Parquet is a file format for persisting structured data. It compresses data, preserves data types, and offers faster reads and writes than CSV or JSON. section 11.2.3
Article data pipeline
News fetchers standardize article records into pandas DataFrames and save the results as Parquet files. section 11.2.3
Listing 11.1 NewsAPI fetcher with Parquet persistence
from newsapi import NewsApiClient
import pandas as pd
from datetime import datetime, timedelta
def fetch_newsapi(keywords, days_back=30, api_key='your-api-key'):
newsapi = NewsApiClient(api_key=api_key)
from_date = (datetime.now() - timedelta(days=days_back)).strftime('%Y-%m-%d')
to_date = datetime.now().strftime('%Y-%m-%d')
query = ' OR '.join([f'"{kw}"' if ' ' in kw else kw for kw in keywords])
response = newsapi.get_everything(
q=query,
from_param=from_date,
to=to_date,
language='en',
sort_by='publishedAt',
page_size=100
) #1
rows = []
for article in response.get('articles', []):
rows.append({
'collection_date': datetime.now().strftime('%Y-%m-%d'),
'title': article.get('title', ''),
'description': article.get('description', ''),
'url': article.get('url', ''),
'source_name': article.get('source', {}).get('name', 'Unknown'),
'published_at': article.get('publishedAt'),
'published_dt': pd.to_datetime(article.get('publishedAt'))
})
df = pd.DataFrame(rows)
df['data_source'] = 'NewsAPI'
return df.drop_duplicates(subset=['url']).reset_index(drop=True)
# Execute and persist (collecting from August 2025 onward)
keywords = ['uranium', 'nuclear energy', 'Kazatomprom', 'Cameco']
df_newsapi = fetch_newsapi(keywords=keywords, days_back=30, api_key='your-api-key')
df_newsapi.to_parquet('newsapi_recent.parquet')
print(f"Saved {len(df_newsapi)} NewsAPI articles")
See a NewsAPI fetcher that collects articles and persists its output as Parquet. section 11.2.3
