Polars Tutorial

Adopted from https://www.youtube.com/playlist?list=PLINDUevGdb7U_ZRLCpqKWutmcY0vCUFOI

Reading CSV Files

import polars as pl
# df = pl.read_csv('pl_data.csv')

Creating New Columns

df = pl.DataFrame({'Name': ['Mario', 'Luigi', 'Wario', 'Mario', 'Mario'], 'Age': [30,28,26,30,30]})
print(df)

shape: (5, 2)
┌───────┬─────┐
│ Name  ┆ Age │
│ ---   ┆ --- │
│ str   ┆ i64 │
╞═══════╪═════╡
│ Mario ┆ 30  │
│ Luigi ┆ 28  │
│ Wario ┆ 26  │
│ Mario ┆ 30  │
│ Mario ┆ 30  │
└───────┴─────┘

score_values = pl.Series([95,99,94,90,96])
df = df.with_columns(Score = pl.lit(score_values),
                     Score_x2 = pl.lit(score_values*2))
print(df)

shape: (5, 4)
┌───────┬─────┬───────┬──────────┐
│ Name  ┆ Age ┆ Score ┆ Score_x2 │
│ ---   ┆ --- ┆ ---   ┆ ---      │
│ str   ┆ i64 ┆ i64   ┆ i64      │
╞═══════╪═════╪═══════╪══════════╡
│ Mario ┆ 30  ┆ 95    ┆ 190      │
│ Luigi ┆ 28  ┆ 99    ┆ 198      │
│ Wario ┆ 26  ┆ 94    ┆ 188      │
│ Mario ┆ 30  ┆ 90    ┆ 180      │
│ Mario ┆ 30  ┆ 96    ┆ 192      │
└───────┴─────┴───────┴──────────┘

Dropping Columns

print(df.drop('Score_x2'))
print(df.drop('Score_x2', 'Score'))
print(df.drop(['Score_x2', 'Score']))
df.drop_in_place('Score_x2')
print(df)

shape: (5, 3)
┌───────┬─────┬───────┐
│ Name  ┆ Age ┆ Score │
│ ---   ┆ --- ┆ ---   │
│ str   ┆ i64 ┆ i64   │
╞═══════╪═════╪═══════╡
│ Mario ┆ 30  ┆ 95    │
│ Luigi ┆ 28  ┆ 99    │
│ Wario ┆ 26  ┆ 94    │
│ Mario ┆ 30  ┆ 90    │
│ Mario ┆ 30  ┆ 96    │
└───────┴─────┴───────┘
shape: (5, 2)
┌───────┬─────┐
│ Name  ┆ Age │
│ ---   ┆ --- │
│ str   ┆ i64 │
╞═══════╪═════╡
│ Mario ┆ 30  │
│ Luigi ┆ 28  │
│ Wario ┆ 26  │
│ Mario ┆ 30  │
│ Mario ┆ 30  │
└───────┴─────┘
shape: (5, 2)
┌───────┬─────┐
│ Name  ┆ Age │
│ ---   ┆ --- │
│ str   ┆ i64 │
╞═══════╪═════╡
│ Mario ┆ 30  │
│ Luigi ┆ 28  │
│ Wario ┆ 26  │
│ Mario ┆ 30  │
│ Mario ┆ 30  │
└───────┴─────┘
shape: (5, 3)
┌───────┬─────┬───────┐
│ Name  ┆ Age ┆ Score │
│ ---   ┆ --- ┆ ---   │
│ str   ┆ i64 ┆ i64   │
╞═══════╪═════╪═══════╡
│ Mario ┆ 30  ┆ 95    │
│ Luigi ┆ 28  ┆ 99    │
│ Wario ┆ 26  ┆ 94    │
│ Mario ┆ 30  ┆ 90    │
│ Mario ┆ 30  ┆ 96    │
└───────┴─────┴───────┘

Drop Duplicate Rows

df = pl.DataFrame({'Name': ['Mario', 'Luigi', 'Wario', 'Mario', 'Mario'], 'Age': [30,28,26,30,30]})
print(df.unique())

shape: (3, 2)
┌───────┬─────┐
│ Name  ┆ Age │
│ ---   ┆ --- │
│ str   ┆ i64 │
╞═══════╪═════╡
│ Mario ┆ 30  │
│ Luigi ┆ 28  │
│ Wario ┆ 26  │
└───────┴─────┘

Iterating Polars Dataframes

for row in df.rows():
    print(row)
    print(row[0])

('Mario', 30)
Mario
('Luigi', 28)
Luigi
('Wario', 26)
Wario
('Mario', 30)
Mario
('Mario', 30)
Mario

Plot Polars Dataframe

import matplotlib.pyplot as plt
df = pl.DataFrame({'Date': ['1/1/2023','1/2/2023','1/3/2023','1/4/2023','1/5/2023','1/6/2023','1/7/2023','1/8/2023','1/9/2023','1/10/2023'],
                   'Price': [15,16,16,15,14,13,14,17,16,18]})
dates = list(df['Date'])
prices = list(df['Price'])
plt.plot(dates, prices)
plt.show()

Converting Polars Dataframe To Pandas

This operation requires that both pandas and pyarrow are installed.

x = df.to_pandas()
print(type(df))
print(type(x))
print(df)
print(x)

<class 'polars.dataframe.frame.DataFrame'>
<class 'pandas.core.frame.DataFrame'>
shape: (10, 2)
┌───────────┬───────┐
│ Date      ┆ Price │
│ ---       ┆ ---   │
│ str       ┆ i64   │
╞═══════════╪═══════╡
│ 1/1/2023  ┆ 15    │
│ 1/2/2023  ┆ 16    │
│ 1/3/2023  ┆ 16    │
│ 1/4/2023  ┆ 15    │
│ 1/5/2023  ┆ 14    │
│ 1/6/2023  ┆ 13    │
│ 1/7/2023  ┆ 14    │
│ 1/8/2023  ┆ 17    │
│ 1/9/2023  ┆ 16    │
│ 1/10/2023 ┆ 18    │
└───────────┴───────┘
        Date  Price
0   1/1/2023     15
1   1/2/2023     16
2   1/3/2023     16
3   1/4/2023     15
4   1/5/2023     14
5   1/6/2023     13
6   1/7/2023     14
7   1/8/2023     17
8   1/9/2023     16
9  1/10/2023     18

Filtering Polars Dataframes

print(df.filter(df['Price'] > 15))
print(df.filter((df['Price'] > 15) & (df['Price'] < 17)))

shape: (5, 2)
┌───────────┬───────┐
│ Date      ┆ Price │
│ ---       ┆ ---   │
│ str       ┆ i64   │
╞═══════════╪═══════╡
│ 1/2/2023  ┆ 16    │
│ 1/3/2023  ┆ 16    │
│ 1/8/2023  ┆ 17    │
│ 1/9/2023  ┆ 16    │
│ 1/10/2023 ┆ 18    │
└───────────┴───────┘
shape: (3, 2)
┌──────────┬───────┐
│ Date     ┆ Price │
│ ---      ┆ ---   │
│ str      ┆ i64   │
╞══════════╪═══════╡
│ 1/2/2023 ┆ 16    │
│ 1/3/2023 ┆ 16    │
│ 1/9/2023 ┆ 16    │
└──────────┴───────┘

Joining/merging Polars Dataframes

df1 = pl.DataFrame({
    'key': ['A', 'B', 'C', 'D'],
    'value_left': [1, 2, 3, 4]
})

df2 = pl.DataFrame({
    'key': ['B', 'C', 'D', 'E'],
    'value_left': [5, 6, 7, 8]
})

joined_df = df1.join(df2, on = 'key')
print(joined_df)
joined_df = df1.join(df2, on = 'key', how = 'left')
print(joined_df)

shape: (3, 3)
┌─────┬────────────┬──────────────────┐
│ key ┆ value_left ┆ value_left_right │
│ --- ┆ ---        ┆ ---              │
│ str ┆ i64        ┆ i64              │
╞═════╪════════════╪══════════════════╡
│ B   ┆ 2          ┆ 5                │
│ C   ┆ 3          ┆ 6                │
│ D   ┆ 4          ┆ 7                │
└─────┴────────────┴──────────────────┘
shape: (4, 3)
┌─────┬────────────┬──────────────────┐
│ key ┆ value_left ┆ value_left_right │
│ --- ┆ ---        ┆ ---              │
│ str ┆ i64        ┆ i64              │
╞═════╪════════════╪══════════════════╡
│ A   ┆ 1          ┆ null             │
│ B   ┆ 2          ┆ 5                │
│ C   ┆ 3          ┆ 6                │
│ D   ┆ 4          ┆ 7                │
└─────┴────────────┴──────────────────┘

Aggregate Data

df = pl.DataFrame({
    'Name': ['Chris', 'Chris', 'Kyle', 'John', 'John', 'Chris', 'Kyle'],
    'Calories': [100,200,50,100,350,150,450]
})
df2 = df.group_by(['Name']).agg(
    sum_ = pl.col('Calories').sum(),
    count_ = pl.col('Calories').count()
)
print(df2)

shape: (3, 3)
┌───────┬──────┬────────┐
│ Name  ┆ sum_ ┆ count_ │
│ ---   ┆ ---  ┆ ---    │
│ str   ┆ i64  ┆ u32    │
╞═══════╪══════╪════════╡
│ Kyle  ┆ 500  ┆ 2      │
│ Chris ┆ 450  ┆ 3      │
│ John  ┆ 450  ┆ 2      │
└───────┴──────┴────────┘

Sorting

print(df.sort(['Name', 'Calories']))

shape: (7, 2)
┌───────┬──────────┐
│ Name  ┆ Calories │
│ ---   ┆ ---      │
│ str   ┆ i64      │
╞═══════╪══════════╡
│ Chris ┆ 100      │
│ Chris ┆ 150      │
│ Chris ┆ 200      │
│ John  ┆ 100      │
│ John  ┆ 350      │
│ Kyle  ┆ 50       │
│ Kyle  ┆ 450      │
└───────┴──────────┘

Renaming Columns

renamed_col = df.rename({'Name': 'ID', 'Calories': 'Consumption'})
print(renamed_col)

shape: (7, 2)
┌───────┬─────────────┐
│ ID    ┆ Consumption │
│ ---   ┆ ---         │
│ str   ┆ i64         │
╞═══════╪═════════════╡
│ Chris ┆ 100         │
│ Chris ┆ 200         │
│ Kyle  ┆ 50          │
│ John  ┆ 100         │
│ John  ┆ 350         │
│ Chris ┆ 150         │
│ Kyle  ┆ 450         │
└───────┴─────────────┘

Select Columns

print(df2.select(['Name', 'count_']))

shape: (3, 2)
┌───────┬────────┐
│ Name  ┆ count_ │
│ ---   ┆ ---    │
│ str   ┆ u32    │
╞═══════╪════════╡
│ Kyle  ┆ 2      │
│ Chris ┆ 3      │
│ John  ┆ 2      │
└───────┴────────┘

Fill/Drop Null Values

df = pl.DataFrame({
    'Name': ['John', 'Steven', 'Chris', None],
    'Age': [20, 25, None, 32],
    'Score': [99, None, 100, None]
})
print(df)
fill_df = df.fill_null('NULL_STRING')
print(fill_df)
fill_df = df.fill_null(0)
print(fill_df)
print(df.drop_nulls())
print(fill_df.drop_nulls())

shape: (4, 3)
┌────────┬──────┬───────┐
│ Name   ┆ Age  ┆ Score │
│ ---    ┆ ---  ┆ ---   │
│ str    ┆ i64  ┆ i64   │
╞════════╪══════╪═══════╡
│ John   ┆ 20   ┆ 99    │
│ Steven ┆ 25   ┆ null  │
│ Chris  ┆ null ┆ 100   │
│ null   ┆ 32   ┆ null  │
└────────┴──────┴───────┘
shape: (4, 3)
┌─────────────┬──────┬───────┐
│ Name        ┆ Age  ┆ Score │
│ ---         ┆ ---  ┆ ---   │
│ str         ┆ i64  ┆ i64   │
╞═════════════╪══════╪═══════╡
│ John        ┆ 20   ┆ 99    │
│ Steven      ┆ 25   ┆ null  │
│ Chris       ┆ null ┆ 100   │
│ NULL_STRING ┆ 32   ┆ null  │
└─────────────┴──────┴───────┘
shape: (4, 3)
┌────────┬─────┬───────┐
│ Name   ┆ Age ┆ Score │
│ ---    ┆ --- ┆ ---   │
│ str    ┆ i64 ┆ i64   │
╞════════╪═════╪═══════╡
│ John   ┆ 20  ┆ 99    │
│ Steven ┆ 25  ┆ 0     │
│ Chris  ┆ 0   ┆ 100   │
│ null   ┆ 32  ┆ 0     │
└────────┴─────┴───────┘
shape: (1, 3)
┌──────┬─────┬───────┐
│ Name ┆ Age ┆ Score │
│ ---  ┆ --- ┆ ---   │
│ str  ┆ i64 ┆ i64   │
╞══════╪═════╪═══════╡
│ John ┆ 20  ┆ 99    │
└──────┴─────┴───────┘
shape: (3, 3)
┌────────┬─────┬───────┐
│ Name   ┆ Age ┆ Score │
│ ---    ┆ --- ┆ ---   │
│ str    ┆ i64 ┆ i64   │
╞════════╪═════╪═══════╡
│ John   ┆ 20  ┆ 99    │
│ Steven ┆ 25  ┆ 0     │
│ Chris  ┆ 0   ┆ 100   │
└────────┴─────┴───────┘

Convert String To Date

df = pl.DataFrame({'Date': ['1/1/2023','1/2/2023','1/3/2023','1/4/2023','1/5/2023','1/6/2023','1/7/2023','1/8/2023','1/9/2023','1/10/2023'],
                   'Price': [15,16,16,15,14,13,14,17,16,18]})
df = df.with_columns(Date2 = df['Date'].str.to_date())
print(df)
df = df.with_columns(Date3 = df['Date2'].dt.strftime('%m%Y'))
print(df)

shape: (10, 3)
┌───────────┬───────┬────────────┐
│ Date      ┆ Price ┆ Date2      │
│ ---       ┆ ---   ┆ ---        │
│ str       ┆ i64   ┆ date       │
╞═══════════╪═══════╪════════════╡
│ 1/1/2023  ┆ 15    ┆ 2023-01-01 │
│ 1/2/2023  ┆ 16    ┆ 2023-02-01 │
│ 1/3/2023  ┆ 16    ┆ 2023-03-01 │
│ 1/4/2023  ┆ 15    ┆ 2023-04-01 │
│ 1/5/2023  ┆ 14    ┆ 2023-05-01 │
│ 1/6/2023  ┆ 13    ┆ 2023-06-01 │
│ 1/7/2023  ┆ 14    ┆ 2023-07-01 │
│ 1/8/2023  ┆ 17    ┆ 2023-08-01 │
│ 1/9/2023  ┆ 16    ┆ 2023-09-01 │
│ 1/10/2023 ┆ 18    ┆ 2023-10-01 │
└───────────┴───────┴────────────┘
shape: (10, 4)
┌───────────┬───────┬────────────┬────────┐
│ Date      ┆ Price ┆ Date2      ┆ Date3  │
│ ---       ┆ ---   ┆ ---        ┆ ---    │
│ str       ┆ i64   ┆ date       ┆ str    │
╞═══════════╪═══════╪════════════╪════════╡
│ 1/1/2023  ┆ 15    ┆ 2023-01-01 ┆ 012023 │
│ 1/2/2023  ┆ 16    ┆ 2023-02-01 ┆ 022023 │
│ 1/3/2023  ┆ 16    ┆ 2023-03-01 ┆ 032023 │
│ 1/4/2023  ┆ 15    ┆ 2023-04-01 ┆ 042023 │
│ 1/5/2023  ┆ 14    ┆ 2023-05-01 ┆ 052023 │
│ 1/6/2023  ┆ 13    ┆ 2023-06-01 ┆ 062023 │
│ 1/7/2023  ┆ 14    ┆ 2023-07-01 ┆ 072023 │
│ 1/8/2023  ┆ 17    ┆ 2023-08-01 ┆ 082023 │
│ 1/9/2023  ┆ 16    ┆ 2023-09-01 ┆ 092023 │
│ 1/10/2023 ┆ 18    ┆ 2023-10-01 ┆ 102023 │
└───────────┴───────┴────────────┴────────┘

Change Column Data Type

df = df.with_columns(Price = df['Price'].cast(pl.Utf8))
print(df)
df = df.with_columns(Price = df['Price'].cast(str))
print(df)
df = df.with_columns(Price = df['Price'].cast(pl.Float64))
print(df)

shape: (10, 4)
┌───────────┬───────┬────────────┬────────┐
│ Date      ┆ Price ┆ Date2      ┆ Date3  │
│ ---       ┆ ---   ┆ ---        ┆ ---    │
│ str       ┆ str   ┆ date       ┆ str    │
╞═══════════╪═══════╪════════════╪════════╡
│ 1/1/2023  ┆ 15    ┆ 2023-01-01 ┆ 012023 │
│ 1/2/2023  ┆ 16    ┆ 2023-02-01 ┆ 022023 │
│ 1/3/2023  ┆ 16    ┆ 2023-03-01 ┆ 032023 │
│ 1/4/2023  ┆ 15    ┆ 2023-04-01 ┆ 042023 │
│ 1/5/2023  ┆ 14    ┆ 2023-05-01 ┆ 052023 │
│ 1/6/2023  ┆ 13    ┆ 2023-06-01 ┆ 062023 │
│ 1/7/2023  ┆ 14    ┆ 2023-07-01 ┆ 072023 │
│ 1/8/2023  ┆ 17    ┆ 2023-08-01 ┆ 082023 │
│ 1/9/2023  ┆ 16    ┆ 2023-09-01 ┆ 092023 │
│ 1/10/2023 ┆ 18    ┆ 2023-10-01 ┆ 102023 │
└───────────┴───────┴────────────┴────────┘
shape: (10, 4)
┌───────────┬───────┬────────────┬────────┐
│ Date      ┆ Price ┆ Date2      ┆ Date3  │
│ ---       ┆ ---   ┆ ---        ┆ ---    │
│ str       ┆ str   ┆ date       ┆ str    │
╞═══════════╪═══════╪════════════╪════════╡
│ 1/1/2023  ┆ 15    ┆ 2023-01-01 ┆ 012023 │
│ 1/2/2023  ┆ 16    ┆ 2023-02-01 ┆ 022023 │
│ 1/3/2023  ┆ 16    ┆ 2023-03-01 ┆ 032023 │
│ 1/4/2023  ┆ 15    ┆ 2023-04-01 ┆ 042023 │
│ 1/5/2023  ┆ 14    ┆ 2023-05-01 ┆ 052023 │
│ 1/6/2023  ┆ 13    ┆ 2023-06-01 ┆ 062023 │
│ 1/7/2023  ┆ 14    ┆ 2023-07-01 ┆ 072023 │
│ 1/8/2023  ┆ 17    ┆ 2023-08-01 ┆ 082023 │
│ 1/9/2023  ┆ 16    ┆ 2023-09-01 ┆ 092023 │
│ 1/10/2023 ┆ 18    ┆ 2023-10-01 ┆ 102023 │
└───────────┴───────┴────────────┴────────┘
shape: (10, 4)
┌───────────┬───────┬────────────┬────────┐
│ Date      ┆ Price ┆ Date2      ┆ Date3  │
│ ---       ┆ ---   ┆ ---        ┆ ---    │
│ str       ┆ f64   ┆ date       ┆ str    │
╞═══════════╪═══════╪════════════╪════════╡
│ 1/1/2023  ┆ 15.0  ┆ 2023-01-01 ┆ 012023 │
│ 1/2/2023  ┆ 16.0  ┆ 2023-02-01 ┆ 022023 │
│ 1/3/2023  ┆ 16.0  ┆ 2023-03-01 ┆ 032023 │
│ 1/4/2023  ┆ 15.0  ┆ 2023-04-01 ┆ 042023 │
│ 1/5/2023  ┆ 14.0  ┆ 2023-05-01 ┆ 052023 │
│ 1/6/2023  ┆ 13.0  ┆ 2023-06-01 ┆ 062023 │
│ 1/7/2023  ┆ 14.0  ┆ 2023-07-01 ┆ 072023 │
│ 1/8/2023  ┆ 17.0  ┆ 2023-08-01 ┆ 082023 │
│ 1/9/2023  ┆ 16.0  ┆ 2023-09-01 ┆ 092023 │
│ 1/10/2023 ┆ 18.0  ┆ 2023-10-01 ┆ 102023 │
└───────────┴───────┴────────────┴────────┘

Shift Values

df = df.with_columns(Shifted = df['Date3'].shift(1)
                    .fill_null('SHIFTED'))
print(df)

shape: (10, 5)
┌───────────┬───────┬────────────┬────────┬─────────┐
│ Date      ┆ Price ┆ Date2      ┆ Date3  ┆ Shifted │
│ ---       ┆ ---   ┆ ---        ┆ ---    ┆ ---     │
│ str       ┆ f64   ┆ date       ┆ str    ┆ str     │
╞═══════════╪═══════╪════════════╪════════╪═════════╡
│ 1/1/2023  ┆ 15.0  ┆ 2023-01-01 ┆ 012023 ┆ SHIFTED │
│ 1/2/2023  ┆ 16.0  ┆ 2023-02-01 ┆ 022023 ┆ 012023  │
│ 1/3/2023  ┆ 16.0  ┆ 2023-03-01 ┆ 032023 ┆ 022023  │
│ 1/4/2023  ┆ 15.0  ┆ 2023-04-01 ┆ 042023 ┆ 032023  │
│ 1/5/2023  ┆ 14.0  ┆ 2023-05-01 ┆ 052023 ┆ 042023  │
│ 1/6/2023  ┆ 13.0  ┆ 2023-06-01 ┆ 062023 ┆ 052023  │
│ 1/7/2023  ┆ 14.0  ┆ 2023-07-01 ┆ 072023 ┆ 062023  │
│ 1/8/2023  ┆ 17.0  ┆ 2023-08-01 ┆ 082023 ┆ 072023  │
│ 1/9/2023  ┆ 16.0  ┆ 2023-09-01 ┆ 092023 ┆ 082023  │
│ 1/10/2023 ┆ 18.0  ┆ 2023-10-01 ┆ 102023 ┆ 092023  │
└───────────┴───────┴────────────┴────────┴─────────┘

Export Polars Dataframe To CSV

There are other file formats you can export the DataFrame into, such as write_excel and write_parquet.

# df.write_csv('df.csv')

Read Multi-Sheet Excel File

# Read the first sheet.
# df = pl.read_excel('data2.xlsx')
# Read the second sheet called "Sheet2".
# df = pl.read_excel('data2.xlsx', sheet_name = 'Sheet2')

Reading Text Files

Read a text file with | delimited columns.

# df = pl.read_csv('pipe_sep.txt', seperator = '|')

Reading CSV Files#

Creating New Columns#

Dropping Columns#

Drop Duplicate Rows#

Iterating Polars Dataframes#

Plot Polars Dataframe#

Converting Polars Dataframe To Pandas#

Filtering Polars Dataframes#

Joining/merging Polars Dataframes#

Aggregate Data#

Sorting#

Renaming Columns#

Select Columns#

Fill/Drop Null Values#

Convert String To Date#

Change Column Data Type#

Shift Values#

Export Polars Dataframe To CSV#

Read Multi-Sheet Excel File#

Reading Text Files#