Add C# (.NET) binding over the C ABI hub (#226)

The first language stecker on the C ABI hub: a .NET binding exposing all 514
indicators as idiomatic `IDisposable` classes, generated from `wickra.h`.

## What's here

- **`bindings/csharp/`** — the `Wickra` .NET 8 package. `[LibraryImport]`
  source-generated P/Invoke (`NativeMethods.g.cs`) plus idiomatic wrappers
  (`Indicators.g.cs`), both generated from the committed `bindings/c/include/wickra.h`.
  The binding owns no indicator maths — it only marshals types across the C ABI.
- **Marshalling, verified end-to-end against the native library.** Opaque handles
  cross as `nint` kept alive per call via a `SafeHandle`; `bool` as
  `[MarshalAs(U1)]` (Rust `bool` is one byte); a self-correcting
  `DllImportResolver` validates the loaded library actually exports the Wickra
  ABI. Tests cover one representative per FFI archetype (scalar, candle, pairwise,
  multi-output, bars, profile, values-profile, order-book / array-input) plus
  exact Sma reference values.
- **NuGet packaging** — `dotnet pack` produces `Wickra.<version>.nupkg`; the
  release pipeline stages prebuilt native libraries under `runtimes/<rid>/native/`
  for six target triples (win/linux/osx × x64/arm64).
- **`examples/csharp/`** — nine examples mirroring `examples/c/`: streaming,
  backtest, multi_timeframe, parallel_assets, three strategies, and
  fetch_btcusdt + live_binance.
- **CI** — a `csharp` job on the three OSes builds the C ABI, tests the binding,
  and runs the offline examples. **Release** — a gated `csharp-publish` job packs
  and pushes to NuGet (gated on `NUGET_API_KEY`, independent of the GitHub-release
  job so a C# hiccup never blocks the C/C++ asset release).
- **Docs consistency wave** — README, CONTRIBUTING, CHANGELOG, examples/README,
  the issue / PR templates, `sync-about.yml`, and `.gitattributes`.

The native Python / Node / WASM bindings and the C ABI are untouched; this is
additive. Publishing to NuGet stays gated behind the release tag and the secret.
This commit is contained in:
kingchenc
2026-06-09 14:32:05 +02:00
committed by GitHub
parent 4caaa1db97
commit 91f6f67257
59 changed files with 38713 additions and 87 deletions
+7
View File
@@ -0,0 +1,7 @@
# .NET build output
bin/
obj/
*.user
# Data fetched at runtime by fetch_btcusdt
**/data/
+19
View File
@@ -0,0 +1,19 @@
<Project>
<!-- Shared settings + references for every C# example. Each example project
only declares <OutputType>Exe</OutputType>. -->
<PropertyGroup>
<TargetFramework>net8.0</TargetFramework>
<LangVersion>latest</LangVersion>
<Nullable>enable</Nullable>
<ImplicitUsings>enable</ImplicitUsings>
<IsPackable>false</IsPackable>
</PropertyGroup>
<ItemGroup>
<ProjectReference Include="..\..\..\bindings\csharp\Wickra\Wickra.csproj" />
<Compile Include="..\_common\MarketData.cs" Link="_common\MarketData.cs" />
<Compile Include="..\_common\Backtest.cs" Link="_common\Backtest.cs" />
</ItemGroup>
</Project>
+29
View File
@@ -0,0 +1,29 @@
# Wickra examples — C# / .NET
Runnable .NET examples for the [Wickra .NET binding](../../bindings/csharp).
Each example is a small console project that references the `Wickra` project and
resolves the native library automatically (from `target/release` during local
development, or the NuGet `runtimes/` layout when packaged).
Build the native library first, then run any example:
```bash
cargo build -p wickra-c --release
dotnet run --project examples/csharp/streaming
```
| Example | What it does | Run |
| --- | --- | --- |
| `streaming` | Feed a synthetic price series through SMA / EMA / RSI / MACD tick by tick. | `dotnet run --project examples/csharp/streaming` |
| `backtest` | Compute a basket of indicators over an OHLCV series and print a summary. | `dotnet run --project examples/csharp/backtest -- <ohlcv.csv>` |
| `multi_timeframe` | Resample a 1-minute series into 5m / 15m and print an indicator per timeframe. | `dotnet run --project examples/csharp/multi_timeframe` |
| `parallel_assets` | SMA(20) batch over a panel of assets, serial vs `Parallel.For`, with speedup. | `dotnet run -c Release --project examples/csharp/parallel_assets -- 200 5000` |
| `strategy_rsi_mean_reversion` | RSI(14) mean-reversion with a PnL / Sharpe / max-DD summary. | `dotnet run -c Release --project examples/csharp/strategy_rsi_mean_reversion` |
| `strategy_macd_adx` | MACD crossover entries gated by ADX(14) > 20. | `dotnet run -c Release --project examples/csharp/strategy_macd_adx` |
| `strategy_bollinger_squeeze` | Bollinger-squeeze breakout with an ATR(14) trailing stop. | `dotnet run -c Release --project examples/csharp/strategy_bollinger_squeeze` |
| `fetch_btcusdt` | Download real BTCUSDT klines from the Binance REST API into a CSV. | `dotnet run --project examples/csharp/fetch_btcusdt` |
| `live_binance` | Stream live Binance klines through EMA(20) over a WebSocket. | `dotnet run --project examples/csharp/live_binance` |
`fetch_btcusdt` and `live_binance` require network access; the rest run offline
on deterministic synthetic data. Shared helpers (synthetic data, CSV loader,
equity summary) live in [`_common/`](_common).
+45
View File
@@ -0,0 +1,45 @@
namespace Wickra.Examples;
/// <summary>Summary statistics for a long-only equity curve.</summary>
public sealed record EquityResult(double TotalReturnPct, double Sharpe, double MaxDrawdownPct, int Trades, double FinalEquity);
/// <summary>
/// Minimal long-only backtest helper: turn a stream of per-bar fractional
/// returns into a PnL / Sharpe / max-drawdown summary. The strategy examples
/// produce the returns; this aggregates them.
/// </summary>
public static class Backtest
{
/// <param name="periodReturns">Per-bar fractional returns (0.01 == +1%).</param>
/// <param name="trades">Number of position entries.</param>
/// <param name="periodsPerYear">Annualisation factor for the Sharpe ratio.</param>
public static EquityResult Summarize(IReadOnlyList<double> periodReturns, int trades, double periodsPerYear = 252.0)
{
double equity = 1.0, peak = 1.0, maxDrawdown = 0.0;
foreach (var r in periodReturns)
{
equity *= 1.0 + r;
peak = Math.Max(peak, equity);
if (peak > 0)
{
maxDrawdown = Math.Max(maxDrawdown, (peak - equity) / peak);
}
}
var mean = periodReturns.Count > 0 ? periodReturns.Average() : 0.0;
var variance = periodReturns.Count > 1
? periodReturns.Sum(x => (x - mean) * (x - mean)) / (periodReturns.Count - 1)
: 0.0;
var stdDev = Math.Sqrt(variance);
var sharpe = stdDev > 1e-12 ? mean / stdDev * Math.Sqrt(periodsPerYear) : 0.0;
return new EquityResult((equity - 1.0) * 100.0, sharpe, maxDrawdown * 100.0, trades, equity);
}
/// <summary>Prints a one-line summary.</summary>
public static void Print(string name, EquityResult r)
{
Console.WriteLine(
$"{name,-26} return={r.TotalReturnPct,8:F2}% sharpe={r.Sharpe,6:F2} maxDD={r.MaxDrawdownPct,6:F2}% trades={r.Trades}");
}
}
+78
View File
@@ -0,0 +1,78 @@
namespace Wickra.Examples;
/// <summary>One OHLCV bar with a millisecond timestamp.</summary>
public readonly record struct Bar(double Open, double High, double Low, double Close, double Volume, long Timestamp);
/// <summary>
/// Deterministic synthetic market data plus a small OHLCV CSV loader, shared by
/// the offline examples so they run without network access.
/// </summary>
public static class MarketData
{
/// <summary>A reproducible price path (trend + two cycles), no randomness.</summary>
public static double[] SyntheticPrices(int count, double start = 100.0)
{
var prices = new double[count];
for (var i = 0; i < count; i++)
{
prices[i] = start + 12.0 * Math.Sin(i * 0.05) + 5.0 * Math.Sin(i * 0.013) + i * 0.01;
}
return prices;
}
/// <summary>A reproducible OHLCV series derived from <see cref="SyntheticPrices"/>.</summary>
public static Bar[] SyntheticCandles(int count, long startTimestamp = 0, long stepMs = 3_600_000)
{
var prices = SyntheticPrices(count + 1);
var bars = new Bar[count];
for (var i = 0; i < count; i++)
{
var open = prices[i];
var close = prices[i + 1];
var high = Math.Max(open, close) + 0.5 + Math.Abs(Math.Sin(i * 0.7));
var low = Math.Min(open, close) - 0.5 - Math.Abs(Math.Cos(i * 0.7));
var volume = 1_000.0 + 500.0 * (1.0 + Math.Sin(i * 0.1));
bars[i] = new Bar(open, high, low, close, volume, startTimestamp + i * stepMs);
}
return bars;
}
/// <summary>
/// Loads an OHLCV CSV. Accepts rows of <c>timestamp,open,high,low,close,volume</c>
/// or <c>open,high,low,close,volume</c>; a non-numeric first row is treated as a header.
/// </summary>
public static Bar[] LoadOhlcvCsv(string path)
{
var bars = new List<Bar>();
foreach (var rawLine in File.ReadLines(path))
{
var line = rawLine.Trim();
if (line.Length == 0)
{
continue;
}
var cols = line.Split(',');
if (!double.TryParse(cols[0], System.Globalization.CultureInfo.InvariantCulture, out _) &&
!long.TryParse(cols[0], out _))
{
continue; // header row
}
double F(int i) => double.Parse(cols[i], System.Globalization.CultureInfo.InvariantCulture);
if (cols.Length >= 6)
{
bars.Add(new Bar(F(1), F(2), F(3), F(4), F(5), long.Parse(cols[0])));
}
else
{
bars.Add(new Bar(F(0), F(1), F(2), F(3), F(4), bars.Count));
}
}
return bars.ToArray();
}
}
+33
View File
@@ -0,0 +1,33 @@
using Wickra;
using Wickra.Examples;
// Compute a basket of indicators over an OHLCV series and print a summary.
// Pass a CSV path (timestamp,open,high,low,close,volume) or run on synthetic data.
var source = args.Length > 0 ? args[0] : "synthetic";
Bar[] bars = args.Length > 0 ? MarketData.LoadOhlcvCsv(args[0]) : MarketData.SyntheticCandles(1000);
Console.WriteLine($"Backtest over {bars.Length} bars ({source}):");
using var sma = new Sma(20);
using var ema = new Ema(50);
using var rsi = new Rsi(14);
using var atr = new Atr(14);
double lastSma = 0, lastEma = 0, lastRsi = 0, lastAtr = 0;
var oversold = 0;
foreach (var b in bars)
{
lastSma = sma.Update(b.Close);
lastEma = ema.Update(b.Close);
lastRsi = rsi.Update(b.Close);
lastAtr = atr.Update(b.Open, b.High, b.Low, b.Close, b.Volume, b.Timestamp);
if (double.IsFinite(lastRsi) && lastRsi < 30.0)
{
oversold++;
}
}
Console.WriteLine($" SMA(20) last = {lastSma:F4}");
Console.WriteLine($" EMA(50) last = {lastEma:F4}");
Console.WriteLine($" RSI(14) last = {lastRsi:F4} ({oversold} oversold bars)");
Console.WriteLine($" ATR(14) last = {lastAtr:F4}");
+5
View File
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
+33
View File
@@ -0,0 +1,33 @@
using System.Globalization;
using System.Text.Json;
// Download real BTCUSDT hourly klines from the Binance REST API into a CSV that the
// other examples can consume. Requires network access (build-only in CI).
const string url = "https://api.binance.com/api/v3/klines?symbol=BTCUSDT&interval=1h&limit=500";
using var http = new HttpClient();
Console.WriteLine($"Fetching {url}");
var json = await http.GetStringAsync(url);
using var doc = JsonDocument.Parse(json);
var dir = Path.Combine(AppContext.BaseDirectory, "data");
Directory.CreateDirectory(dir);
var path = Path.Combine(dir, "btcusdt_1h.csv");
using var writer = new StreamWriter(path);
writer.WriteLine("timestamp,open,high,low,close,volume");
var count = 0;
foreach (var kline in doc.RootElement.EnumerateArray())
{
// Binance kline array: [openTime, open, high, low, close, volume, ...]
var ts = kline[0].GetInt64();
var o = kline[1].GetString();
var h = kline[2].GetString();
var l = kline[3].GetString();
var c = kline[4].GetString();
var v = kline[5].GetString();
writer.WriteLine(string.Create(CultureInfo.InvariantCulture, $"{ts},{o},{h},{l},{c},{v}"));
count++;
}
Console.WriteLine($"Wrote {count} klines to {path}");
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
+40
View File
@@ -0,0 +1,40 @@
using System.Globalization;
using System.Net.WebSockets;
using System.Text;
using System.Text.Json;
using Wickra;
// Stream live BTCUSDT 1-minute klines from Binance and feed each close through EMA(20).
// Requires network access (build-only in CI). Runs for up to 60 seconds.
var uri = new Uri("wss://stream.binance.com:9443/ws/btcusdt@kline_1m");
Console.WriteLine($"Connecting to {uri} (up to 60s)...");
using var ws = new ClientWebSocket();
using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(60));
using var ema = new Ema(20);
var buffer = new byte[8192];
try
{
await ws.ConnectAsync(uri, cts.Token);
while (ws.State == WebSocketState.Open && !cts.IsCancellationRequested)
{
var result = await ws.ReceiveAsync(buffer, cts.Token);
if (result.MessageType == WebSocketMessageType.Close)
{
break;
}
using var doc = JsonDocument.Parse(Encoding.UTF8.GetString(buffer, 0, result.Count));
if (doc.RootElement.TryGetProperty("k", out var k))
{
var close = double.Parse(k.GetProperty("c").GetString()!, CultureInfo.InvariantCulture);
var value = ema.Update(close);
Console.WriteLine($"close={close:F2} EMA(20)={value:F2}");
}
}
}
catch (OperationCanceledException)
{
Console.WriteLine("Done (time limit reached).");
}
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
@@ -0,0 +1,44 @@
using Wickra;
using Wickra.Examples;
// Resample a 1-minute series into higher timeframes and run an indicator per timeframe.
var oneMinute = MarketData.SyntheticCandles(1200, startTimestamp: 0, stepMs: 60_000);
Console.WriteLine("EMA(20) of close across timeframes (resampled from 1-minute bars):");
foreach (var factor in new[] { 1, 5, 15 })
{
var bars = Resample(oneMinute, factor);
using var ema = new Ema(20);
double last = 0;
foreach (var b in bars)
{
last = ema.Update(b.Close);
}
Console.WriteLine($" {factor,2}m: {bars.Length,5} bars EMA(20) last = {last:F4}");
}
static Bar[] Resample(Bar[] source, int factor)
{
if (factor <= 1)
{
return source;
}
var output = new List<Bar>();
for (var i = 0; i < source.Length; i += factor)
{
var end = Math.Min(i + factor, source.Length);
double high = double.MinValue, low = double.MaxValue, volume = 0;
for (var j = i; j < end; j++)
{
high = Math.Max(high, source[j].High);
low = Math.Min(low, source[j].Low);
volume += source[j].Volume;
}
output.Add(new Bar(source[i].Open, high, low, source[end - 1].Close, volume, source[i].Timestamp));
}
return output.ToArray();
}
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
@@ -0,0 +1,47 @@
using System.Diagnostics;
using Wickra;
using Wickra.Examples;
// Run SMA(20) batch over a panel of assets, serial vs Parallel.For, and report the speedup.
var assets = args.Length > 0 ? int.Parse(args[0]) : 500;
var bars = args.Length > 1 ? int.Parse(args[1]) : 20_000;
var panel = new double[assets][];
for (var a = 0; a < assets; a++)
{
panel[a] = MarketData.SyntheticPrices(bars, start: 50.0 + a * 0.1);
}
// Warm up the JIT and thread pool so the comparison is fair.
using (var warm = new Sma(20))
{
warm.Batch(panel[0]);
}
var sink = 0.0;
var sw = Stopwatch.StartNew();
for (var a = 0; a < assets; a++)
{
using var sma = new Sma(20);
var result = sma.Batch(panel[a]);
sink += result[^1];
}
sw.Stop();
var serialMs = sw.Elapsed.TotalMilliseconds;
var lasts = new double[assets];
sw.Restart();
Parallel.For(0, assets, a =>
{
using var sma = new Sma(20);
var result = sma.Batch(panel[a]);
lasts[a] = result[^1];
});
sw.Stop();
var parallelMs = sw.Elapsed.TotalMilliseconds;
Console.WriteLine($"{assets} assets x {bars} bars, SMA(20) batch:");
Console.WriteLine($" serial {serialMs,8:F1} ms");
Console.WriteLine($" parallel {parallelMs,8:F1} ms ({serialMs / Math.Max(parallelMs, 1e-9):F1}x speedup)");
GC.KeepAlive(sink);
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
@@ -0,0 +1,46 @@
using Wickra;
using Wickra.Examples;
// Breakout: when Bollinger bandwidth is tight (a "squeeze") and price closes above the
// upper band, go long with an ATR(14) trailing stop.
var bars = args.Length > 0 ? MarketData.LoadOhlcvCsv(args[0]) : MarketData.SyntheticCandles(2000);
using var bollinger = new BollingerBands(20, 2.0);
using var atr = new Atr(14);
var returns = new List<double>();
var trades = 0;
var inPosition = false;
var entry = 0.0;
var stop = 0.0;
foreach (var b in bars)
{
var bands = bollinger.Update(b.Close);
var atrValue = atr.Update(b.Open, b.High, b.Low, b.Close, b.Volume, b.Timestamp);
if (bands is not { } band || !double.IsFinite(atrValue))
{
continue;
}
var bandwidth = band.Middle != 0.0 ? (band.Upper - band.Lower) / band.Middle : double.MaxValue;
if (!inPosition && bandwidth < 0.06 && b.Close > band.Upper)
{
inPosition = true;
entry = b.Close;
stop = b.Close - 2.0 * atrValue;
trades++;
}
else if (inPosition)
{
stop = Math.Max(stop, b.Close - 2.0 * atrValue); // trail the stop up
if (b.Close < stop)
{
returns.Add((b.Close - entry) / entry);
inPosition = false;
}
}
}
Backtest.Print("Bollinger squeeze", Backtest.Summarize(returns, trades));
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
@@ -0,0 +1,42 @@
using Wickra;
using Wickra.Examples;
// Trend follower: enter long on a MACD histogram cross up, but only when ADX(14) > 20
// confirms a trend; exit when the histogram crosses back below zero.
var bars = args.Length > 0 ? MarketData.LoadOhlcvCsv(args[0]) : MarketData.SyntheticCandles(2000);
using var macd = new MacdIndicator(12, 26, 9);
using var adx = new Adx(14);
var returns = new List<double>();
var trades = 0;
var inPosition = false;
var entry = 0.0;
var prevHistogram = double.NaN;
foreach (var b in bars)
{
var m = macd.Update(b.Close);
var a = adx.Update(b.Open, b.High, b.Low, b.Close, b.Volume, b.Timestamp);
if (m is not { } macdValue || a is not { } adxValue)
{
continue;
}
var trending = adxValue.Adx > 20.0;
if (!inPosition && trending && double.IsFinite(prevHistogram) && prevHistogram <= 0.0 && macdValue.Histogram > 0.0)
{
inPosition = true;
entry = b.Close;
trades++;
}
else if (inPosition && macdValue.Histogram < 0.0)
{
returns.Add((b.Close - entry) / entry);
inPosition = false;
}
prevHistogram = macdValue.Histogram;
}
Backtest.Print("MACD + ADX trend", Backtest.Summarize(returns, trades));
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
@@ -0,0 +1,34 @@
using Wickra;
using Wickra.Examples;
// Mean reversion: go long when RSI(14) drops below 30, exit when it recovers above 50.
var bars = args.Length > 0 ? MarketData.LoadOhlcvCsv(args[0]) : MarketData.SyntheticCandles(2000);
using var rsi = new Rsi(14);
var returns = new List<double>();
var trades = 0;
var inPosition = false;
var entry = 0.0;
foreach (var b in bars)
{
var value = rsi.Update(b.Close);
if (!double.IsFinite(value))
{
continue;
}
if (!inPosition && value < 30.0)
{
inPosition = true;
entry = b.Close;
trades++;
}
else if (inPosition && value > 50.0)
{
returns.Add((b.Close - entry) / entry);
inPosition = false;
}
}
Backtest.Print("RSI mean-reversion", Backtest.Summarize(returns, trades));
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>
+29
View File
@@ -0,0 +1,29 @@
using Wickra;
using Wickra.Examples;
// Feed a synthetic price series through several indicators tick by tick (O(1) each).
var prices = MarketData.SyntheticPrices(500);
using var sma = new Sma(20);
using var ema = new Ema(20);
using var rsi = new Rsi(14);
using var macd = new MacdIndicator(12, 26, 9);
double lastSma = 0, lastEma = 0, lastRsi = 0;
MacdOutput? lastMacd = null;
foreach (var price in prices)
{
lastSma = sma.Update(price);
lastEma = ema.Update(price);
lastRsi = rsi.Update(price);
lastMacd = macd.Update(price);
}
Console.WriteLine($"Streamed {prices.Length} prices through SMA(20), EMA(20), RSI(14), MACD(12,26,9):");
Console.WriteLine($" SMA = {lastSma:F4}");
Console.WriteLine($" EMA = {lastEma:F4}");
Console.WriteLine($" RSI = {lastRsi:F4}");
if (lastMacd is { } m)
{
Console.WriteLine($" MACD = {m.Macd:F4} signal={m.Signal:F4} hist={m.Histogram:F4}");
}
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>