Data sources and ingestion¶

ASRQuant separates data acquisition from numerical research. Remote calls occur only when the user explicitly constructs or invokes a provider.

Local data¶

Supported through QuantLab.from_file(...) or asrquant.data.load_prices(...):

Format	Extension	Optional dependency
CSV/text	`.csv`, `.txt`	none
Parquet	`.parquet`, `.pq`	`pyarrow`
Excel	`.xlsx`, `.xls`	`openpyxl` or another pandas engine
JSON	`.json`	none
Feather	`.feather`, `.ft`	`pyarrow`
SQL	connection + query	database driver chosen by the user

The default parser treats the first column as the timestamp unless date_column is supplied. It converts selected value columns to floats, sorts timestamps, and rejects duplicate timestamps or infinite values.

Canonical price-panel contract¶

A QuantLab price panel has:

a unique DatetimeIndex;
one column per asset;
positive finite prices;
a declared missing-data policy: raise, drop, or ffill;
no implicit network request or hidden field selection.

Use lab.quality or data_quality_report(...) to inspect row count, date range, missingness, duplicate timestamps, sampling gaps, and constant columns.

OHLCV data¶

validate_ohlcv(...) checks canonical Open, High, Low, and Close columns and optional non-negative Volume. resample_ohlcv(...) uses finance-consistent aggregation:

Open: first;
High: maximum;
Low: minimum;
Close: last;
Volume: sum;
other fields: last.

Remote providers¶

Alpha Vantage¶

AlphaVantageProvider supports daily, weekly, monthly, and intraday equity time series. Supply an API key through the constructor or ALPHAVANTAGE_API_KEY.

lab = QuantLab.from_provider(
    "alpha_vantage",
    "IBM",
    provider_kwargs={"api_key": "..."},
    interval="daily",
)

Binance¶

BinanceProvider retrieves public Spot kline/candlestick data and requires no key for the implemented endpoint.

lab = QuantLab.from_provider("binance", "BTCUSDT", interval="1h", limit=1000)

FRED¶

FREDProvider retrieves economic series. Supply a key through the constructor or FRED_API_KEY.

lab = QuantLab.from_provider(
    "fred",
    ["DGS10", "DGS2"],
    field="Value",
    provider_kwargs={"api_key": "..."},
    observation_start="2015-01-01",
)

Yahoo¶

YahooProvider is an optional convenience adapter based on yfinance.

pip install "asrquant[data]"

lab = QuantLab.from_provider("yahoo", ["SPY", "QQQ"], start="2018-01-01")

Near-real-time polling¶

PollingFeed repeatedly invokes a provider's latest-observation method at an explicit interval. It stamps each observation with received_at.

provider = BinanceProvider()
feed = PollingFeed(provider, "BTCUSDT", interval_seconds=60)
for quote in feed.stream(max_updates=5, interval="1m", limit=2):
    print(quote)

Polling is not a WebSocket, exchange gateway, broker connection, or order router. It is suitable for research dashboards and simple data collection, subject to provider limits and terms.

Reproducibility and licensing¶

For every remote dataset, record:

provider and endpoint;
symbol or series identifier;
fields and adjustment convention;
requested interval and date range;
timezone and timestamp interpretation;
retrieval timestamp;
entitlement, redistribution, and citation requirements;
raw response or immutable snapshot when permitted;
data fingerprint after normalization.

Provider availability, schemas, prices, entitlements, rate limits, and revision policies are upstream properties and may change independently of ASRQuant.

ECB Data Portal¶

import asrquant as asr

ecb = asr.ECBProvider()
curve_history = ecb.yield_curve_history(
    maturities=("3M", "6M", "1Y", "2Y", "5Y", "10Y", "30Y"),
    start="2020-01-01",
)

The generic ECBProvider.history() accepts an ECB dataflow and series key. The yield-curve convenience method uses the YC family, aligns maturities and converts percent-per-annum observations to decimals. Network access remains explicit.