Dataframe and many-to-many relationship

[evdoks]

Is it possible to use django-pandas with models that stay in many-to-many relationship? It seems that currently this feature is not supported. E.g., for

class Topping(models.Model):
    name = models.CharField(max_length=30)

class Pizza(models.Model):
    name = models.CharField(max_length=50)
    toppings = models.ManyToManyField(Topping)

qs_pizza_with_toppings = Pizza.objects.all().prefetch_related('toppings')
df_pizza_with_toppings = read_frame(qs_pizza_with_toppings)

df_pizza_with_toppings does not contain any topping names.

[paulochf]

I think I'm having the same problem.

I have models like

class Category(models.Model):
    name    = models.CharField()
    showing = models.BooleanField(default=False)

    objects = DataFrameManager()

class Food(models.Model):
    name    = models.CharField()

    objects = DataFrameManager()

class FoodCategory(models.Model):
    food     = models.ForeignKey('Food')
    category = models.ForeignKey('Category')
    primary  = models.BooleanField(default=False)

    objects  = DataFrameManager()

Then, I try

foods = FoodCategory.objects.only("food__name").filter(category__showing=True)
df = foods.to_dataframe()

df.head()

which prints

id  food category    is_primary
0   1   Food object    Category object   True
1   2   Food object    Category object   True
2   4   Food object    Category object   True
3   5   Food object    Category object   True
4   6   Food object    Category object   True

and then I cannot/don't know how to access the inner object.

df.loc[:,"food"]

0     Food object
1     Food object
2     Food object
3     Food object
...

How to access Food name attribute?

Thanks!

[bufke]

Same for reverse foreign key relations.

[chrisdev]

@evdoks sorry for the late response. You are correct!! I'll fix in in the next release. In the mean time here is a workaround. Using pandas and values_list QuerySets

import pandas as pd
pd.DataFrame.from_records(list(Pizza.objects.all().values_list('name', 'toppings__name')))

[chrisdev]

@paulochf @bufke again sorry for the very tardy response.
I'm working on an IPython notebook/blog post with some examples of usage.

df = FoodCategory.objects.only("food__name").filter(category__showing=False).to_dataframe()
# Show the frame to see the column names 
df.head()

#to get the food column you specify a list with the column names that you want ['food']
df[['food']]
#to get food and primary
df[['food, 'primary']]

Note, you must define the __str__ method in your Models for the verbose resolution to work
That's why in your example above you have Food Object in the rows of your pandas DataFrame.

@python_2_unicode_compatible
class Category(models.Model):
    name = models.CharField(max_length=200)
    showing = models.BooleanField(default=False)

    objects = DataFrameManager()
    def __str__(self):
        return "{0}-{1}".format(
            self.name,
            self.showing
        )

@python_2_unicode_compatible
class Food(models.Model):
    name = models.CharField(max_length=200)

    objects = DataFrameManager()

    def __str__(self):
        return "{}".format(self.name)

Also you can get a DateFrame with exactly the columns you want by specifying them in a list as the first or fieldnames argument in the to_dataframe method. You can use __ for relationship spanning

FoodCategory.objects.filter(category__showing=False).to_dataframe(
    ['category__name', 'food__name', 'primary']
)

[paulochf]

Nice, @chrisdev! Thank you so much for explaining this!