DS Exercises · DS_part3

The advanced-introductory band: everything runs over real (small) text datasets, ends in a mini capstone, and stays inside course-taught C++.

PF-DS-13 · Two-Column File Statistics (CSV)

Level: Intermediate · Lectures: L29–L30 · Prerequisites: DS-09 Problem: File scores.csv holds name,score lines (no spaces in names; score 0..100). Print loaded, then the mean (2 dp), the max with its owner (first owner wins ties), and the count of skipped lines with reasons split into no-comma and bad-score. Formula first: mean over loaded only; skip classification is a decision tree: (no comma) → no-comma; (comma but non-numeric / out of range) → bad-score. Edge cases: empty file; file with header line name,score — treat it as a data line and let it classify as bad-score (document the choice; a header-aware version is an extension). Dataset: datasets/scores.csv (11 lines incl. the name,score header: 7 loaded, 1 no-comma, 3 bad-score — the header counts as bad-score per the stated rule; mean of the 7 valid 88, 72, 95, 61, 79, 90, 100 is 83.57; max 100 owned by Jade). Hints: find(',') on each line; substr both fields; istringstream the score.

PF-DS-14 · Group Means by Key (file)

Level: Advanced · Lectures: L28–L30 · Prerequisites: DS-13 Problem: File students.txt holds name age gpa lines. Print mean age and mean GPA overall, then for each distinct name (first-appearance order): <name>: <gpa-mean> (<k> rows) 2 dp. Formula first: mean(c,k) = Σ gpa where name = k / rows(k) — a two-pass group-by without maps: pass 1 discovers keys (first appearance), pass 2 accumulates each group (the ds_tabular_records.cpp pattern). Edge cases: a single-row group; all rows one group; out-of-range rows (age 16..90, gpa 0..4) rejected and counted before grouping. Dataset: datasets/students.txt (12 lines: 9 valid, 1 bad age, 1 bad gpa, 1 non-numeric gpa; Ana appears 3× with gpa mean 3.63; overall mean age 20.89, mean gpa 3.37). Hints: struct Student { string name; int age; double gpa; }; + std::vector<Student>; a parallel bool grouped[] marks joined rows.

PF-DS-15 · Data Quality Audit (file)

Level: Advanced · Lecture: L30 · Prerequisites: DS-09, DS-13 Problem: File messy_grades.txt holds lines that are one of: name score (valid), name NA (missing), name <junk> (corrupt), or garbage with no name. Produce an audit: counts per class, then for valid rows: min, max, mean, median (2 dp), and the percentage of the file that was usable, usable: <p>% (1 dp). Formula first: usable% = 100 · loaded / total_lines (blank lines excluded from both); median by the sorted rule from DS-05. Edge cases: zero valid rows (statistics n/a, audit still prints, exit 0); a line that is only whitespace; duplicate names (fine — grades may repeat). Dataset: datasets/messy_grades.txt (20 lines: 13 valid, 2 missing, 2 corrupt values, 3 no-name lines (bare scores or lone NA — classify them as corrupt with a reason, your table must state the rule); mean 74.92, median 79.00, usable 65.0 %, min 41, max 100). Hints: build on DS-09's classifier, then add statistics; keep classes separate in the report — merging them destroys the audit's value.

PF-DS-16 · Moving Average

Level: Advanced · Lectures: L17–L18 · Prerequisites: DS-08 Problem: Read n, window w (1 ≤ w ≤ n), then n doubles. Print the w-day moving averages, one per line, 2 dp — there are exactly n − w + 1 of them. Formula first: MA(i) = (x(i) + x(i+1) + … + x(i+w−1)) / w for i = 1 .. n−w+1 (1-based). Naive version: re-sum each window (O(n·w)) — fine at this scale; say so in a comment. Edge cases: w = n (exactly one average = the overall mean); w = 1 (the data echoed back, formatted). Sample: 5 3 2 4 6 8 10 → 4.00 / 6.00 / 8.00 Hints: one outer loop over window starts, one inner loop over the window; a running sum version (add the entering value, drop the leaving one) is the extension — implement naive first, then compare outputs.

PF-DS-17 · Reproducible Simulation

Level: Advanced · Lectures: L10/L18 · Prerequisites: DS-06 Problem: Simulate n rolls of a fair six-sided die with a FIXED seed (42), std::mt19937, and uniform_int_distribution<int> die{1,6}. Print per-face counts and empirical proportions (4 dp) plus the mean of faces shown, then the same table for n trials of TWO dice summed, sum2 2..12 (13 buckets, index shift). Formula first: expected p = 1/6 per face; for two dice, expected counts follow the triangle 1,2,3,4,5,6,5,4,3,2,1 over sums 2..12 — print your empirical proportions beside k/36 computed from the triangle. Edge cases: n = 0 (no trials, proportions printed as 0.0000 with a note); large n (long long counts). Sample behavior: with n = 60000, every face proportion lands within ±0.01 of 0.1667 and the two-dice peak sits at sum 7 — verify and state both in the test table. Same seed → identical output every run: reproducibility is the point; say why in a comment. Hints: exactly the ds_simulation.cpp pattern extended; no other randomness sources allowed (no rand()).

PF-DS-18 · Mini Capstone: Station Report

Level: Advanced Introductory · Lectures: L24–L30 · Prerequisites: DS-14, DS-15 Problem: Files stations/station_*.txt (see datasets/stations/) each hold day temperature lines. Build one report across ALL station files (argument: directory-free version — filenames listed in stations/index.txt, one per line): per station (loaded/missing/invalid + mean), then across all stations: the hottest single reading with station and day, the coldest likewise, and each station's mean as a ranked list (hottest mean first; stable ties by name). Formula first: means per group (DS-14); ranking = sort an array of struct StationMean { string name; double mean; } with your own sort (descending by mean; equal means keep alphabetical name order — a stated, implemented tie rule). Edge cases: a station file entirely NA (mean n/a — still ranked? No: excluded from the ranking, listed in an excluded: line); an index entry whose file cannot open → warning line, exit code still 0 if at least one station loaded, else 2. Dataset: datasets/stations/ — three stations, ~10 days each, seeded defects; expected headline values are in the datasets README for self-checking. Hints: this composes DS-09 + DS-11 + DS-14; keep each piece a function with a contract comment — the report is the product, the functions are the pipeline.