ohlcv-tools

Small CLI toolkit for working with OHLCV/tick market data.

Subcommands

• vwap — volume-weighted average price from a trade/tick CSV
• sma / ema — simple/exponential moving averages over a price series
• returns — simple percentage returns between consecutive prices
• drawdown — max peak-to-trough drawdown over a price series
• stats — min/max/mean/median for a price series
• correlation — Pearson correlation between two equal-length price series
• aggregate — turn tick/trade data into OHLCV candles at a given interval
• resample — re-bucket already-aggregated OHLCV candles into a larger interval
• completions — generate a shell completion script

Every subcommand above (except completions) supports --format json for scripting; see JSON output.

Usage

cargo run -- info
cargo run -- vwap trades.csv
cargo run -- sma prices.csv --window 14
cargo run -- ema prices.csv --period 14
cargo run -- returns prices.csv
cargo run -- drawdown prices.csv
cargo run -- stats prices.csv --format json
cargo run -- correlation a.csv b.csv
cargo run -- aggregate ticks.csv --interval 60
cargo run -- resample candles.csv --interval 300
cargo run -- completions bash > ohlcv-tools.bash

trades.csv must have price and volume columns:

price,volume
100,10
101,5
99,20

prices.csv (for sma/ema) must have a price column:

price
100
101
99

ticks.csv (for aggregate) must have timestamp (unix seconds), price, and volume columns. Ticks don't need to be pre-sorted; output is one candle per line: timestamp,open,high,low,close,volume.

candles.csv (for resample) must have timestamp,open,high,low,close,volume columns — the same shape aggregate produces. Candles don't need to be pre-sorted. Within each new, larger bucket: open comes from the earliest-timestamped candle, close from the latest, high/low are the extremes across all candles in the bucket, and volume is summed. This lets you build higher-timeframe candles (e.g. 1m to 1h) without re-reading the original ticks.

All numeric output is rounded via --precision (default 6 decimal places).

Every subcommand accepts - in place of a file path to read CSV from stdin, e.g. cat trades.csv | ohlcv-tools vwap -.

Every computation subcommand rejects non-finite values (NaN, inf, -inf) in its numeric input columns with a clear error instead of silently propagating them into the result (e.g. a single NaN price would otherwise make vwap/sma/drawdown return NaN with no indication anything went wrong).

JSON output

Every computation subcommand accepts --format json for scripting instead of the default plain-text output:

• Single-value results (vwap, drawdown, correlation) print a named JSON object, e.g. {"vwap":17.5}, {"max_drawdown":0.25}, {"correlation":1.0}.
• Series results (sma, ema, returns) print a JSON array of numbers, e.g. [1.5,2.5,3.5].
• aggregate prints a JSON array of candle objects, e.g. [{"timestamp":0,"open":100.0,"high":101.0,"low":100.0,"close":101.0,"volume":3.0}].
• stats prints a JSON object with min/max/mean/median keys.

cargo run -- vwap trades.csv --format json
cargo run -- sma prices.csv --window 14 --format json
cargo run -- aggregate ticks.csv --interval 60 --format json

Development

cargo build
cargo test

Tests include unit tests per module, property-based tests (proptest) for numeric invariants, plus tests/cli.rs, which exercises the compiled binary end-to-end (argument parsing, exit codes, real stdout/stderr output).

cargo bench

Runs the criterion benchmark suite in benches/.