gallery_server_client_s190js/tools/runFaceClusteringScheduler.js

// tools/runFaceClusteringScheduler.js
const db = require('../db/knex');
const { spawn } = require('child_process');

let running = false;

async function getUsersNeedingClustering() {

  const result = await db.raw(`
    SELECT
        p.user
    FROM faces f
    JOIN photos p
        ON p.id = f.photo_id
    LEFT JOIN (
        SELECT
            user,
            MAX(completed_at) AS last_cluster
        FROM face_cluster_runs
        GROUP BY user
    ) r
        ON r.user = p.user
    GROUP BY p.user
    HAVING
        MAX(f.created_at)
        >
        COALESCE(
            MAX(r.last_cluster),
            '1970-01-01'
        )
  `);

  return Array.isArray(result)
    ? result
    : result?.rows || [];
}

function clusterUser(user) {
  return new Promise((resolve, reject) => {

    console.log(
      '[clusterScheduler] start:',
      user,
    );

    const child = spawn(
      process.execPath,
      [
        'tools/runFaceClustering.js',
        '--user',
        user,
        '--persist',
      ],
      {
        stdio: 'inherit',
      },
    );

    child.on('exit', (code) => {

      if (code === 0) {
        console.log(
          '[clusterScheduler] done:',
          user,
        );

        resolve();
        return;
      }

      reject(
        new Error(
          `Clustering exit code ${code}`,
        ),
      );
    });

    child.on('error', reject);
  });
}

async function clusterUserIfNeeded(user) {

  const result = await db.raw(
    `
    SELECT
      MAX(f.created_at) AS last_face,
      (
        SELECT MAX(completed_at)
        FROM face_cluster_runs
        WHERE user = ?
      ) AS last_cluster
    FROM faces f
    JOIN photos p
      ON p.id = f.photo_id
    WHERE p.user = ?
    `,
    [user, user],
  );

  const rows =
    Array.isArray(result)
      ? result
      : result?.rows || [];

  const row = rows[0];

  if (!row) {
    return false;
  }

  const lastFace =
    row.last_face;

  const lastCluster =
    row.last_cluster;

  if (!lastFace) {
    return false;
  }

  if (
    lastCluster
    && new Date(lastFace)
      <= new Date(lastCluster)
  ) {

    console.log(
      '[clusterScheduler] already up to date:',
      user,
    );

    return false;
  }

  console.log(
    '[clusterScheduler] clustering required:',
    user,
  );

  await clusterUser(user);

  return true;
}

async function runCycle() {

  if (running) {
    return;
  }

  running = true;

  try {

    const users =
      await getUsersNeedingClustering();

    if (!users.length) {

      console.log(
        '[clusterScheduler] no users to cluster',
      );

      return;
    }

    for (const row of users) {

      const user =
        row.user || row.USER;

      if (!user) {
        continue;
      }

      await clusterUser(user);
    }

  } finally {
    running = false;
  }
}

module.exports = {
  runCycle,
  clusterUserIfNeeded,
};