marketgoblin

Download, store, and load financial market data — fast, without fuss.

pip install marketgoblin

---

Features

• Multi-dataset fetch & persist

  ---

  Download OHLCV, shares-outstanding, or dividends by symbol and date range via a Dataset enum. When save_path is set, data is automatically sliced into monthly Parquet files on disk.

  API Reference

• Lazy by default

  ---

  Built on Polars. Every data path returns a pl.LazyFrame — nothing is computed until you call .collect().

• Batch fetching

  ---

  fetch_many() uses a ThreadPoolExecutor with a token-bucket rate limiter. Failed symbols are logged and skipped — they never crash the batch.

• Pluggable sources

  ---

  Subclass BaseSource, implement one method, register in one line. Ships with YahooSource and TiingoSource out of the box.

• Reliable by default

  ---

  YahooSource retries transient failures with exponential backoff. Writes are atomic (.tmp rename). JSON sidecars record metadata per slice.

• Predictable storage layout

  ---

  Uniform across datasets: {save_path}/{provider}/{dataset}/{SYMBOL}/{SYMBOL}_{YYYY-MM}.pq. OHLCV adjusted + raw share one file, distinguished by the is_adjusted column. Every slice is inspectable and portable with any Parquet reader.

---

Quick start

Fetch & saveShares outstandingDividendsLoad from diskBatch fetchTiingo fundamentals

import polars as pl
from marketgoblin import MarketGoblin

goblin = MarketGoblin(provider="yahoo", save_path="./data")

# OHLCV is a tidy stacked frame — each day appears twice
# (is_adjusted=True / False). Filter to pick a variant.
lf = goblin.fetch("AAPL", "2024-01-01", "2024-03-31", parse_dates=True)
print(lf.filter(pl.col("is_adjusted")).collect())

from marketgoblin import Dataset, MarketGoblin

goblin = MarketGoblin(provider="yahoo", save_path="./data")

shares = goblin.fetch(
    "AAPL",
    "2024-01-01",
    "2024-03-31",
    dataset=Dataset.SHARES,
    parse_dates=True,
)
print(shares.collect())

from marketgoblin import Dataset, MarketGoblin

goblin = MarketGoblin(provider="yahoo", save_path="./data")

dividends = goblin.fetch(
    "AAPL",
    "2024-01-01",
    "2024-03-31",
    dataset=Dataset.DIVIDENDS,
    parse_dates=True,
)
print(dividends.collect())

from marketgoblin import MarketGoblin

goblin = MarketGoblin(provider="yahoo", save_path="./data")

# No network call — reads straight from Parquet
lf = goblin.load("AAPL", "2024-01-01", "2024-03-31", parse_dates=True)
print(lf.collect())

from marketgoblin import MarketGoblin

goblin = MarketGoblin(provider="yahoo", save_path="./data")

results = goblin.fetch_many(
    ["AAPL", "MSFT", "GOOGL", "NVDA"],
    start="2024-01-01",
    end="2024-03-31",
    max_workers=4,
    requests_per_second=2.0,
)

for symbol, lf in results.items():
    print(f"{symbol}: {lf.collect().height} rows")

from marketgoblin import Dataset, MarketGoblin

# Tiingo covers the full catalog (splits, daily & quarterly fundamentals).
# Reads its key from api_key= or the TIINGO_API_KEY env var.
goblin = MarketGoblin(provider="tiingo", save_path="./data")

statements = goblin.fetch(
    "AAPL",
    "2023-01-01",
    "2024-03-31",
    dataset=Dataset.FUNDAMENTALS_STATEMENTS,
    parse_dates=True,
)
print(statements.collect())

Full API Reference Contributing

---

Data conventions

Property	Value
Date on disk	int32 YYYYMMDD (e.g. 20240101) — use parse_dates=True to get pl.Date
OHLC columns	float32
Volume column	int64
is_adjusted column	bool (OHLCV only — distinguishes adjusted vs raw rows)
Shares column	int64
Dividend column	float32
Parquet path	{save_path}/{provider}/{dataset}/{SYMBOL}/{SYMBOL}_{YYYY-MM}.pq (uniform across datasets)
JSON sidecar	Same path, .json — row count, date range, per-dataset stats (OHLCV also records has_adjusted/has_raw and missing trading days)

Datasets

Dataset	Providers	Notes
Dataset.OHLCV	yahoo, tiingo	Daily open/high/low/close/volume. Tidy stacked: each day appears twice with is_adjusted=True/False. Adjusted Open/High/Low are derived locally from Adj Close / Close (zero numerical drift vs auto_adjust=True, half the network calls).
Dataset.SHARES	yahoo, tiingo	Shares outstanding — sparse, corporate-action-driven cadence. Deduplicated to one row per day (last value wins).
Dataset.DIVIDENDS	yahoo, tiingo	Cash dividend events — typically quarterly. Filtered to the requested [start, end] range.
Dataset.SPLITS	tiingo	Forward/reverse split events as a split_factor multiplier.
Dataset.FUNDAMENTALS_DAILY	tiingo	Daily market cap, enterprise value, and valuation ratios. Paid endpoint.
Dataset.FUNDAMENTALS_STATEMENTS	tiingo	Full quarterly income statement, balance sheet, cash flow & overview — every line item in both as-reported and restated variants. Paid endpoint.

See Providers & Capabilities for the full matrix.