/**
 * Human-readable diagnostic dumps for Metro / adb logcat.
 * Designed so a host can paste the terminal output into a support chat.
 */

import { Platform } from 'react-native';

import type { AccelerationDiagnostics } from '../types/acceleration';

const PREFIX = '[PoseTracker]';

/** Always-on console sink — shows up in `npx expo start` / Metro. */
export function defaultDiagnosticLogger(message: string): void {
  // Keep a single prefix so hosts can filter: `npx expo start | grep PoseTracker`
  console.log(PREFIX, message.startsWith('[posetracker]') ? message : message);
}

export function logPlatformBanner(): void {
  console.log(
    PREFIX,
    `platform=${Platform.OS} version=${String(Platform.Version)} ` +
      `constants=${JSON.stringify({
        isPad: (Platform as { isPad?: boolean }).isPad ?? null,
        isTV: Platform.isTV,
      })}`,
  );
}

/**
 * Multi-line dump of the warm-up / acceleration verdict. Call after
 * preload()/warmup() so Android CPU-fallback reasons are visible in Metro.
 */
export function logAccelerationReport(
  diag: AccelerationDiagnostics | null,
  extra?: Record<string, unknown>,
): void {
  const lines: string[] = [];
  lines.push('======== ACCELERATION REPORT ========');
  lines.push(`platform: ${Platform.OS} ${String(Platform.Version)}`);
  if (!diag) {
    lines.push('diagnostics: null (warm-up not finished)');
  } else {
    lines.push(`state: ${diag.state}`);
    lines.push(`runtime: ${diag.runtime ?? 'webview'}`);
    lines.push(`tfjsBackend: ${diag.tfjsBackend ?? 'n/a'}`);
    if (diag.delegate) {
      lines.push(`delegate: ${diag.delegate}`);
    }
    lines.push(
      `medianInferenceMs: ${diag.medianInferenceMs != null ? diag.medianInferenceMs.toFixed(1) : 'n/a'}`,
    );
    lines.push(
      `warmUpRunsMs: [${diag.inferenceTimesMs.map((ms) => Math.round(ms)).join(', ')}]`,
    );
    lines.push(`maxAcceptableInferenceMs: ${diag.maxAcceptableInferenceMs}`);
    lines.push(`contextLossCount: ${diag.contextLossCount}`);

    const caps = diag.capabilities;
    if (caps) {
      lines.push(`GL.renderer: ${caps.renderer ?? 'n/a'}`);
      lines.push(`GL.vendor: ${caps.vendor ?? 'n/a'}`);
      lines.push(`GL.version: ${caps.glVersion ?? 'n/a'}`);
      lines.push(`GL.maxTextureSize: ${caps.maxTextureSize ?? 'n/a'}`);
      lines.push(
        `GL.extensions(shimmed=${caps.extensionQueriesShimmed}): ` +
          `float=${caps.textureFloat} halfFloat=${caps.textureHalfFloat} ` +
          `colorBufferFloat=${caps.colorBufferFloat} colorBufferHalfFloat=${caps.colorBufferHalfFloat}`,
      );
    } else {
      lines.push('GL.capabilities: null (no GL context — typically means tfjs backend=cpu)');
    }

    lines.push(`tfjsFlags: ${JSON.stringify(diag.flags)}`);
    if (diag.reasons.length === 0) {
      lines.push('reasons: (none)');
    } else {
      lines.push('reasons:');
      for (const reason of diag.reasons) {
        lines.push(`  - ${reason}`);
      }
    }

    if (diag.state === 'gpu' && diag.runtime === 'vision') {
      lines.push(
        'INTERPRETATION: Apple Vision (VNDetectHumanBodyPoseRequest) — Neural Engine / GPU. Up to 19 joints.',
      );
    } else if (diag.runtime === 'webview' || diag.runtime === undefined) {
      lines.push(
        'INTERPRETATION: offline WebView MoveNet Lightning (ANGLE WebGL, bundled model) — ' +
          'same stack as the PoseTracker WebView product. Expected real-time on both platforms.',
      );
    } else if (diag.state === 'cpu-fallback') {
      lines.push(
        'INTERPRETATION: non-accelerated path (cpu-fallback). Expect low FPS.',
      );
    } else if (diag.state === 'gpu') {
      lines.push('INTERPRETATION: GPU path accepted by health check.');
    } else if (diag.state === 'unavailable') {
      lines.push('INTERPRETATION: no usable backend — model could not run.');
    }
  }

  if (extra) {
    lines.push(`extra: ${JSON.stringify(extra)}`);
  }
  lines.push('======== END ACCELERATION REPORT ========');

  for (const line of lines) {
    console.log(PREFIX, line);
  }
}

/** Throttled FPS / latency line for the camera loop (Metro-friendly). */
export function logFrameStats(stats: {
  fps: number;
  medianLatencyMs: number | null;
  backend: string;
  acceleration: string;
  frames: number;
  keypointsAbove03: number;
  meanScore: number;
}): void {
  console.log(
    PREFIX,
    `FRAMESTATS fps=${stats.fps} ` +
      `medianMs=${stats.medianLatencyMs != null ? Math.round(stats.medianLatencyMs) : 'n/a'} ` +
      `backend=${stats.backend} acceleration=${stats.acceleration} ` +
      `frames=${stats.frames} kp≥0.3=${stats.keypointsAbove03}/17 ` +
      `meanScore=${stats.meanScore.toFixed(2)}`,
  );
}