Final Exam Model Paper 7

i. The loop keeps the biggest value seen so far: 420, then 510.

Biggest: 510

ii. max is a built-in function, but line 2 made a variable called max that shadows it: from then on, the name max means the number 510, and a number can’t be called like a function. Give the variable another name, such as biggest, in lines 2, 5 and 6; then line 7’s max() is the built-in again and prints Check: 510.

• Nodes: the farms. Edges: join two farms that share a pasture; they go both ways (sharing is mutual), so each edge is written in both farms’ lists when the graph is built from the pasture lists.
• Search: a breadth-first search from the sick farm, keeping each farm’s steps in a dictionary. It finds the farms in rings: steps 1 (sharing a pasture with the sick farm, vaccinate first), then steps 2. List the farms with steps 1 or 2, nearest first.

First check exactly which rows the condition picks, then delete with the same WHERE:

SELECT * FROM sales WHERE sale_id = 57;DELETE FROM sales WHERE sale_id = 57;

The farmer, crop, kg and day are the same in rows 56 and 57, so a WHERE on them would delete both, the real sale too. sale_id is the primary key, different in every row, so it picks exactly one.

• i.Line 5 makes one row of three charts side by side, axes: small multiples. Each turn of the loop picks one cooperative’s rows and draws its milk month by month, as a line with dots, in its own chart, with the cooperative’s name as the title.
• ii.So that all three charts have the same scales: a point at the same height means the same number of liters in each. Without them, each chart would stretch its own axis to fit its own numbers, so Kalika’s line, about 300 liters, would fill its chart just as Ratnanagar’s 3,000 fills its own, and the three cooperatives would look about the same size.

• i.He doesn’t know which change fixed it, so if it happens again he won’t know what to do; and the two changes that weren’t needed may have broken something else that will show up later (the new driver, for example).
• ii.Change one thing at a time, and test after each. If a change doesn’t help, undo it before trying the next, and keep a short note of what he tried and what happened.

• i.Line 4, a TypeError: + was given an int (total, 0) and a str, and Python can’t add a number to text. The text came from input(), which always gives a string: “12” × 50 is not 600, it is the text “12” repeated 50 times, so the multiplication worked and the addition failed.
• ii.Turn the answer into a number when it is read: bags = int(input("Bags of rice? ")). Then the program prints Total: 600 kg.

• i.At every token the model has chances for many next tokens, and with a temperature above 0 it picks at random, weighted by those chances. The digits of a number are tokens too, so now and then a less likely digit is picked, and the answer changes; once a wrong digit is written, the rest of the answer goes on from it.
• ii.All five answers come from the same model and the same training text. If a wrong number was common in that text, the most common answer can be wrong too. Asking again only shows what the model finds likely, not what is true. A real check is a reliable source: here, 8,586 m is the height given by the Survey Department and in atlases.

It doesn’t. A test score only says how the model does on things like its test set, and their apple leaves, apple diseases, mountain light and winter photos were not in it: the model may be much worse on them, and one total can hide a group where it is weak. The group should ask for (or make) a test set of their own photos, with the diseases checked by an expert, count the accuracy for apples separately, and judge the harm of each kind of mistake (a disease missed can cost a season’s crop) before relying on it.

The choices of how to design it, whether to use it, where to deploy it (put it into service), and how to regulate it (what rules to make). Examples: design, storing a year in two digits to save memory (the Year 2000 problem); use, whether a shop takes cash or only digital payments; deploy, which towns get a new mobile tower first; regulate, a rule with a review clause or a sandbox for a new payment service.

It is a choice that will cost more the longer it lasts. Preeti stores Devanagari as the codes of English letters, so on any computer or phone without that font, and in today’s programs, the records turn into gibberish; they can’t be searched or sorted properly in Nepali, sent to a bank or a government office, or read in twenty years. Each year adds more documents to convert. A better choice is to type new records in Unicode Nepali now, and convert the old ones, keeping the originals until the copies are checked.

i. For example: read_weights() asks for the kg of each grade (input, a side effect); payment(kg_by_grade) works out the total kg and the amount; print_slip(name, total_kg, amount) prints the slip (a side effect). payment should be pure: it only takes the weights in and gives the numbers back.

ii.

"""Paying farmers at the orange collection center.""" PRICES = {"A": 80, "B": 60, "C": 35}     # rupees a kg, by grade  def payment(kg_by_grade):    """Given the kg of each grade, as a dictionary, return the total kg    and the amount to pay in rupees."""    total_kg = 0    amount = 0    for grade in kg_by_grade:        total_kg = total_kg + kg_by_grade[grade]        amount = amount + kg_by_grade[grade] * PRICES[grade]    return total_kg, amount  if __name__ == "__main__":    print(payment({"A": 40, "B": 25, "C": 10}), "should be (75, 5050)")

40 × 80 + 25 × 60 + 10 × 35 = 3,200 + 1,500 + 350 = Rs 5,050, for 75 kg.

iii. A returned value can be used: printed on a slip today, but also added to the day’s totals, saved in the cooperative’s records or shown in an app, and it can be tested by checking the numbers. A function that printed would decide for every caller what to do with them.

i.

• ii.5 canals: Field 6 was first reached from Field 5, Field 5 from Field 3, Field 3 from the Lower branch, the Lower branch from the Main canal, and that from the Intake: Intake → Main canal → Lower branch → Field 3 → Field 5 → Field 6.
• iii.The graph is directed: the canal runs from Field 7 to Field 6, so water (and the search, which follows the arrows only forwards) can’t go up it from Field 6. No place’s list contains Field 7, so nothing leads to it and it gets no water from the intake.

i. The two biggest sales, by kg:

ii. One row per crop, with the average of its prices: ginger (95 + 100 + 90) ÷ 3 = 95, orange (60 + 55 + 65) ÷ 3 = 60.

iii. The ginger sales of farmers in ward 2, each joined to its farmer’s row, the most first (Kamala Rana sold ginger too, but she is in ward 5):

iv. The ginger and cardamom sales of under 100 kg, the smallest first (sale 1, of 120 kg, is left out):

• i.Each dot is one farm: across, its land; up, its harvest. The color shows its district. So the chart shows three variables at once (a multivariate chart).
• ii.Each district’s name with its group of rows, one district at a time, so each district is drawn in its own color with its own label. The legend is the key that says which color is which district; without it the colors mean nothing.
• iii.The farms run from 2 to 300 ropani, and the harvests from 700 to 61,000 kg: several powers of ten. On ordinary scales the big farms would sit at the far corner and the small ones would be squashed together near 0. On a log scale each step is ten times the one before, so small and big farms all get room. A reader must be told, or they will think the gaps are equal.
• iv.Shows: in these few farms, Kaski’s farms harvested more for the same land than Dhading’s. Doesn’t show: why (irrigation, soil or the crops grown could explain it), or that it holds for all farms in the districts: there are only three farms from each, so it could be chance.

• i.Keeping readings, each with its time, is logging, and the file of readings is a data log.
• ii.So that the manager knows when each reading was taken: to see how the temperature changes through the night and day, to find when something went wrong (and match it to the power cut), and to see which readings are missing.
• iii.It is plain text, one line per reading, with the values separated by commas, so the device can simply add a line each time, and almost any program (a spreadsheet, pandas) can read it.
• iv.As a gap: the line should break, or show only the points, with nothing between 01:00 and 03:00. A line drawn straight across would look like a measurement of temperatures nobody measured.

The function printed the amount and returned nothing, so it gave back None, and 0 + None is a TypeError. It must return the amount; the printing belongs in main.py.

"""How much seed to buy for a field."""import math RATES = {"rice": 2.5, "wheat": 6, "maize": 1.5}    # kg of seed for each ropani  def seed_needed(crop, ropani):    """The kg of seed to buy for this crop and land, rounded up to a whole kg."""    return math.ceil(RATES[crop] * ropani)  if __name__ == "__main__":    # tests: an exact amount, and amounts that must be rounded up    print(seed_needed("wheat", 3), "should be 18")    print(seed_needed("maize", 5), "should be 8")    print(seed_needed("rice", 1), "should be 3")

Run by itself, seeds.py prints its tests:

18 should be 188 should be 83 should be 3

import seeds total = 0for crop, ropani in [("rice", 4), ("wheat", 3), ("maize", 5)]:    kg = seeds.seed_needed(crop, ropani)    print(crop + ":", kg, "kg")    total = total + kgprint("Buy", total, "kg of seed in all")

rice: 10 kgwheat: 18 kgmaize: 8 kgBuy 36 kg of seed in all

Maize: 1.5 × 5 = 7.5 kg, rounded up to 8.

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

import sqlite3 db = sqlite3.connect(":memory:")db.execute("CREATE TABLE plots (plot_id INTEGER PRIMARY KEY, farmer TEXT, ward INTEGER, crop TEXT, ropani INTEGER)")db.execute("""CREATE TABLE soil_tests (test_id INTEGER PRIMARY KEY,              plot_id INTEGER REFERENCES plots(plot_id),              month TEXT, ph REAL, nitrogen TEXT)""")db.executemany("INSERT INTO plots VALUES (?, ?, ?, ?, ?)", [    (1, "Ram Shrestha", 3, "potato", 8),    (2, "Sita Lama", 5, "tomato", 4),    (3, "Hari Dahal", 3, "maize", 10),    (4, "Gita Tamang", 7, "potato", 6),    (5, "Bina Thapa", 5, "cauliflower", 5),])db.executemany("INSERT INTO soil_tests VALUES (?, ?, ?, ?, ?)", [    (1, 1, "Baisakh", 5.2, "low"),    (2, 2, "Baisakh", 6.4, "medium"),    (3, 3, "Baisakh", 5.8, "low"),    (4, 4, "Jestha", 4.9, "low"),    (5, 1, "Kartik", 5.6, "medium"),    (6, 5, "Kartik", 6.8, "high"),    (7, 2, "Kartik", 6.1, "medium"),    (8, 3, "Kartik", 5.5, "low"),])  def ask(sql):    """Print each row the query gives, then a blank line."""    for row in db.execute(sql):        print(row)    print()  ask("SELECT test_id, plot_id, ph FROM soil_tests WHERE ph < 5.5 ORDER BY ph")ask("SELECT plot_id, COUNT(*), MIN(ph) FROM soil_tests GROUP BY plot_id")ask("""SELECT plots.farmer, plots.crop, soil_tests.ph FROM soil_tests       JOIN plots ON soil_tests.plot_id = plots.plot_id       WHERE soil_tests.month = 'Kartik' ORDER BY soil_tests.ph""")

(4, 4, 4.9)(1, 1, 5.2) (1, 2, 5.2)(2, 2, 6.1)(3, 2, 5.5)(4, 1, 4.9)(5, 1, 6.8) ('Hari Dahal', 'maize', 5.5)('Ram Shrestha', 'potato', 5.6)('Sita Lama', 'tomato', 6.1)('Bina Thapa', 'cauliflower', 6.8)

Test 8, with a pH of exactly 5.5, is not below 5.5, so it isn’t in the first answer. (The tests are made up for this paper.)

rolpa = {    "Liwang": ["Sulichaur", "Mirul", "Jaimakasala"],    "Sulichaur": ["Liwang", "Holeri"],    "Mirul": ["Liwang", "Thawang", "Korchabang"],    "Jaimakasala": ["Liwang", "Holeri"],    "Holeri": ["Sulichaur", "Jaimakasala", "Ghartigaun"],    "Thawang": ["Mirul", "Jelbang"],    "Korchabang": ["Mirul", "Jelbang"],    "Jelbang": ["Thawang", "Korchabang", "Rangsi"],    "Rangsi": ["Jelbang"],    "Ghartigaun": ["Holeri"],}  def fewest_trails(graph, start, goal):    """The fewest trails from start to goal, or None if there is no way."""    steps = {start: 0}    to_visit = [start]    while len(to_visit) > 0:        village = to_visit.pop(0)        if village == goal:            return steps[village]        for nxt in graph[village]:            if nxt not in steps:                steps[nxt] = steps[village] + 1                to_visit.append(nxt)    return None  for goal in ["Rangsi", "Ghartigaun", "Thawang"]:    print("Liwang to", goal + ":", fewest_trails(rolpa, "Liwang", goal), "trails") print("Stay the night at:")for village in rolpa:    if fewest_trails(rolpa, "Liwang", village) > 2:        print(village)

Liwang to Rangsi: 4 trailsLiwang to Ghartigaun: 3 trailsLiwang to Thawang: 2 trailsStay the night at:JelbangRangsiGhartigaun

Every village can be reached, so the function never returns None here; on a map where one couldn’t, None > 2 would be an error, and the loop would need to check for None first.

• i.Seed is sold in whole kg, and too little seed leaves part of the field unsown, so the amount must never be rounded down. round() goes to the nearest whole number, sometimes down: 1 ropani of rice needs 2.5 kg, and round(2.5) gives 2 (Python rounds a half to the even number), half a kg short. math.ceil(2.5) gives 3.
• ii.A plot is tested many times (plot 1 in Baisakh and in Kartik), so plot_id is repeated in soil_tests and can’t be a primary key, which must be different in every row. It is a foreign key pointing to the plot; test_id names each test. One plot has many tests: a one-to-many relationship.
• iii.When a village is first added, its steps are its neighbor’s steps plus one, kept in the steps dictionary. The queue takes places in the order they were added, so all villages 1 trail away are visited before any 2 away, and so on in rings; the first time the goal comes off the queue, no shorter route can be left.