Skip to content

About

Quantitative market analysis toolkit written in Rust

Topics

Resources

Stars

1 star

Watchers

1 watching

Forks

Latest commit

 

History

297 Commits

Folders and files

Repository files navigation

example workflow

quantitative_data_analysis_rs

Quantitative market analysis toolkit written in Rust. It ingests historical stock data plus fundamentals from J-Quants and Yahoo Finance to produce chart-ready signals, summaries, and SVG/JSON artifacts for reporting.

Features

  • Hexagonal-style architecture (domain, use_case, controller, presenter, infrastructure, shared) keeps market logic testable and interchangeable.
  • Trend analysis engine detects crossover patterns, MACO signals, candlestick patterns (marubozu, doji), and aggregates them into summaries.
  • Financial indicator analyzer correlates statements with the latest available prices to generate JSON-friendly metrics.
  • CSV/TSV ingestion utilities plus ready-made assets under assets/ accelerate local experiments.
  • Async-first data access via J-Quants REST API and Yahoo Finance scrapers, including helper tooling for token rotation.

Scoring universe policy

When selecting the universe for scoring from J-Quants master data, we use both ProdCat and Mkt:

  • ProdCat = "011" keeps ordinary domestic listed stocks and excludes ETFs, REITs, foreign stocks, and similar products.
  • Mkt = "0111" | "0112" | "0113" keeps only Prime, Standard, and Growth listings and excludes other market codes such as TOKYO PRO MARKET.

This keeps the scoring candidate set focused on the common Japanese equity universe used by the analyzers.

Installation

  1. Install the Rust toolchain specified in rust-toolchain.toml (Rust 1.90.0). rustup automatically picks it up when you enter the repo.
  2. Install build essentials for your OS (LLVM/Clang on macOS, build-essential on Linux).
  3. Clone the repository and fetch dependencies:
cargo fetch

This repository intentionally tracks Cargo.lock to keep CI and J-Quants connectivity checks reproducible.

Usage

  • Financial indicators:

    cargo run --example financial_indicator
  • Trend analysis (SVG + JSON outputs):

    cargo run --example trend_analysis
  • Company master TSV generator:

    cargo run --example make_companies

Generated artifacts are written to examples/ (SVG charts) and compared against the golden files in assets/ for regression safety.

Configuration

J-Quants financial disclosures

The J-Quants statement repository retrieves every /v2/fins/summary page and returns disclosure rows in ascending disclosure date and time (JST). Rows with the same timestamp retain their source order; disclosure numbers do not establish their chronological order. A failed page, invalid response, or malformed nonempty field is an error, not a partial history. Rate limiting and retries remain the caller's responsibility.

FetchScoringData.financial_disclosures contains the API's FY, quarterly, and forecast-revision rows. Each entry retains its disclosure identifier and document type, period and fiscal-year dates, and operating-profit actual and forecast fields, including the NC variants. Amounts are in the units provided by J-Quants (normally yen). Missing values are null; they do not mean unchanged or withdrawn. FOP refers to the current fiscal year, NxFOP to the next fiscal year. Compare both fiscal-year boundaries before associating forecasts.

The API's OP is preserved with its original period dates and operating_profit_period: "unspecified". Its interpretation as cumulative or single-quarter is not guaranteed here; no single-quarter differences are calculated. The NC-less fields may represent consolidated or individual values depending on document type; neither accounting basis nor consolidation status is inferred when the document does not establish it. J-Quants does not provide the time at which a disclosure became available through its API, and provider corrections can overwrite previously supplied values. These records therefore do not guarantee point-in-time reconstruction. J-Quants also does not guarantee snapshot consistency if data changes during pagination.

latest_statement and full_year_sales remain FY projections, separate from the full history. get_row_statements returns the full disclosure rows, whereas get_statements remains an FY-only projection. Generated statements TSVs now include a disclosure_json column; examples/make_statements.rs writes every disclosure and stops on fetch failure. Existing TSV data lacks the history and must be regenerated before using it with FetchScoringData; it is rejected rather than silently producing incomplete history. The JSON contract adds financial_disclosures and old JSON must be regenerated. The example uses assets/statements.tsv, so regeneration and deployment of that input must be handled separately.

Stock price and CSV input contract

Stock requires adj_open, adj_high, adj_low, adj_close, and adj_volume. Adjusted prices account for splits/consolidations and supported corporate actions, but not cash dividends. Adjusted volume uses the corresponding split/consolidation share basis and is stored as f64 to preserve fractional values. Missing or invalid required values are errors, not a reason to silently drop a row or substitute source values.

  • J-Quants maps AdjO, AdjH, AdjL, AdjC, and AdjVo directly. Its supported price adjustments include rights issues; rights issues do not adjust volume.
  • Yahoo maps its provided OHLC and volume to both the source and adjusted fields. The analysis adj_close is Yahoo's close, not its dividend-adjusted adjclose. No additional split adjustment is applied.
  • Source fields (open, high, low, close, volume) preserve provider values and are not guaranteed to be unadjusted.
  • Yahoo's split-adjusted OHLC and volume were corroborated for AAPL's 2020 4-for-1 split. This is not a verified guarantee for all instruments, consolidations, or other corporate actions. Providers may differ in supported events and rounding; cross-provider numerical equivalence is not guaranteed.
  • Adjusted history may change after subsequent corporate actions. It does not guarantee point-in-time reconstruction.

CSV input requires these columns (the shown order is also the headerless order):

Date,Open,High,Low,Close,Adj Open,Adj High,Adj Low,Adj Close,Volume,Adj Volume

All fields are required. Dates use YYYY-MM-DD, prices must be finite and positive, Volume is a nonnegative integer, and Adj Volume is finite and nonnegative. Legacy seven-column CSVs are rejected; their Adj Close must not be copied blindly because it may include dividend adjustments. The producer is responsible for supplying a consistent adjustment basis; the reader cannot infer it from the numbers.

The stock fixtures assets/2031.T.csv, assets/8473.T.csv, and assets/9223.T.csv were replaced with J-Quants daily bars fetched on 2026-10-24. Their periods are respectively 2024-02-15, 2022-09-09 through 2023-09-08, and 2023-12-25 through 2024-02-16. Fixture-based tests require no API access. Technical analysis uses adjusted OHLC and volume consistently: SMA, MACD, RSI, Bollinger bands, ATR, volume ratios, price/SMA comparisons, crossover returns, high/low direction, candlestick patterns, and chart/table values. Crossover returns are price returns excluding dividends, not total returns. ATR uses adjusted high/low and previous adjusted close with the existing ta EMA smoothing (2 / (period + 1)), not Wilder smoothing. Volume ratios compare the current adjusted volume against the previous N bars, excluding the current bar. Analysis-result volume and chart-view volumes() now use f64 to avoid truncating fractional adjusted volume. Financial valuation inputs and stored source fields are unchanged.

  • Provide J-Quants API key via environment variable before executing any example or binary:

    export JQUANTS_API_KEY="your-jquants-api-key"
  • Optional inputs (dates, markets, crossover filters) are passed directly when calling controllers or editing the example binaries.

Contribution

  • Read RUST_GUIDELINES.md and the referenced Microsoft Pragmatic Rust Guidelines before sending PRs.
  • Adhere to the layered architecture; new logic should live in domain/use_case, while adapters go under infrastructure/presenter.
  • Include documentation/comments for non-obvious algorithms and add/adjust tests in tests/ or under the relevant module.

Testing

Run formatting and the full test suite locally:

cargo fmt --all
cargo clippy --all-targets --all-features
cargo test

Examples double as integration tests; compare their outputs with the JSON/SVG fixtures in assets/ when modifying presenters.

License

The project has not declared an explicit license yet. Please ask the maintainers before using it in production or redistributing.

Acknowledgements

  • J-Quants for providing authenticated access to fundamental and price data.
  • Yahoo Finance for historical market data access.
  • The Rust community and the authors of the crates listed in Cargo.toml for the ecosystem support.

About

Quantitative market analysis toolkit written in Rust

Topics

Resources

Stars

1 star

Watchers

1 watching

Forks

Releases

Packages

Used by

Contributors

Languages