Programming, problem solving, and algorithms

CPSC 203, 2026 W1

October 1, 2026

Announcements

• You must sit Examlet 1 in ORCA between today and Monday.
• POTWs

Where Are We?

What have we covered so far in CPSC 203?

• Complexity.
  • How does the amount of work grow as the size of the problem grows?
• Data representation.
  • Computers represent everything as binary integers.
  • We can map other types of data to numbers: characters, colours, etc.
• Arbitrary sized data: Lists and tuples.
• Compound data: Dataclasses.

Today

• Tabular data: dataframes and the pandas library
• Dates and times: datetime library

Pandas and Data Frames

Tabular Data

We often encounter data in tabular form.

• Consider the Billboard Hot 100.

• The data from the week of January 31, 2026:

date	title	artist	rank	peakPos	lastPos	isNew	weeks
2026-01-31	I Just Might	Bruno Mars	1	1	1	False	2
2026-01-31	Man I Need	Olivia Dean	2	2	4	False	22
2026-01-31	The Fate Of Ophelia	Taylor Swift	3	1	2	False	16
2026-01-31	Golden	HUNTR/X: EJAE, Audrey Nuna & REI AMI	4	1	3	False	31
2026-01-31	Ordinary	Alex Warren	5	1	5	False	49

CSV Format

Acronym for comma separated values

• Each line is one row of the table.
• Each column contains the information between commas.
  • Strings containing spaces or unusual symbols (such as commas) may be surrounded by double quotes.
• Use extension .csv for files with this formatting.

date,title,artist,rank,peakPos,lastPos,isNew,weeks
2026-01-31,I Just Might,Bruno Mars,1,1,1,False,2
2026-01-31,Man I Need,Olivia Dean,2,2,4,False,22
2026-01-31,The Fate Of Ophelia,Taylor Swift,3,1,2,False,16
2026-01-31,Golden,"HUNTR/X: EJAE, Audrey Nuna & REI AMI",4,1,3,False,31
2026-01-31,Ordinary,Alex Warren,5,1,5,False,49

• Show CSV file in text editor.
• Spreadsheet programs (MS Excel, Google sheets) can read and write CSV files, but usually choose to store their data in more complex formats.
  • CSV can only store simple data. The fancier formats allow for equations, plots, etc.
  • Show CSV in Excel.

Processing Tabular Data in CPSC 103

Typical CPSC 103 workflow for CSV files:

• Identify or create a data type for the information in each column of interest.
• Create Compound data type to store a row.
• Create List[Compound] data type to store the full data set.
• Write read and parse functions to get the data from the file.
• Write analyze and visualize functions to process the data and produce output.

Go back to the table and identify the type for each column of the Hot 100 data.

CPSC 103 Approach Example

Create a compound for each row of the Hot 100:

from dataclasses import dataclass

@dataclass
Class Song:
        date: date
        title: str
        artist: str
        rank: int
        peakPos: int
        lastPos: int
        isNew: bool
        weeks: int

Create a list[Song] to store the table.

In CPSC 103, we treated tables as arbitrary size (list) containing one compound for each row.

This approach is very general, in that the compound can contain other things: another compound, or even an arbitrary size.

For example, what if you wanted to store the audio of the song in each row? Each song has a different length.

That said, in practice most of the tables we deal with have atomic data (or other simple data) as entries. CSV files, for example, cannot store complex data entries. These are the situations when we can draw upon tools that are specialized for tabular data.

CPSC 103 Queries

Answering a query about the data in the table requires loop(s).

For example: What is the name of the artist of the song which has been on the chart for the most weeks?

# chartlist: list[Song]
# Find artist on charts the longest
most_weeks = 0
most_artist = ''
for song in chartlist:
    if song.weeks > most_weeks:
        most_weeks = song.weeks
        most_artist = song.artist

print(most_artist, most_weeks)

Go to HtDP and look at the templates for list and compound.

CPSC 103 Approach: Pros and Cons

• Pro: You think carefully about the information and the data representation before coding.
• Pro: You practice your HtDD and HtDF processes and get correct implementations.
• Con: Your code is very specific to a single data set.
• Con: You can easily manipulate rows, but not columns.
• Con: New functions and loops are required for even simple filter or map operations.
• Con: Your code is not particularly efficient.

Alternative: List of Dictionaries

• Each row is represented as a dictionary.
  
• The keys of the dictionary are the column names.
• The table is a list of dictionaries.

A Better Way: DataFrames

A DataFrame is a 2D table: rows = records, columns = attributes.

DataFrames have built-in operations for filtering, aggregating, and more.

.csv files for Pandas

If our data is in a .csv file:

df = pd.read_csv('billboard.csv')

.csv files must have attribute names in row 1, and data beginning in row 2.

title,artist,rank,last_week,...
Die With A Smile,Lady Gaga,1,1,...
APT.,Rose & Bruno Mars,2,2,...

The Payoff: One-Liners!

Task	With loops	With DataFrames
Count new songs	5+ lines	df[df['isNew']].shape[0]
Average peak	4 lines	df['peakPos'].mean()
Filter long-running	4 lines	df[df['weeks'] > 10]
Find biggest mover	8+ lines	df.loc[df['gradient'].abs().idxmax()]

So what does the template look like for dataframes? Its complicated…

Remember that we sometimes use the arbitrary-size template for lists, and sometimes we use the compound template? For dataframes, there are many, many common templates.

The Pandas cheat sheet encompasses them all, but it will take some practice before you can easily read it.

You will have access to the cheat sheet during exams in which you are expected to use dataframes.

Selecting rows

• df[df[col] > val] — filter by condition

• df.nlargest(n, col) — top n by column

• By far the most common is filter by condition.
• I sometimes use head, tail or iloc to look at example rows when the table is too big to look at all at once.

Adding a column

• df['new_col'] = expression

  Adds a column to the DataFrame containing the computed value for every row.

• df[df['new_col'] > val]

  Filter rows by the new column.

Very useful to create columns containing intermediate expressions when you are working on a complex map, filter or aggregation.

Demo 1: Counting with a condition

Demo: How many Taylor Swift songs?

Goal: Count songs by a specific artist.

Strategy:

1. Filter rows where artist contains “Taylor”
2. Count the resulting rows

Parse through df[df['artist'].str.contains('Taylor')]. Show what df['artist'].str.contains('Taylor') returns.

Demo: How many songs at #1 peak?

Goal: Count songs that peaked at #1.

Strategy:

1. Filter where peakPos == 1
2. Count with len() or .shape[0]

Parse through df[df['peakPos'] == 1]. Show what df['peakPos'] == 1 returns.

Parse through peaked_at_one[['rank', 'title', 'peakPos']].head(). Why the double [[? The outer set is indicating that we are going to index, while the inner set is creating a list[str]. Why use head()?

Demo 2: Aggregation (mean, max, min)

Demo: Average weeks on chart

Goal: Find the average weeks on chart.

Strategy:

• Select a column: df['weeks']
• Apply aggregation: .mean()

Explain the :.1f notation for floating point numbers.

Demo: Average weeks for top 10 only

Goal: Average weeks, but only for top 10 songs.

Strategy:

1. Filter first: df[df['rank'] <= 10]
2. Then aggregate: ['weeks'].mean()

You could do the filter, column select and aggregation all in one statement, rather than creating the intermediate variable top10.

• However, remember that your goal is human readability, not minimum number of lines of code.

Demo 3: Finding extremes with idxmax

Demo: Longest-running song

Goal: Find the song with the most weeks on chart.

Strategy:

• idxmax() returns the row index of the maximum
• df.loc[idx] retrieves that row

This combination of idxmax and loc is a common dataframes pattern.

Demo: Biggest riser this week

Goal: Which song moved up the most?

Problem: New songs have lastPos = 0 (they weren’t on the chart last week).

Strategy:

1. Filter out new songs first
2. Compute change: lastPos - rank
3. Find max with idxmax()

Handling special values

The lastPos = 0 problem

New songs weren’t on the chart last week, so lastPos = 0.

If we compute change = lastPos - rank:

• New song at #23: 0 - 23 = -23
• Looks like it fell 23 spots!

Solution: Filter them out first.

It is not unusual to have special values in tabular data that denote something weird. Ideally, it would be encoded with an Optional data type, so the weird case would not be a valid (if unusual) value for the not-weird case. But it can sometimes be infeasible to do it the smart way.

We need to identify and filter out such weird cases before we accidentally include them in some otherwise reasonable operation (such as arithmetic).

Here is a case where you might want to create a new column isNew which has precomputed which songs are new.

Let’s Write Code

Open CA5.1 Billboard, and load STUDENT_nb.py.

PrairieLearn Activity

Pandas and Dataframes: A Sharp Blade

Compared to the CPSC 103 process, Pandas:

• Allows us to quickly^† perform complex data manipulations.
• Allows us to quickly^† generate bugs that are difficult to spot and even harder to remove.

^† In terms of both coding and execution speed.

Dates and Times

Warm Up

datetime: a Python library that simplifies date computation.

Objects we need

• date — a calendar day (year-month-day)
• timedelta — a number of days you can add/subtract

Core calls

• today=date.today(): today’s date
• today.weekday(): an integer 0..6 with Mon=0, … , Sun=6
• today - timedelta(days=k): the date k days earlier

Ref: https://docs.python.org/3/library/datetime.html

Billboard constraint - The library expects Saturday chart dates. We need the most recent Saturday at or before today.

Last Saturday

Let \(w =\) today.weekday() \(\in \{0,\dots,6\}\) and note Saturday \(= 5\).

offset: How many days to step back to reach Saturday.

• If today is Saturday (\(w=5\)): offset \(=0\)
• If today is Sunday (\(w=6\)): offset \(=1\)
• If today is Monday (\(w=0\)): offset \(=2\)

\[ \texttt{offset} = (w - 5)\bmod 7 \]

\[ \texttt{lastSaturday} = \texttt{today} - \texttt{timedelta(days=offset)}. \]

Write the clock on a piece of paper, and count around it (counter-clockwise because we are subtracting). Demo for Tuesday or Wednesday.

Addition / subtraction on a clock (rather than an infinite number line) is modular arithmetic.

Code

Modify to Mark Carney’s birthday today = date(year=1965,month=3,day=16) (born in Fort Smith, NWT). That was a Tuesday, so previous Saturday is March 13.

Resources

https://pymotw.com/2/datetime/

https://www.dataschool.io/best-python-pandas-resources/

https://pandas.pydata.org/Pandas_Cheat_Sheet.pdf

https://queirozf.com/entries/pandas-dataframe-plot-examples-with-matplotlib-pyplot