Search code examples
pythongeneratorgoogle-scholar

Python: How to access the elements in a generator object and put them in a Pandas dataframe or in a dictionary?

I am using the scholarly module in python to search for a keyword. I am getting back a generator object as follows:

import pandas as pd
import numpy as np
import scholarly

search_query = scholarly.search_keyword('Python')
print(next(search_query))

{'_filled': False,
 'affiliation': 'Juelich Center for Neutron Science',
 'citedby': 75900,
 'email': '@fz-juelich.de',
 'id': 'zWxqzzAAAAAJ',
 'interests': ['Physics', 'C++', 'Python'],
 'name': 'Gennady Pospelov',
 'url_picture': 'https://scholar.google.com/citations?view_op=medium_photo&user=zWxqzzAAAAAJ'}

I want to access the element 'citedby' but when I try to do next(search_query)['citedby'] it returns TypeError: 'Author' object is not subscriptable.

My question is how can I access elements in the generator object? and How can I convert that object to a Pandas dataframe?

Solution

  • This is not a generator problem. The objects the generator produces are not dictionaries.

    Granted, the scholary library does not help matters by giving the Author instances that you are given a dictionary-like string conversion, and not actually documenting what API that class does support.

    Each of the 'keys' in the Author representation is actually an attribute on the object:

    author = next(search_query)
print(author.citedby)

    You can get a dictionary for the object by using the vars() function:

    author_dict = vars(author)

    The data doesn't necessarily map to a dataframe directly, though. How would the interests list be represented in the dataframe tabular data structure, for example? And you wouldn't want to include the _filled internal attribute either (that's a flag to record if author.fill() has been called yet).

    That said, you could just create a dataframe from the dictionaries by mapping the generator over the vars function:

    search_query = scholarly.search_keyword('Python')
df = pd.DataFrame(map(vars, search_query))

    and then drop the _filled column if necessary, and convert the interests column into something a bit more structured, such as separate columns with 0 / 1 values or similar.

    Note that this is going to be slow, because the scholarly library pages through the Google search results sequentially, and the library deliberately delays requests with a random sleep interval of 5-10 seconds each time to avoid Google blocking the requests. So you'll have to be patient as the Python keyword search easily produces nearly 30 pages of results.