Scrapped data from https://ereceipt.tn.gov.in/cmprf/Interface/CMPRF/MonthWiseReport
scrapper.py
contains the working scraper using Python Selenium and Pandas.scrapper
folder contains a Scrapy spider which attempted to do the same. The pagination couldn't be handled and will probably scrap the data faster if pagination is figured out.
All the data is in the data
folder as CSV files named <year>_<month>.csv
. These are dumps from Pandas Data Frames and might have inconsistent column ordering. Best used for programmatic manipulation.