How saisantoshv3/electoral_bonds works
What this project is (type + purpose)
This repo is a data-wrangling / dataset-building pipeline (not a running service). Its purpose is to produce donor- and recipient-wise “matched” electoral bond transactions by linking two SBI/ECI datasets via Electoral Bond Code. The README explicitly mentions:
• PDFs as upstream source (ECI site; SBI-provided)
• Tabula used to extract CSVs from PDFs (manual/external step)
• OpenRefine used for clustering/deduping purchaser names (manual/external step)
• A Python script that likely performs the deterministic “matching” and outputs cleaned datasets
File tree analysis → components you must reflect
Top-level structure:
• raw_files/
• encashed_parties.pdf
• purchased_companies.pdf
These are the raw upstream artifacts (PDFs). They are *inputs* to the pipeline.
• data/
• encashed.csv
• purchased.csv
• final.csv
This is the intermediate and final data products. encashed.csv and purchased.csv are extracted/cleaned versions of the PDFs; final.csv is the matched dataset output.
• script/
• electoral_bonds.py
• README.md
data/purchased.csv → OpenRefine
• (Optionally also show encashed.csv being cleaned if applicable; README only explicitly mentions purchaser names.)
data/final.csv → End users
This is the core dataflow. Don’t invent APIs/databases—stay faithful to the repo.
Architectural pattern / design principle
Call it what it is:
• Batch ETL / data pipeline
• Artifact-based workflow (files as boundaries)
• Deterministic matching step (join by Electoral Bond Code)
• Human-in-the-loop cleaning (OpenRefine) is a first-class stage
• *Manual steps* vs *Automated steps*
• *Raw* vs *Intermediate* vs *Final* data
Relevant technologies
Explicitly label:
• Tabula (PDF → CSV)
• OpenRefine (data cleaning)
• Python script (electoral_bonds.py)
“Run locally / command-line execution” (since there’s no deployment structure indicated).
What NOT to include (to keep it accurate)
• Don’t imply OpenRefine/Tabula are Python dependencies; they’re external tools in the workflow.