zid•6mo ago

Best practice for testing recursive, potentially costly mutation

I'm nervous as there seems to be no manual kill switch for a recursive mutation call in convex. I've setup basic logic to exit out, however still, I'd appreciate a second look along with some ideas for best practices regarding these costly fns.

const MAX_RECURSION_DEPTH = 20; 

export const deactivateThreads = internalMutation({
  args: {
    cursor: v.optional(v.union(v.string(), v.null())),
    userId: v.id("users"),
    depth: v.optional(v.number()),
  },
  handler: async ({ db, scheduler }, { cursor, userId, depth = 0 }) => {
    if (depth > MAX_RECURSION_DEPTH) {
      throw new Error("Maximum recursion depth exceeded");
    }
    try {
      const user = await db.get(userId);
      const savedCursor =
        cursor ||
        (user?.deactivation?.nextCursor ? user.deactivation.nextCursor : null);

      const batchSize = 2;

      const { page, continueCursor, isDone, pageStatus } = await db
        .query("threads")
        .withIndex("by_userId", (q) => q.eq("userId", userId))
        .paginate({
          numItems: batchSize,
          cursor: savedCursor,
        });

      for (let i = 0; i < page.length; i++) {
        const currentThread = page[i];
        await db.delete(currentThread._id);
      }

      if (isDone) {
        // await resetCursor(db);
        await db.patch(userId, {
          deactivation: {
            ...user.deactivation,
            nextCursor: null,
          },
        });
        return;
      }

      if (!isDone) {

        await db.patch(userId, {
          deactivation: {
            count: user?.deactivation?.count ? user.deactivation.count + 1 : 0,
            nextCursor: continueCursor,
          },
        });

        await scheduler.runAfter(0, internal.internalUser.deactivateThreads, {
          cursor: continueCursor,
          userId: userId,
          depth: depth + 1,
        });
      }
    } catch (error) {
      throw error;
    }
  },
});

const MAX_RECURSION_DEPTH = 20; 

export const deactivateThreads = internalMutation({
  args: {
    cursor: v.optional(v.union(v.string(), v.null())),
    userId: v.id("users"),
    depth: v.optional(v.number()),
  },
  handler: async ({ db, scheduler }, { cursor, userId, depth = 0 }) => {
    if (depth > MAX_RECURSION_DEPTH) {
      throw new Error("Maximum recursion depth exceeded");
    }
    try {
      const user = await db.get(userId);
      const savedCursor =
        cursor ||
        (user?.deactivation?.nextCursor ? user.deactivation.nextCursor : null);

      const batchSize = 2;

      const { page, continueCursor, isDone, pageStatus } = await db
        .query("threads")
        .withIndex("by_userId", (q) => q.eq("userId", userId))
        .paginate({
          numItems: batchSize,
          cursor: savedCursor,
        });

      for (let i = 0; i < page.length; i++) {
        const currentThread = page[i];
        await db.delete(currentThread._id);
      }

      if (isDone) {
        // await resetCursor(db);
        await db.patch(userId, {
          deactivation: {
            ...user.deactivation,
            nextCursor: null,
          },
        });
        return;
      }

      if (!isDone) {

        await db.patch(userId, {
          deactivation: {
            count: user?.deactivation?.count ? user.deactivation.count + 1 : 0,
            nextCursor: continueCursor,
          },
        });

        await scheduler.runAfter(0, internal.internalUser.deactivateThreads, {
          cursor: continueCursor,
          userId: userId,
          depth: depth + 1,
        });
      }
    } catch (error) {
      throw error;
    }
  },
});

2 Replies

lee•6mo ago

this code looks good to me. i like how the depth argument guards against infinite loops. if you do end up with an infinite loop, note you can always go to Convex dashboard -> Settings -> Pause Deployment, which will stop all scheduled jobs but still let you cancel them or update your code with npx convex dev.

zidOP•6mo ago

oh so there is a kill switch, nice! And thanks for reviewing the code, @lee . I have a follow up question somewhat related to this but i'll create a separate ticket.

Best practice for testing recursive, potentially costly mutation

Did you find this page helpful?