Final Exam Model Paper 9

• i.forecast is a dictionary (a JSON object), and forecast["days"] is a list of dictionaries, one for each day (a JSON array).
• ii.The place and the number of days, the two days over 40 km/h, then the last day (index −1) and its rain:

Jomsom 3Mon windy: 52 km/hTue windy: 45 km/hTue 0

About 2 seconds. Each village is visited once (it is marked as seen when it is first added, and never added again), and each village’s list of neighbors is read once, so every trail is looked at a fixed number of times. The work grows in step with the number of villages and trails, linear growth: twice the map, about twice the work.

SELECT AVG(mm) FROM rainWHERE station = 'Pokhara' AND day BETWEEN '2026-07-01' AND '2026-07-15'; SELECT day, mm FROM rainWHERE station = 'Pokhara' AND day BETWEEN '2026-07-01' AND '2026-07-15' AND mm > 100ORDER BY mm DESC;

BETWEEN includes both ends, and dates written year-month-day sort in the right order as text.

i.

import pandas as pd rain = pd.read_csv("pokhara_shrawan.csv")print(rain.sort_values("mm", ascending=False).head(3))

   day     mm5    6  151.01    2  118.04    5   96.5

ii. They are the table’s index, the place each row had in the file, counted from 0: day 6 was row 5. sort_values gives back the same rows in a new order, and each row keeps its own index, so they no longer go 0, 1, 2.

• i.Nobody can see her computer, and her question tells them nothing: not what she was doing, what went wrong or what she tried. Almost any cause fits, so a helper would have to ask many questions before even starting, and most people pass over a question that is hard to answer.
• ii.What she was trying to do and what she expected; what happened instead, with the exact error message (copied, or a screenshot with personal details covered); what she has already tried, and what each try did; and her device, operating system and the program’s version. (Often, writing all this makes her see the answer herself.)

• i.The Recycle Bin (4 GB: files already deleted still take space until it is emptied) and the Downloads (9 GB: installers can be downloaded again, and the film isn’t the office’s work). 4 + 9 = 13 GB.
• ii.Never delete the only copy. Copy them to another disk or the cloud, check that the copy opens, and only then think about removing them from the laptop (better, keep a copy in two places).

• i.It sees only “two villages”. After those words, “are” is the commoner next word in its training text, and the word that decides, “road”, is five words back, out of its sight. Each step is likely, so the sentence sounds right but is wrong.
• ii.A transformer uses attention: for each new token it looks back over all the tokens in its context window and weighs those that matter most. It has learned to link the place of the verb to its subject, “road” (as it learned to link “its” to the noun it refers to), so it writes “is”.

The note doesn’t make the plan safe. The bank chooses to put the chatbot on its own website to answer customers, so what it says is information from the bank, like any other page there. In the Air Canada case (2024), the airline argued that its chatbot was responsible for its own words; the tribunal held the airline responsible for all the information on its website, chatbot included. And a customer can’t tell which answers are the wrong ones, while a wrong answer about fees costs real money. Instead, the bank should have the chatbot answer from its official fee pages (grounding), test it before use with real customers’ questions whose answers are checked against its rules, show the official fees with each answer, and honor or put right its mistakes.

OpenStreetMap is a free map of the world that anyone can edit, like Wikipedia. After the earthquake of 25 April 2015, Kathmandu Living Labs and the Humanitarian OpenStreetMap Team split the affected areas into small squares on a website; thousands of volunteers, in Nepal and around the world, each took a square and traced roads and buildings from satellite photos, and mappers in Nepal added names, paths and blocked roads. Aid teams used the maps to reach damaged villages.

It is a poor decision. A signed 32-bit count of seconds since 1970 overflows in January 2038, and the logger will then think it is 1901, so its dates, and anything worked out from them (how long a turbine ran, when a reading was taken), go wrong. Small computers in machines are often left running for decades, so the loggers installed now will very likely still be in use in 2038, and finding every one later, inside a dam, will cost far more than choosing devices that store time in 64 bits today. The Year 2000 problem showed that programs and devices outlive everyone’s expectations: design for longer than you think it will last.

i. In order the values are 0, 4, 10, 30; the two middle ones are 4 and 10, and their mean is 7.0 (the median sorts the values itself).

ii.

"""Summaries of daily rainfall."""import statistics HEAVY = 50      # mm in a day, or more, counts as a heavy day  def summary(daily):    """Given a list of daily rain in mm, return (mean rounded to 1 decimal,    median, number of heavy days), or None if the list is empty."""    try:        mean = statistics.mean(daily)    except statistics.StatisticsError:        return None    heavy = 0    for mm in daily:        if mm >= HEAVY:            heavy = heavy + 1    return round(mean, 1), statistics.median(daily), heavy  if __name__ == "__main__":    print(summary([]), "should be None")    print(summary([10, 50, 60]), "should be (40, 50, 2)")

iii.

import rainfall result = rainfall.summary([12.5, 0, 64.0, 30.5, 0, 8.0, 51.5])if result is None:    print("No readings")else:    mean, median, heavy = result    print("Mean", mean, "mm, median", median, "mm,", heavy, "heavy days")

Mean 23.8 mm, median 12.5 mm, 2 heavy days

The 7 days add up to 166.5 mm, a mean of 23.79; in order the middle value is 12.5; 64.0 and 51.5 are heavy.

i. Each row’s 1s, in order:

links = {    "Kathmandu": ["Hetauda", "Damauli"],    "Hetauda": ["Kathmandu", "Bharatpur"],    "Bharatpur": ["Hetauda", "Damauli", "Butwal"],    "Damauli": ["Kathmandu", "Bharatpur", "Pokhara"],    "Pokhara": ["Damauli", "Butwal"],    "Butwal": ["Bharatpur", "Pokhara"],}

ii. Each station is noted when it is added, so none is added twice:

Butwal is 3 hops away: Butwal → Bharatpur → Hetauda → Kathmandu. (Butwal → Bharatpur → Damauli → Kathmandu, or Butwal → Pokhara → Damauli → Kathmandu, are 3 hops too.)
iii. The table would have 2,000 × 2,000 = 4 million cells, almost all 0, since each row has only about three 1s, and finding a gauge’s neighbors would mean reading a whole row of 2,000. A dictionary stores only the links that exist (about 3 names for each gauge) and finds a gauge’s neighbors in one lookup.

• i.The stations above 2,000 m, the highest first: Jomsom, then Jumla.
• ii.Each station’s hottest reading:

• iii.The stations with more than 10 mm of rain on 11 July: Pokhara 112.5 and Biratnagar 12.0 (Jumla had 0).
• iv.Dates as text are compared character by character from the left. With the day first, they would sort by the day of the month: 02/08/2026 (2 August) would come before 11/07/2026 (11 July). And BETWEEN '01/07/2026' AND '15/07/2026' would also catch 05/08/2026, 5 August, since “05/08” falls between “01/07” and “15/07” as text. Written year-month-day, the biggest part comes first, so the order of the text is the order of the dates.

• i.Only how the July rain at one gauge, Kathmandu airport, changed over these ten years. It doesn’t say why, and it says nothing about the rest of Nepal.
• ii.From one rain gauge in one place. Rain in Nepal differs enormously from place to place (Pokhara gets several times as much as Mustang), and one gauge can’t speak for all of Nepal.
• iii.Not very: it has only 10 values, one a year, and the rain in one month changes a lot from year to year by chance, so a line that goes down over ten years could be chance. Many more years, and the spread from year to year, would be needed.
• iv.Any one: the other months of the monsoon (June to September); every other station; anything earlier than 2016. Honest: “July rain at Kathmandu airport was lower in recent years; one gauge and ten years can’t show whether Nepal’s monsoon is changing.”

• i.Almost all the electricity a computer uses ends up as heat, which must be taken away or the computers overheat, so the data center uses electricity again for cooling (and for lights, batteries and losses in its wiring).
• ii.Power usage effectiveness: the electricity used by the whole data center divided by the electricity used by its computers. Here 15 ÷ 10 = 1.5: for every 100 units the computers use, 50 more go to cooling and the rest.
• iii.Cool outside air or cold water can take the heat away instead of air conditioning, which uses much less electricity (a lower PUE).
• iv.Electricity from coal puts out a great deal of CO₂ (and smoke for the people nearby); hydropower puts out very little. The same number of units can do very different harm depending on how they were made.

"""Monthly rain totals from weekly readings."""  def monthly_totals(weekly):    """Given (month, mm) pairs, return a dictionary from each month    to its total mm, in the order the months first appear."""    totals = {}    for month, mm in weekly:        if month not in totals:            totals[month] = 0        totals[month] = totals[month] + mm    return totals  if __name__ == "__main__":    print(monthly_totals([]), "should be {}")    print(monthly_totals([("Asar", 10), ("Asar", 5), ("Shrawan", 2)]), "should be Asar 15, Shrawan 2")

Run by itself, monsoon.py prints its tests:

{} should be {}{'Asar': 15, 'Shrawan': 2} should be Asar 15, Shrawan 2

import monsoon weekly = [("Jestha", 120.5), ("Asar", 310.0), ("Asar", 295.5), ("Shrawan", 402.0),          ("Shrawan", 380.5), ("Bhadra", 260.0), ("Bhadra", 190.0), ("Jestha", 88.0)] totals = monsoon.monthly_totals(weekly)for month in totals:    print(month, totals[month], "mm") print("Over 500 mm:")for month in totals:    if totals[month] > 500:        print(month)

Jestha 208.5 mmAsar 605.5 mmShrawan 782.5 mmBhadra 450.0 mmOver 500 mm:AsarShrawan

(The readings are made up for this paper.)

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

import sqlite3 db = sqlite3.connect(":memory:")db.execute("CREATE TABLE municipalities (muni_id INTEGER PRIMARY KEY, name TEXT, district TEXT)")db.execute("""CREATE TABLE landslides (slide_id INTEGER PRIMARY KEY,              muni_id INTEGER REFERENCES municipalities(muni_id),              day TEXT, houses INTEGER, road_blocked INTEGER)""")db.executemany("INSERT INTO municipalities VALUES (?, ?, ?)", [    (1, "Jugal", "Sindhupalchok"),    (2, "Melamchi", "Sindhupalchok"),    (3, "Bhimeshwar", "Dolakha"),    (4, "Kalinchok", "Dolakha"),    (5, "Sunkoshi", "Sindhupalchok"),])db.executemany("INSERT INTO landslides VALUES (?, ?, ?, ?, ?)", [    (1, 2, "2026-06-28", 4, 1),    (2, 1, "2026-07-03", 0, 1),    (3, 3, "2026-07-05", 2, 0),    (4, 2, "2026-07-19", 7, 1),    (5, 5, "2026-08-02", 1, 0),    (6, 4, "2026-08-10", 3, 1),    (7, 2, "2026-08-21", 0, 0),    (8, 3, "2026-09-02", 5, 1),])  def ask(sql):    """Print each row the query gives, then a blank line."""    for row in db.execute(sql):        print(row)    print()  ask("SELECT slide_id, muni_id, houses FROM landslides WHERE day LIKE '2026-07%' ORDER BY day")ask("SELECT muni_id, COUNT(*), SUM(houses) FROM landslides GROUP BY muni_id ORDER BY muni_id")ask("""SELECT municipalities.name, landslides.day FROM landslides       JOIN municipalities ON landslides.muni_id = municipalities.muni_id       WHERE landslides.road_blocked = 1 AND municipalities.district = 'Sindhupalchok'       ORDER BY landslides.day""")

(2, 1, 0)(3, 3, 2)(4, 2, 7) (1, 1, 0)(2, 3, 11)(3, 2, 7)(4, 1, 3)(5, 1, 1) ('Melamchi', '2026-06-28')('Jugal', '2026-07-03')('Melamchi', '2026-07-19')

day BETWEEN '2026-07-01' AND '2026-07-31' is right for i too. (The landslides are made up for this paper.)

• i.It gives each key, a month’s name, one at a time, and totals[month] gives that month’s total. A dictionary keeps its keys in the order they were first added: Jestha was added by the first reading, and its second reading only added to the total already there.
• ii.% matches any number of characters (none too), so it matches every day starting with 2026-07, all of July. _ matches exactly one character, so '2026-07-_' would match only a day such as 2026-07-5, written with one digit, which none of ours is: it would find nothing.
• iii.With Gokyo closed, it is never added to the list, so the list runs out and the function returns None: “no way”, which is right. With Lukla closed, nothing changes: the start is put in came_from and the list before any check, so it still gives the 6-trail route. If a closed start means nobody can set out, that is wrong: check it first, if start in closed: return None.