How does pandas Dataframe.loc accept the [...] syntax?

Question

I have read this documentation:

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.loc.html

You can use a syntax like df.loc[df['shield'] > 6, ['max_speed']].

I tried using Github and found out:

Suppose you have a pandas.core.frame.DataFrame object, i.e. a DataFrame called df.

The type of df.loc is pandas.core.indexing._LocIndexer.

Nevertheless, I could not sort out these questions:

How do you make a Python function/class accepting a syntax like above?
Where in the source code of pandas.core.frame.DataFrame is the property self.loc defined??

Look at this file on github. github.com/pandas-dev/pandas/blob/master/pandas/core/frame.py and this github.com/pandas-dev/pandas/blob/… — Poojan
– Poojan, Commented Jul 29, 2019 at 12:33
@Poojan Thanks, but where does get_loc become self.loc?? Sorry for being so stupid, but I don't get it :/ — Michael S
– Michael S, Commented Jul 29, 2019 at 12:45
@Poojan I looked at the get_loc function but I cannot find any @property decorator or any other syntax defining self.loc. I git cloned the pandas repo. Used git grep -n self.loc and git grep -A1 @property | grep loc but I can't find it. Could you solve my issue? — Michael S
– Michael S, Commented Jul 29, 2019 at 14:17
I am unable to find exact location of code where .loc is implemented but it is implemented somewhere in this file. github.com/pandas-dev/pandas/blob/master/pandas/core/…. Implementation of loc is more complex. You can search in indexing.py file for some in depth explanation . Write answer if you find a good explanation for this. — Poojan
– Poojan, Commented Jul 29, 2019 at 15:29

Konrad Rudolph · Accepted Answer · 2021-10-28 16:35:31Z

6

How you make a class accept that syntax in general is by implementing __getitem__ which is an example of operator overloading. This allows an object of that class to be indexed with []. For example:
```
class get_item_example(object):
 def __getitem__(self, key):
         print(key)
```
Try it out:
```
>>> gi = get_item_example()
>>> gi['a']
a
>>> gi[['a','b','c']]
['a', 'b', 'c']
>>> gi['a','b','c']
('a', 'b', 'c')
```
In the case of df.loc[df['shield'] > 6, ['max_speed']] what happens is that the key passed to __getitem__ is a tuple containing the pandas series returned by df['shield'] > 6 and the single item list ['max_speed'].
In the pandas source, pandas.core.indexing._LocIndexer inherits an implementation of __getitem__ from pandas.core.indexing. _LocationIndexer. The implementation is here: https://github.com/pandas-dev/pandas/blob/61362be9ea4d69b33ae421f1f98b8db50be611a2/pandas/core/indexing.py#L1374

edited Oct 28, 2021 at 16:35

Konrad Rudolph

549k142 gold badges967 silver badges1.3k bronze badges

answered Mar 28, 2020 at 21:16

Alex L

611 silver badge3 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

user3058724 · Accepted Answer · 2021-10-28 16:29:24Z

0

If you want to have a property that has indexer (df.loc[] vs. df[]), you can create another class that will process the indexer and that gets created and passed back when the property is accessed.

class Helper(self)
   def __init__(self, parent):
       self.parent = parent

   def __getitem__(self, idx):
       ### do your indexing here

class Parent(self):
   ...

   @property
   def fancy_indexer(self):
       return Helper(self)

answered Oct 28, 2021 at 16:29

user3058724

1

Collectives™ on Stack Overflow

How does pandas Dataframe.loc accept the [...] syntax?

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related