Aplicando transformações personalizadas

Transformação de Dados com Polars

Liam Brannigan

Data Scientist & Polars Contributor

Conjunto de dados: Venues

venues = pl.read_csv("venues.csv")

shape: (4, 6)
| business         | location    | type       | hygiene_rating | review | price |
| ---              | ---         | ---        | ---            | ---    | ---   |
| str              | str         | str        | i64            | f64    | i64   |
|------------------|-------------|------------|----------------|--------|-------|
| 7burgers         | Wakey Wakey | restaurant | 4              | 4.2    | 15    |
| Bang Bang Burger | Forest Rd.  | restaurant | 3              | 3.8    | 12    |
| Costa Coffee     | City Point  | café       | 5              | 4.5    | 8     |
| The Queens Head  | Denman St.  | bar        | 5              | 4.7    | 25    |

Tarefa: redimensionar a coluna review

Redimensionar reviews

def rescale_review(x):
    return 2 * x

Redimensionar reviews

def rescale_review(x):
    return 2 * x

venues.with_columns(
    pl.col("review")
)

Redimensionar reviews

def rescale_review(x):
    return 2 * x

venues.with_columns(
    pl.col("review").map_elements(rescale_review)
)

shape: (4, 6)
| business         | location    | type       | hygiene_rating | review | price |
| ---              | ---         | ---        | ---            | ---    | ---   |
| str              | str         | str        | i64            | f64    | i64   |
|------------------|-------------|------------|----------------|--------|-------|
| 7burgers         | Wakey Wakey | restaurant | 4              | 8.4    | 15    |
| Bang Bang Burger | Forest Rd.  | restaurant | 3              | 7.6    | 12    |
| Costa Coffee     | City Point  | café       | 5              | 9.0    | 8     |
| The Queens Head  | Denman St.  | bar        | 5              | 9.4    | 25    |

Especificar return_dtype para controle

venues.with_columns(
    pl.col("review").map_elements(
        rescale_review, return_dtype=pl.Float64
    )
)

shape: (4, 6)
| business         | location    | type       | hygiene_rating | review | price |
| ---              | ---         | ---        | ---            | ---    | ---   |
| str              | str         | str        | i64            | f64    | i64   |
|------------------|-------------|------------|----------------|--------|-------|
| 7burgers         | Wakey Wakey | restaurant | 4              | 8.4    | 15    |
| Bang Bang Burger | Forest Rd.  | restaurant | 3              | 7.6    | 12    |
| Costa Coffee     | City Point  | café       | 5              | 9.0    | 8     |
| The Queens Head  | Denman St.  | bar        | 5              | 9.4    | 25    |

Aviso do Polars

PolarsInefficientMapWarning:

Expr.map_elements é significativamente mais lento que a API de expressões nativa.
Use apenas se você REALMENTE NÃO PUDER implementar sua lógica de outra forma.
Substitua esta expressão...
  - pl.col("review").map_elements(rescale_review)
por esta aqui:
  + 2 * pl.col("review")

Redimensionar reviews (expressão nativa)

venues.with_columns(
    (2 * pl.col("review")).alias("review")
)

shape: (4, 6)
| business         | location    | type       | hygiene_rating | review | price |
| ---              | ---         | ---        | ---            | ---    | ---   |
| str              | str         | str        | i64            | f64    | i64   |
|------------------|-------------|------------|----------------|--------|-------|
| 7burgers         | Wakey Wakey | restaurant | 4              | 8.4    | 15    |
| Bang Bang Burger | Forest Rd.  | restaurant | 3              | 7.6    | 12    |
| Costa Coffee     | City Point  | café       | 5              | 9.0    | 8     |
| The Queens Head  | Denman St.  | bar        | 5              | 9.4    | 25    |

Escolha a ferramenta certa

Expressões nativas

Padrão preferido
Rápidas
Otimizadas

`.map_elements()`

Use quando precisar
Legível
Pacotes de terceiros

Padronizar texto de location

Denman St. -> DENMAN STREET
Forest Rd. -> FOREST ROAD

Tarefa: padronizar o formato da coluna location

Padronizar locations

venues.with_columns(
    pl.col("location")



)

Padronizar locations

venues.with_columns(
    pl.col("location")
    .str.replace_many(["St.", "Rd."], ["Street", "Road"])


)

Padronizar locations

venues.with_columns(
    pl.col("location")
    .str.replace_many(["St.", "Rd."], ["Street", "Road"])
    .str.to_uppercase()
    .alias("location_clean")
)

shape: (4, 7)
| business         | location    | ... | location_clean |
| ---              | ---         | --- | ---            |
| str              | str         | ... | str            |
|------------------|-------------|-----|----------------|
| 7burgers         | Wakey Wakey | ... | WAKEY WAKEY    |
| Bang Bang Burger | Forest Rd.  | ... | FOREST ROAD    |
| Costa Coffee     | City Point  | ... | CITY POINT     |
| The Queens Head  | Denman St.  | ... | DENMAN STREET  |

Armazenar uma expressão em variável

standardize_locations_expr = (
    pl.col("location")
    .str.replace_many(["St.", "Rd."], ["Street", "Road"])
    .str.to_uppercase()
    .alias("location_clean")
)


type(standardize_locations_expr)

polars.Expr

Reutilizar a expressão

venues.with_columns(
    standardize_locations_expr
)

shape: (4, 7)
| business         | location    | ... | location_clean |
| ---              | ---         | --- | ---            |
| str              | str         | ... | str            |
|------------------|-------------|-----|----------------|
| 7burgers         | Wakey Wakey | ... | WAKEY WAKEY    |
| Bang Bang Burger | Forest Rd.  | ... | FOREST ROAD    |
| Costa Coffee     | City Point  | ... | CITY POINT     |
| The Queens Head  | Denman St.  | ... | DENMAN STREET  |

Adicionar um método de expressão customizado

def standardize(input):

Adicionar um método de expressão customizado

def standardize(input):
    return (
        input
        .str.replace_many(["St.", "Rd."], ["Street", "Road"])
        .str.to_uppercase()
    )

Adicionar um método de expressão customizado

def standardize(input):
    return (
        input
        .str.replace_many(["St.", "Rd."], ["Street", "Road"])
        .str.to_uppercase()
    )

pl.Expr.standardize = standardize

Usar o método customizado

restaurants.with_columns(
    pl.col("address").standardize().alias("address_clean")
)

shape: (4, 7)
| business         | address     | ... | address_clean  |
| ---              | ---         | --- | ---            |
| str              | str         | ... | str            |
|------------------|-------------|-----|----------------|
| 7burgers         | Wakey Wakey | ... | WAKEY WAKEY    |
| Bang Bang Burger | Forest Rd.  | ... | FOREST ROAD    |
| Costa Coffee     | City Point  | ... | CITY POINT     |
| The Queens Head  | Denman St.  | ... | DENMAN STREET  |

Vamos praticar!

Transformação de Dados com Polars