divided by n using whole-number division.
p is the largest sum of any K consecutive samples, not the largest individual sample unless K is 1.
Treat each profile line independently. Stop at exit. This summary does not time commands.

Hint

After summing the first window, can you update it by adding one value and subtracting one value?

Where This Skill Takes You

The same calculation appears in the window-statistics Prequel. Here you wrap it in a shell command and a labeled report. The timings are supplied data, not evidence of real shell performance.

What Comes Next?

Summarize a known command trace using counts and a deterministic tie rule.

Level 5 of 6

Code

5 executions per level

Loading ALGO Monaco Editor...

Input:Case 1: Sliding window finds the peak

Output:

SAMPLES:5 TOTAL_US:750 AVG_US:150 PEAK_WINDOW_US:360