与for循环相比，选择数据帧不同部分的方法更快？

Question

I have a dataframe of instacart orders

order_id    product_id  add_to_cart_order   reordered   product_name    
32          49683       7                   1           Cucumber Kirby  
52          49683       4                   1           Cucumber Kirby  
88          49683       20                  0           Cucumber Kirby  
95          49683       12                  1           Cucumber Kirby  
111         49683       5                   1           Cucumber Kirby  

reordered is either 1 or 0 indicating whether the customer had ordered that product in a previous order or not.

I want to get info on a per-product basis, for example I would like to know which products have the most number of reorders (among other things). The only way I can think of to do this is to iterate through the dataframe, selecting only rows by product name one at a time, and summing the values of reordered for each product. Only problem is there are about 92k different products and this is breaking my computer and taking forever. Here's my code. I'm saving the results to a dictionary but I'm open to other approaches. There must be a more efficient way to do this?

reordersums = {}
for product in list(products.product_name):
# Select the rows whose product name matches the product we are checking, sum the values in column "reordered"
    reordersum = order_products[order_products.product_name == product].reordered.sum()

    reordersums[product]=reordersum    
print(reordersums)

Answer 1

Try using the group_by interface:

# Group up the dataframe by product
group_products = products.groupby('product_name')

# Sum the groups on the reordered column
reordered_sums = group_products['reordered'].agg('sum')

Answer 2

Please try below, However i'm not sure it this is what you are looking for:

Your illustrated DataFrame Structure:

   order_id  product_id  add_to_cart_order  reordered    product_name
0        32       49683                  7          1  Cucumber Kirby
1        52       49683                  4          1  Cucumber Kirby
2        88       49683                 20          0  Cucumber Kirby
3        95       49683                 12          1  Cucumber Kirby
4       111       49683                  5          1  Cucumber Kirby

Solution: groupby + DataFrame.filter + sum()

>>> df.groupby('reordered').filter(lambda x: len(x) > 1).groupby(['product_name']).sum().reset_index()
     product_name  order_id  product_id  add_to_cart_order  reordered
0  Cucumber Kirby       290      198732                 28          4

OR , as suggested by @Amit in comment Section.

>>> df[df.reordered==1].groupby('product_name').sum().reset_index()
     product_name  order_id  product_id  add_to_cart_order  reordered
0  Cucumber Kirby       290      198732                 28          4

OR , In case you want only to see product_name & reordered

df.set_index('product_name').reordered.ge(1).sum(level=0).astype(int).reset_index()
     product_name  reordered
0  Cucumber Kirby          4