Creating a Description

This tutorial will show you how to create a customized description from dataframe columns with a jinja template. We will start by using a simple, hard coded, jinja template then show how to create and use a dynamic template.

We will largely be working with the create.jinja wrangle if you are unfamiliar with it.

Hard Coded Template

The easiest way to create a new description is to hard code a jinja template that can be used across the entire dataset. We will use the table below as a example.

Brand	Product Type	Features
Bosch	Angle Grinder	4.5", 120v
Makita	Impact Driver
DeWalt	Reciprocating Saw	20v
Milwaukee	Orbital Sander	With dust extraction

wrangles:
  - rename:
      Product Type: Product_Type
      
  - create.jinja:
      output: Description
      template:
        string: {{ Brand }} {{ Product_Type }} ({{ Features }})
        
  - rename:
      Product_Type: Product Type

In the recipe above, we first renamed the Product Type column to Product_Type because jinja does not allow variables to contain spaces. The jinja templated uses variables (wrapped in double squigly brackets) which correspond to column names. Notice that anything not wrapped in double squigly brackets is passed through to the output, in this case parenthesis.

Brand	Product Type	Features	Description
Bosch	Angle Grinder	4.5", 120v	Bosch Angle Grinder (4.5", 120v)
Makita	Impact Driver		Makita Impact Driver ()
DeWalt	Reciprocating Saw	20v	DeWalt Reciprocating Saw (20v)
Milwaukee	Orbital Sander	With dust extraction	Milwaukee Orbital Sander (With dust extraction)

As you can see in the second row, the Features column is empty which resulted in an empty set of parenthesis. Let's add an if statement to our template to deal with empty Features.

wrangles:
  - rename:
      Product Type: Product_Type
      
  - create.jinja:
      output: Description
      template:
        string: {{ Brand }} {{ Product_Type }} {% if Features %} ({{ Features }}) {% endif %}
        
  - rename:
      Product_Type: Product Type

In the new template, we used a conditional control structure (ie if statement) to keep from outputting empty parenthesis. In jinja, control structures are wrapped in "{% %}".

Brand	Product Type	Features	Description
Bosch	Angle Grinder	4.5", 120v	Bosch Angle Grinder (4.5", 120v)
Makita	Impact Driver		Makita Impact Driver
DeWalt	Reciprocating Saw	20v	DeWalt Reciprocating Saw (20v)
Milwaukee	Orbital Sander	With dust extraction	Milwaukee Orbital Sander (With dust extraction)

Dynamic Template

By emplementing yaml aliases, along with a custom python function, we can utilize different templates for different categories of data. In this example, we will use Product Category to decide what sort of template we want to emplement.

The table below illustrates our example data:

Brand	Product Type	Product Category	Voltage	Handle Length	TPI
Makita	Impact Driver	Power Tools	18v
Klein	Diagonal Cutting Pliers	Hand Tools		4.5"
Diablo	Reciprocating Saw Blade	Power Tool Accessories			11

alias:
  - &Templates {
      Power Tools: '{{ Brand }} {{ Product_Type }} ({{ Voltage }})',
      Hand Tools: '{{ Brand }} {{ Product_Type }} ({{ Handle_Length }})',
      Power Tool Accessories: '{{ Brand }} {{ Product_Type }} ({{ TPI }}tpi)'
  }

wrangles:
  - rename:
      Product Type: Product_Type
      Product Category: Product_Category
      Handle Length: Handle_Length

  - custom.template_builder:
      category: Product_Category
      alias: *Templates
      output: Template

  - create.jinja:
      output: Description
      template: 
        column: Template

  - rename:
      Product_Type: Product Type
      Product_Category: Product Category
      Handle_Length: Handle Length

write:
  - dataframe:
      not_columns:
        - Template

def template_builder(df, category, alias, output):
    temp_list = []
    for i in range(len(df)):
        temp_list.append(alias[df[Category][i]])
    df[output] = temp_list
    return df

df = wrangles.recipe.run(recipe='temporary.wrgl.yml', dataframe=data, functions=template_builder)

In the above code, a custom function (template_builder) is used to step into the alias and output a column of templates based on category.

Brand	Product Type	Product Category	Voltage	Handle Length	TPI	Description
Makita	Impact Driver	Power Tools	18v			Makita Impact Driver (18v)
Klein	Diagonal Cutting Pliers	Hand Tools		4.5"		Klein Diagonal Cutting Pliers (4.5")
Diablo	Reciprocating Saw Blade	Power Tool Accessories			11	Diablo Reciprocating Saw Blade (11tpi)

As you can see in our output, each product utilizes a different template based on their respective category.

There are many ways to implement jinja templates to write descriptions, and this is a very simple example. For loops, if statements and many more python-like functions can be used in jinja to taylor your output to your needs. See the jinja documentation if this example does not fit your needs.