pitedgar 0.1.0

Point-in-time SEC EDGAR financial data pipeline

pitedgar

Point-in-time SEC EDGAR financial data pipeline.

Downloads SEC EDGAR companyfacts.zip, parses XBRL JSON facts into a local parquet file, and exposes a query API with zero look-ahead bias — every value is stamped with the filed date (when the data was actually available to the market), not the period-end date.

---

Installation

pip install pitedgar
# or with Poetry
poetry install

---

Quick start

from pathlib import Path
from pitedgar import PitEdgarConfig, build_cik_map, download_bulk, parse_all, PitQuery

config = PitEdgarConfig(
    edgar_identity="Mario Rossi [e-mail]",  # required by SEC
    data_dir=Path("./data"),
)

# Step 1 — one-shot ticker → CIK mapping
tickers = ["AAPL", "MSFT", "JPM", "GOOGL"]
cik_map = build_cik_map(tickers, config)

# Step 2 — download ~1.5 GB bulk ZIP (do this periodically, not every run)
download_bulk(config)

# Step 3 — parse JSON → parquet (sub-minute for 500 companies)
master = parse_all(config, cik_map)

# Step 4 — query
q = PitQuery(config.data_dir / "pit_financials.parquet")

# What revenue figure was available to the market on 2022-06-30?
result = q.as_of(["AAPL", "MSFT"], "us-gaap:Revenues", "2022-06-30")

# Full history
hist = q.history("AAPL", "us-gaap:NetIncomeLoss", freq="A")

# Portfolio cross-section signal
xs = q.cross_section("us-gaap:NetIncomeLoss", "2023-12-31")

---

CLI

# Resolve tickers (tickers.txt has one ticker per line)
pitedgar map --tickers tickers.txt --identity "Name [e-mail]"

# Download bulk ZIP
pitedgar fetch --identity "Name [e-mail]"

# Parse to parquet
pitedgar build --identity "Name [e-mail]"

# Query a single value
pitedgar query --ticker AAPL --concept us-gaap:Revenues --as-of 2023-06-30

---

Key design decisions

Decision	Rationale
filed as PIT timestamp	The date the filing was submitted to SEC — this is when information became public
Deduplication keeps latest filed per (concept, end)	Companies sometimes refile restated figures; keep the superseding value
Raw USD values, no scale conversion	SEC reports values as-filed; downstream code applies any needed normalization
Local parquet, no runtime HTTP	Queries run at DataFrame speed with no network dependency

---

Supported XBRL concepts (defaults)

See pitedgar.config.DEFAULT_CONCEPTS for the full list, which includes revenues, net income, assets, liabilities, equity, EPS, cash, debt, operating cash flow, capex, and R&D expense.

---

Examples

• examples/fcf_sp500.py — S&P 500 free cash flow benchmark: fetches constituents, builds the parquet, and queries FCF cross-sections across 20 quarters. Useful as an end-to-end performance reference.

---

Contributing

Contributions are welcome. See CONTRIBUTING.md for setup instructions, coding conventions, and the PR process.

---

License

MIT