Module 2 — Design boundaries and algorithm evidence
EDU-SWE-201 · Module 2 of 4 · Estimated effort: 2 hours including practice
Before you begin: Complete Module 1 or demonstrate its evidence criteria. Use the course environment and downloads.
Outcomes
Separate policy from presentation; compare two deduplication strategies using an explicit workload; teach the difference between operation counts and elapsed time.
Explain and demonstrate
A small program can teach architecture without requiring a distributed system. Keep booking policy in a function, input conversion at the boundary and printing in the command-line interface. The function should not read from a keyboard or depend on a web request. This separation allows the same policy to be tested directly and reused in a later interface.
Consider deduplicating a list of submission identifiers while preserving the first occurrence. A novice solution scans the results already collected before adding each new value. It is easy to trace, but on a list of distinct values the comparisons grow as 0 + 1 + … + (n − 1). A set can support membership checks more efficiently on average, while a separate output list preserves order. The tradeoff includes additional memory and a requirement for hashable values. “Use a set” is therefore an incomplete explanation.
Worked example: an architecture decision record
| Concern | List scan | Set plus output list |
|---|---|---|
| First-occurrence order | Preserved | Preserved by the output list |
| Distinct-value membership work | Quadratic comparisons | Expected near-linear total membership work |
| Values | Equality-comparable objects | Hashable objects |
| Teaching purpose | Exposes the repeated scan | Exposes a time/memory tradeoff |
For four distinct items, the scan makes 6 comparisons. For eight, it makes 28. Doubling the input did not merely double the count. This is evidence about this algorithm and this workload. It is not a benchmark of a particular computer, and it does not establish the time taken by Python’s built-in operations.
Lab — explain a design tradeoff
Setup: Use the supplied offline lab.py and a text editor. Work in a new folder; no service or database is required.
- Run
python lab.py design. Expectn=4; comparisons=6andn=8; comparisons=28. - Trace
unique_scan([1, 2, 1])by hand. Record the result list after each iteration and the number of equality checks. - Add a function using a
seenset and a result list. Compare its returned values with the original on empty, duplicate and distinct inputs. - Explain why a list containing nested lists needs a different policy: nested lists are not hashable. Do not silently stringify arbitrary objects to make the error disappear.
- Write a short decision record with context, chosen option, rejected option and a trigger for revisiting the choice.
Expected observations: [1, 2, 1] produces [1, 2] with two comparisons. Both designs should preserve output order for supported inputs. Fault injection: Remove the membership check from the new function; show that the duplicate test detects the problem. Restore it before submission.
Evidence: Keep the trace table, test cases and decision record. Cleanup: Remove the defective variant; retain the working code and environment version.
Common pitfalls and teaching decisions
A stopwatch measurement on one small input is noisy and does not prove asymptotic complexity. Counting only successful matches misses unsuccessful comparisons. Mixing UI and policy makes a unit test unnecessarily depend on keyboard input. Ask learners to make a prediction first, observe evidence second and revise their explanation third.
Retrieval and reflection
Why can a faster membership structure still be the wrong choice?
It may require unsupported value types, extra memory or different semantics. State the constraints before choosing a structure.
What evidence would show that the refactor preserved behaviour?
The same contract-based cases should produce the same ordered results, including empty and duplicate inputs. Operation counts alone do not establish correctness.
Official references
Python data structures · Git tutorial for versioned changes. Examples and comparison counts are original Academy fixtures, not vendor benchmark results.