Activity 11: Design a Struct for Real Data

Sheet A: weather stations

Odd groups. If your number is even, you want Sheet B.

Group members:

How this works

Design, ~6 minutes. On this sheet, write:
- a struct for one row of your data, with a type for every field
- a separate struct for the whole collection, holding a Vec of the first one
- three method signatures, each marked &self or &mut self. One of them has to change something
Write one method body, ~3 minutes. Real code if you can, pseudocode if you can't
Swap sheets, ~4 minutes. Trade with a group that did the other task, and do the three jobs at the bottom of this sheet

A worked example, from a different data set

street          species   height_m  healthy
Comm Ave        maple     11.2      yes
Comm Ave        oak       14.8      yes
Bay State Rd    elm        6.4      no

#![allow(unused)]
fn main() {
struct Tree { species: String, height_m: f64, healthy: bool }

struct Block { street: String, trees: Vec<Tree> }

impl Block {
    /// Height of the tallest tree on the block, or 0.0 if there are none.
    fn tallest(&self) -> f64 {
        let mut tallest = 0.0;
        for tree in &self.trees {
            if tree.height_m > tallest { tallest = tree.height_m; }
        }
        tallest
    }
}
}

Your data: five stations report once an hour. Here is the 7am round:

station   temp_f   wind_mph   precip_in   sensor_ok
BOS-01     54.3      12.0       0.00        yes
BOS-02     51.9       8.5       0.12        yes
BOS-03   -999.0       0.0       0.00        no
BOS-04     49.7      15.2       0.31        yes
BOS-05     53.1       6.4       0.00        yes

Design a Reading for one row, and a DailyReport that holds a date and all the readings for that day.

Write this method: average_temp, on DailyReport. The average temperature across the day's readings.

Look at BOS-03 before you write it.

Before you design it

Your collection struct holds the date. Pull one row out of it and hand it to someone else. Does that row still know when it happened?

What else would you want on a row, and where would it come from?

Your design

Row struct:

Collection struct:

Three method signatures, each marked &self or &mut self:

The method body you were asked to write:

If you finish early

What else would be useful on Reading itself? One example: the temperature in Celsius. Write the signature and a line of pseudocode for the body
Grouping by day is one choice. What else could you group by? You might want one station followed over time, with summary statistics of its own. Sketch that struct: what does it hold, and what methods would it have?

When you get the other group's sheet

Write on their sheet, and sign your group number.

Check every &self and &mut self. Mark any you disagree with and why
Add one other method that could be useful. Get creative! Signature, the right self, and a line of pseudocode saying what the body does
Name one field whose type you'd change, and say why

Sheet B: shuttle trips

Even groups. If your number is odd, you want Sheet A.

Group members:

How this works

Design, ~6 minutes. On this sheet, write:
- a struct for one row of your data, with a type for every field
- a separate struct for the whole collection, holding a Vec of the first one
- three method signatures, each marked &self or &mut self. One of them has to change something
Write one method body, ~3 minutes. Real code if you can, pseudocode if you can't
Swap sheets, ~4 minutes. Trade with a group that did the other task, and do the three jobs at the bottom of this sheet

A worked example, from a different data set

street          species   height_m  healthy
Comm Ave        maple     11.2      yes
Comm Ave        oak       14.8      yes
Bay State Rd    elm        6.4      no

#![allow(unused)]
fn main() {
struct Tree { species: String, height_m: f64, healthy: bool }

struct Block { street: String, trees: Vec<Tree> }

impl Block {
    /// Height of the tallest tree on the block, or 0.0 if there are none.
    fn tallest(&self) -> f64 {
        let mut tallest = 0.0;
        for tree in &self.trees {
            if tree.height_m > tallest { tallest = tree.height_m; }
        }
        tallest
    }
}
}

Your data: one row per shuttle arrival. Times are minutes after midnight, so 8:05am is 485:

route     stop             scheduled_min   actual_min   riders
CommAve   Kenmore Sq           485             489         31
CommAve   Silber Way           492             492         44
1BU       Danielsen Hall       495             493         12
CommAve   Marsh Plaza          500             514         58
1BU       Agganis Way          505             507          7

Design a Trip for one row, and a ServiceDay that holds a date and all of the day's trips.

Write this method: late_count, on ServiceDay. How many trips arrived more than threshold minutes after they were scheduled. It takes threshold as a second parameter, after self.

Look at the Danielsen trip before you write it.

Before you design it

Your collection struct holds the date. Pull one row out of it and hand it to someone else. Does that row still know when it happened?

What else would you want on a row, and where would it come from?

Your design

Row struct:

Collection struct:

Three method signatures, each marked &self or &mut self:

The method body you were asked to write:

If you finish early

What else would be useful on Trip itself? One example: the scheduled time as hours and minutes, so 485 reads as 8:05. Write the signature and a line of pseudocode for the body
Grouping by day is one choice. What else could you group by? You might want one route followed over time, with summary statistics of its own. Sketch that struct: what does it hold, and what methods would it have?

When you get the other group's sheet

Write on their sheet, and sign your group number.

Check every &self and &mut self. Mark any you disagree with and why
Add one other method that could be useful. Get creative! Signature, the right self, and a line of pseudocode saying what the body does
Name one field whose type you'd change, and say why

Solutions

Sheet A: weather stations

struct Reading {
    station: String,
    temp_f: f64,
    wind_mph: f64,
    precip_in: f64,
    sensor_ok: bool,
}

struct DailyReport {
    date: String,
    readings: Vec<Reading>,
}

Sheet A: `average_temp`

impl DailyReport {
    fn average_temp(&self) -> f64 {
        let mut sum = 0.0;
        let mut count = 0;
        for r in &self.readings {
            if !r.sensor_ok { continue; }
            sum += r.temp_f;
            count += 1;
        }
        if count == 0 { 0.0 } else { sum / count as f64 }
    }

    fn add_reading(&mut self, r: Reading) {
        self.readings.push(r);
    }
}

BOS-03 reports -999.0. Skip it and the average is 52.25. Average it in and you get -158.0.

Sheet B: shuttle trips

One row per shuttle arrival

struct Trip {
    route: String,
    stop: String,
    scheduled_min: u32,
    actual_min: u32,
    riders: u32,
}

struct ServiceDay {
    date: String,
    trips: Vec<Trip>,
}

Sheet B: `late_count`

impl ServiceDay {
    fn late_count(&self, threshold: u32) -> u32 {
        let mut count = 0;
        for t in &self.trips {
            if t.actual_min > t.scheduled_min + threshold {
                count += 1;
            }
        }
        count
    }

    fn add_trip(&mut self, t: Trip) {
        self.trips.push(t);
    }
}

The finish-early prompts. Celsius is fn temp_c(&self) -> f64 { (self.temp_f - 32.0) * 5.0 / 9.0 }. Hours and minutes is fn scheduled_hm(&self) -> (u32, u32) { (self.scheduled_min / 60, self.scheduled_min % 60) }, or a String. Both go on the row struct, not the collection: the method lives on whichever type holds the data it needs.

Before you design it. Neither Reading nor Trip carries a date (only the collection holds it). Both answers are defensible though. A date on every row adds lots copies of the same string, but no date means a row is not a standalone piece of info.

Another way to group it. StationHistory { station: String, readings: Vec<Reading> } with something like max_temp or hours_offline, or RouteHistory { route: String, trips: Vec<Trip> } with on_time_rate. This relates to the last question: when you want to regroup by station instead of by day, every Reading needs to know its own hour and date.

Keyboard shortcuts

DS210 Fa26 Course Site