Final Exam Model Paper 8

• i.Everything after the ?: district=Mugu&service=vaccines. They are name=value pairs joined by &: district with the value Mugu, and service with the value vaccines. They are the request’s arguments, like a function’s, and the API’s documentation lists the names it understands.
• ii.The same address, with the values changed:

https://api.health.example.np/posts?district=Humla&service=dental

• i.Many routers keep a graph of the network: routers are nodes, links are edges, with a cost on each link. From it the router works out the best route to the destination (with Dijkstra’s algorithm, nearest first by cost) and sends the data to the first router on that route, the best next hop.
• ii.The cut cable is an edge removed from the graph. The routers learn that the link is gone, work out the best routes again on the changed graph, and send the data by another path, if one exists. A network with more than one route between places keeps working when one link fails.

SELECT doctors.name, visits.patientFROM visitsJOIN doctors ON visits.doctor_id = doctors.doctor_idWHERE visits.day = '2026-10-12'ORDER BY doctors.name;

The JOIN puts each visit beside its doctor’s row, matched by doctor_id; WHERE keeps that day’s visits.

i.

import pandas as pd bp = pd.read_csv("bp.csv")high = bp[bp["systolic"] >= 140]print(high)print(len(high), "of", len(bp), "need a check-up")

ii. The comparison gives True or False for each row, and the outer brackets keep the rows that are True, with their index numbers from the file’s order:

   patient  age  systolic1        2   62       1513        4   70       1655        6   49       1427        8   58       1404 of 8 need a check-up

Patient 8, at exactly 140, is included because the rule is >=; patient 5, at 139, is not.

• i.Any two: save, save as and export (a report saved as a PDF); undo (Ctrl + Z), so he can try things freely; select, then act; the menus and the search for commands; help (F1); shortcuts such as Ctrl + C and Ctrl + V. Most programs share these ideas, so he only has to learn the two or three of the spreadsheet’s own, such as cells and formulas.
• ii.“Learn the program” is too big a job to know when it is done, and watching someone else feels like learning but isn’t. He should start with a small, real task (this month’s patients by ward, with a total), read the built-in help, follow a tutorial doing each step on a copy of the file, and check what he learned by doing it again a few days later without help.

i. The program’s language and the exact error, in quotes:

python "PermissionError: [Errno 13] Permission denied"

ii. Her folders and file names (C:\Users\Kabita\Desktop\clinic\report.py, patients.xlsx) and line 12: nobody else has them, so they only hide the pages that would help, and the path shows her name. (The usual answer she will find: the file is open in another program, such as a spreadsheet, or she isn’t allowed to write there.)

• Single letters: the vocabulary would be tiny, but every sentence would be very long in tokens, and the model must look back over everything written so far at each step, so every step would be much more work; and a single letter means almost nothing on its own.
• Whole words: each token would mean something, but the vocabulary would be enormous (गर्छ, गर्छन्, गरेको, गर्दै…, every name, every misspelling, every word of every language), and a word not in the list, such as a new village’s name, couldn’t be read at all.

Real tokenizers do something in between: common words are one token each, and rarer or longer words are cut into a few pieces that appear in other words too, so any word can still be read, at worst letter by letter.

The test doesn’t show that. Turing’s game measures whether a machine can pass for a person in a short written chat, which is not the same as thinking or understanding: ELIZA, a few rules that turned people’s words back as questions, made people feel understood. People disagree about whether a model understands (Searle’s Chinese room says rule-following isn’t understanding; others say predicting this well needs some kind of it), and it is very different from us: no body or goals of its own, and parameters frozen after training. A fair claim: “It writes in a way people can’t tell from a person, in five minutes.”

• An Act is passed or amended by the Federal Parliament: the law itself.
• Rules, directives and guidelines are made under an Act by the government, a ministry or a regulator (such as Nepal Rastra Bank or the Nepal Telecommunications Authority). They fill in the details and are quicker to change, but must stay within what the Act allows.

It is a trap. Suresh would be a money mule: someone whose account is used to pass on money, usually the money of scams, theft or other crime, so that the criminals stay hidden. The bank and the police trace the money to his account, under his name: he could have the account frozen and be charged, even though he “only passed it on”. Honest businesses don’t need a stranger’s account. He should refuse, not reply, and tell his bank or the police if he has already received money.

• i.Any two: that it is widely used and looked after (a recent version, documentation, problems answered), or its mistakes won’t be fixed; that it is the real one, its name copied letter by letter from the official page, since harmful packages use names one letter off and installing runs their code; that its license allows their use; and whether they need it at all, since BMI is one line they could write and test themselves.
• ii.a and c work. b gives a TypeError: the * in the signature means everything after it must be given by name, so 2 can’t be passed as a third value.

iii.

from bmi_tools import bmi       # the library’s module patients = [("Sita", 52, 1.55), ("Ram", 70, 0)]for name, weight, height in patients:    try:        print(name, bmi(weight, height))    except ValueError:        print("Check " + name + "’s height")

It prints Sita 21.6 (52 ÷ 1.55² = 21.64…) and then Check Ram’s height.
iv. So that a call says what each number is: bmi(60, 1.6, digits=2) can’t be mistaken, while bmi(60, 1.6, 2) could be read as a third measurement. It also lets the writer add or reorder options later without breaking anyone’s calls.

i.

• ii.4 trails. Kimathanka was first reached from Hatiya, Hatiya from Bala, Bala from Chainpur, Chainpur from Khandbari: Khandbari → Chainpur → Bala → Hatiya → Kimathanka.
• iii.Hatiya and Seduwa, the third ring.

i. Each post’s quantities added up, the biggest first:

ii. The tablets with fewer than 250 left, the fewest first (Paracetamol at Gamgadhi and Iron have more):

• iii.The rows with no expiry date recorded: stock 5 (Pulu) and stock 7 (Gamgadhi).
• iv.NULL means “no value, unknown”, so nothing is equal to it, not even another NULL: expires = NULL is never true, for any row. Empty fields must be found with IS NULL.

• i.Line 10 moves the clock to when the next person arrives: the gap is random, about 3 minutes on average for 20 people an hour (60 ÷ 20). Line 11 stops when the time passes 240 minutes, 14:00: no more new people.
• ii.Line 13 picks the desk free first (the first one when they tie), and line 14 starts the person at the later of arriving and the desk being free.

(With one desk, the second person would have waited 5.7 − 4.0 = 1.7 minutes.)

• iii.Line 16: random.uniform(2, 4) never gives more than 4 minutes, so the slow registrations are left out of the model. Every slow person holds up everyone behind them, so the real waits, the longest ones above all, would be longer than the simulation says, and the organizers might set up too few desks. A simulation is only as good as its model: they should time real registrations at a camp and use those times.
• iv.One desk can register about one person every 3 minutes (2 to 4, 3 on average), 20 an hour, and 20 an hour arrive: the desk is busy almost all the time, so whenever several people come close together a line builds up, and the desk never has spare time to clear it. A second desk can serve about 40 an hour, so lines clear quickly.

• i.Replication. If one copy is lost (a failed disk, a server down), another copy already has everything, so the records aren’t lost and the system keeps running.
• ii.Copies help only if they don’t fail together: a power cut, fire, flood or earthquake in that building takes both servers. A zone is a separate building, a few kilometers away, with its own power, cooling and network (a region is a different city or country).
• iii.A backup is a copy kept apart, from an earlier time. Replication copies every change, the mistakes too: if someone deletes records by accident (or malware scrambles them), every copy loses them within seconds. Only a backup from before the mistake still has them.
• iv.A change sent to every server at once can break every copy at once, since they all run the same program. A staged rollout (a few servers first, watched, then the rest) limits the harm, and a rollback (a way to undo the change) brings it back quickly.

"""When each token is seen at the eye camp.""" START = 10 * 60          # 10:00, in minutes after midnightEACH = 6                 # minutes for each patientLUNCH = 13 * 60          # lunch starts at 13:00LUNCH_LENGTH = 30        # minutes  def seen_at(token):    """The time token number `token` is seen, as text such as "10:24".    Raises ValueError for a token below 1."""    if token < 1:        raise ValueError("tokens start at 1")    minutes = START + (token - 1) * EACH    if minutes >= LUNCH:        minutes = minutes + LUNCH_LENGTH    return f"{minutes // 60}:{minutes % 60:02}"  if __name__ == "__main__":    # tests: the first token, and either side of lunch    print(seen_at(1), "should be 10:00")    print(seen_at(30), "should be 12:54")    print(seen_at(31), "should be 13:30")

Run by itself, camp.py prints its tests:

10:00 should be 10:0012:54 should be 12:5413:30 should be 13:30

import camp for i in range(3):    token = int(input("Token number? "))    print("Token", token, "is seen at", camp.seen_at(token))

Token number? 5Token 5 is seen at 10:24Token number? 31Token 31 is seen at 13:30Token number? 50Token 50 is seen at 15:24

Token 50: 49 × 6 = 294 minutes after 10:00 is 14:54, plus 30 for lunch, 15:24.

The same program, with these three lines in place of the last one:

import sqlite3 db = sqlite3.connect(":memory:")db.execute("CREATE TABLE donors (donor_id INTEGER PRIMARY KEY, name TEXT, blood_group TEXT, ward INTEGER)")db.execute("""CREATE TABLE donations (donation_id INTEGER PRIMARY KEY,              donor_id INTEGER REFERENCES donors(donor_id),              day TEXT, ml INTEGER)""")db.executemany("INSERT INTO donors VALUES (?, ?, ?, ?)", [    (1, "Anil Rai", "O+", 4),    (2, "Sarita Limbu", "A+", 7),    (3, "Bikash Shrestha", "B+", 4),    (4, "Puja Tamang", "O−", 12),    (5, "Rajesh Yadav", "AB+", 7),    (6, "Nirmala Rai", "O+", 9),])db.executemany("INSERT INTO donations VALUES (?, ?, ?, ?)", [    (1, 1, "2026-01-15", 450),    (2, 2, "2026-01-15", 350),    (3, 3, "2026-04-20", 450),    (4, 1, "2026-05-02", 450),    (5, 4, "2026-05-02", 450),    (6, 6, "2026-08-11", 350),    (7, 1, "2026-09-05", 450),    (8, 2, "2026-09-05", 450),    (9, 5, "2026-09-05", 350),])  def ask(sql):    """Print each row the query gives, then a blank line."""    for row in db.execute(sql):        print(row)    print()  ask("SELECT donation_id, donor_id, day FROM donations WHERE ml = 350 ORDER BY day DESC")ask("SELECT ward, COUNT(*) FROM donors GROUP BY ward ORDER BY ward")ask("""SELECT donors.name, donors.blood_group FROM donations       JOIN donors ON donations.donor_id = donors.donor_id       WHERE donations.day = '2026-09-05' ORDER BY donors.name""")

(9, 5, '2026-09-05')(6, 6, '2026-08-11')(2, 2, '2026-01-15') (4, 2)(7, 2)(9, 1)(12, 1) ('Anil Rai', 'O+')('Rajesh Yadav', 'AB+')('Sarita Limbu', 'A+')

Dates written year-month-day sort correctly as text, so ORDER BY day DESC puts the latest first. (The donors are made up for this paper.)

links = [("Chatara", "Barahakshetra"), ("Barahakshetra", "Prakashpur"),         ("Prakashpur", "Inaruwa"), ("Chatara", "Dharan"), ("Dharan", "Itahari"),         ("Itahari", "Inaruwa"), ("Inaruwa", "Rajbiraj"),         ("Prakashpur", "Hanumannagar"), ("Hanumannagar", "Rajbiraj")]  def add_link(graph, a, b):    """Add a two-way link between a and b."""    if a not in graph:        graph[a] = []    if b not in graph:        graph[b] = []    graph[a].append(b)    graph[b].append(a)  def fewest_hops(graph, start, goal):    """The stations on a route with the fewest hops, or None."""    came_from = {start: None}    to_visit = [start]    while len(to_visit) > 0:        station = to_visit.pop(0)        if station == goal:            route = []            while station is not None:                route.append(station)                station = came_from[station]            route.reverse()            return route        for nxt in graph[station]:            if nxt not in came_from:                came_from[nxt] = station                to_visit.append(nxt)    return None  radio = {}for a, b in links:    add_link(radio, a, b) for start, goal in [("Chatara", "Rajbiraj"), ("Dharan", "Hanumannagar")]:    route = fewest_hops(radio, start, goal)    print(" → ".join(route), "(" + str(len(route) - 1), "hops)") add_link(radio, "Biratnagar", "Itahari")print(" → ".join(fewest_hops(radio, "Chatara", "Biratnagar")))

Chatara → Barahakshetra → Prakashpur → Inaruwa → Rajbiraj (4 hops)Dharan → Chatara → Barahakshetra → Prakashpur → Hanumannagar (4 hops)Chatara → Dharan → Itahari → Biratnagar

Chatara to Rajbiraj by Hanumannagar, and Dharan to Hanumannagar by Itahari and Inaruwa, are also 4 hops; any route with the fewest hops is right.

• i.minutes % 60 is the minutes past the hour (the remainder after the whole hours), and :02 writes it with at least two digits, adding a 0 in front when needed. Token 2 is seen at 10:06; without :02 it would be shown as 10:6, which reads like 10:60 or is easily misread.
• ii.After a JOIN, the result has the columns of both tables, and both have a column called donor_id, so the table’s name says which one is meant. With a bare donor_id, SQLite can’t tell and stops with an error: ambiguous column name: donor_id. (name is only in donors, so donors.name isn’t required, but it makes the query clearer.)
• iii.Breadth-first search adds each station’s neighbors in the order of its list, and the first note made in came_from is kept. Chatara’s list starts with Barahakshetra (its link came first), so Prakashpur is added before Itahari, and Prakashpur’s list has Inaruwa before Hanumannagar, so Inaruwa is visited first and Rajbiraj is noted as reached from it. Giving the links in another order (for example Prakashpur–Hanumannagar before Prakashpur–Inaruwa) changes the lists’ order, and the function would print another route, still of 4 hops.