passing Id to 3rd party integration in beforeOperation hook & How to make Field "unique" but also null (Using Stripe)

[wchorski]

---

dg-publish: true

[!note] Payment Processing Modularity
The big goal here (and what is causing this birds nest of an issue) is to keep my payment processor as a plugin style architecture. Having Stripe as an optional add-on to this template is non-negotiable. The app should be able to work with our without Stripe.

Problem 1. Passing Product Id on Creation

[!info] Example
In this example I'll be referring to creating and updating a Stripe Product (in parody with the app's db), but this also extends to creating Customers, Coupons, Subscriptions, etc.

Using afterOperation hook

My initial thought was to create the Stripe Product so I can retrieve the id of the newly created object through the item property

the stripeProductCreate() is a wrapper function for Node Stripe API. Only thing to note is item.id is set to the metadata

{
	metadata: {
		productId: item.id
	}
}

afterOperation: {
	create: async ({ item, resolvedData, context }) => {
		const {
			id,
			name,
			excerpt,
			status,
			author,
			price,
			stripeProductId,
			stripePriceId,
			image,
		} = item

		try {
			const createdProduct = await stripeProductCreate({
				id,
				name: String(name),
				price: Number(price),
				excerpt,
				category: "product",
				status,
				authorId: author?.connect?.id || "no_author_id",
				type: "product",
				image,
				url: envs.FRONTEND_URL + `/products/${id}`,
				stripeProductId,
				stripePriceId,
				billing_interval: undefined,
			})

			if (createdProduct) {
				resolvedData.stripeProductId = createdProduct.id
				resolvedData.stripePriceId = String(createdProduct.default_price)
			}
		} catch (error: any) {
			console.log("!!! 💳 STRIPE:: ", {
				code: error.code,
				message: error.raw.message,
			})
		}
	},
	update: async ({ resolvedData, context, item }) => {
		await stripeProductUpdate({
			stripeProductId: item.stripeProductId,
			stripePriceId: item.stripePriceId,
			image: resolvedData.image as string | undefined,
			price: resolvedData.price as number | undefined,
			name: resolvedData.name as string | undefined,
			status: resolvedData.status as string | undefined,
			// category: resolvedData.category,
			category: "product",
			excerpt: resolvedData.excerpt as string | undefined,
			authorId: resolvedData.author?.connect?.id,
			billing_interval: undefined,
		}).then(async (res) => {
			if (!res) return

			if (res.default_price) {
				resolvedData.stripePriceId = String(res.default_price)
			}
		})

		resolvedData.dateModified = new Date().toISOString()
	},
},

[!error] resolvedData does not set new data
data set with resolvedData in the afterOperation does NOT save to the database.

Update Via Context

Ok... so instead I'll use a context.db.Product.updateOne right? Except that will create an infinite loop of the app

updating db -> hook triggers stripe api -> `.updateOne` re-triggers Product hook -> repeat

I could just run a check and see if all data is the same? like price, name, status, url. But that seems very cumbersome.

I could use a combination of beforeOperation and afterOperation, creating the product and immediately updating only the metadata. But I don't like this idea of multiple round trips to solve a seemingly trivial problem.

Using beforeOperation hook

After running into the afterOperation brick wall, It seemed more logical to create a Stripe Product in the beforeOperation

The code is very similar with one key exception

beforeOperation: {
	create: async ({ item, resolvedData, context }) => {
		const {
			id,
			name,
			excerpt,
			status,
			author,
			price,
			stripeProductId,
			stripePriceId,
			image,
		} = resolvedData

		try {
			const createdProduct = await stripeProductCreate({
				name: String(name),
				price: Number(price),
				excerpt,
				category: "product",
				status,
				authorId: author?.connect?.id || "no_author_id",
				type: "product",
				image,
				url: envs.FRONTEND_URL + `/products/${id}`,
				stripeProductId,
				stripePriceId,
				billing_interval: undefined,
			})

			if (createdProduct) {
				resolvedData.stripeProductId = createdProduct.id
				resolvedData.stripePriceId = String(createdProduct.default_price)
			}
		} catch (error: any) {
			console.log("!!! 💳 STRIPE:: ", {
				code: error.code,
				message: error.raw.message,
			})
		}
	},
	update: async ({ resolvedData, context, item }) => {
		await stripeProductUpdate({
			stripeProductId: item.stripeProductId,
			stripePriceId: item.stripePriceId,
			image: resolvedData.image as string | undefined,
			price: resolvedData.price as number | undefined,
			name: resolvedData.name as string | undefined,
			status: resolvedData.status as string | undefined,
			// category: resolvedData.category,
			category: "product",
			excerpt: resolvedData.excerpt as string | undefined,
			authorId: resolvedData.author?.connect?.id,
			billing_interval: undefined,
		}).then(async (res) => {
			if (!res) return

			if (res.default_price) {
				resolvedData.stripePriceId = String(res.default_price)
			}
		})

		resolvedData.dateModified = new Date().toISOString()
	},
},

For linking the stripeProductId via resolvedData this works great... except now I can't pass the Product's Id to metadata because the object hasn't been created yet. I know it doesn't make sense, but if there was a way to "pass Id to 3rd party integration in beforeOperation hook" that would theoretically solve all of this.

Using Stripe Ids to search

No matter, I'll use Stripe's provided Id's to identify my database items. This leads right into my second problem "Finding Product during Webhook"

Custom Unique or Blank Validation

Brief pitstop here. I'll need custom validation to allow these Stripe Id fields to be unique OR be blank (null) to prevent any overlap between data.

isIndexed 'Unique'

How about I set stripeProductId field to isIndexed: 'Unique'? That way I can where: ProductWhereUniqueInput! and use { connect: stripeProductId } or .findOne() easily during search or creation of data.

Except the stripeProductId is possibly a blank field (if the dev decides to forgo using Stripe). I realize I can make this app depend on Stripe and always set the stripeProductId, but I will re-emphasize the introductory note above

[!note] Payment Processing Modularity
The big goal here (and what is causing this birds nest of an issue) is to keep my payment processor as a plugin style architecture. Having Stripe as an optional add-on to this template is non-negotiable. The app should be able to work with our without Stripe.

Instead of the built in isIndexed: 'unique' I'll create a custom validate hook inside the field

Product.ts schema fields

...
stripeProductId: text({
	isIndexed: true,
	hooks: {
		validate: {
			create: async ({ resolvedData, context, addValidationError }) => {
				//? custom `unique` validation that also allows the field to be empty is not using stripe at all
				if (!resolvedData.stripeProductId) return
				const subPlans = await context.sudo().db.Product.findMany({
					where: {
						stripeProductId: {
							equals: resolvedData.stripeProductId,
						},
					},
				})
				if (subPlans.length > 0)
					addValidationError(
						"!!! Product can not share same stripeProductId with others"
					)
			},
			update: async ({ resolvedData, context, addValidationError }) => {
				if (!resolvedData.stripeProductId) return
				const subPlans = await context.sudo().db.Product.findMany({
					where: {
						stripeProductId: {
							equals: resolvedData.stripeProductId,
						},
					},
				})
				if (subPlans.length > 1)
					addValidationError(
						"!!! Product can not share same stripeProductId with others"
					)
			},
		},
	},
}),
stripePriceId: text({
	isIndexed: true,
	hooks: {
		validate: {
			create: async ({ resolvedData, context, addValidationError }) => {
				//? custom `unique` validation that also allows the field to be empty is not using stripe at all
				if (!resolvedData.stripePriceId) return
				const subPlans = await context.sudo().db.Product.findMany({
					where: {
						stripePriceId: {
							equals: resolvedData.stripePriceId,
						},
					},
				})
				if (subPlans.length > 0)
					addValidationError(
						"!!! Product can not share same stripePriceId with others"
					)
			},
			update: async ({ resolvedData, context, addValidationError }) => {
				if (!resolvedData.stripePriceId) return
				const subPlans = await context.sudo().db.Product.findMany({
					where: {
						stripePriceId: {
							equals: resolvedData.stripePriceId,
						},
					},
				})
				if (subPlans.length > 1)
					addValidationError(
						"!!! Product can not share same stripePriceId with others"
					)
			},
		},
	},
}),
...

I'm calling this a "custom unique validation" that also allows the field to be empty if dev is omitting the use of Stripe. This is a necessary step to prevent end users from the foot gun of linking the 2 or more products to the same Stripe Product

Problem 2. Finding Product during Webhook

With correct validation set, only using stripe's Ids as keys to search the app's database still comes with a few undesired side effects.

Data Search Workaround

Ok so I can't use where: ProductWhereUniqueInput!, but I can still use db.Product.findMany() to

async function findProductId(stripePriceId:string){
  const products = await keystoneContext.sudo().db.Product.findMany({
    where: {
      stripePriceId: {
        equals: stripePriceId,
      },
    },
  })  
  if(!products[0].id) throw new Error(`!!! products not found with stripePriceId:: ${stripePriceId}`, )
  //? schema field validation should not allow multiple items in this array
  return products[0].id
}

while the code works, the integrity still feels fragile.

Stripe Webhook

In this scenario, I've successfully completed a Stripe checkout session and Stripe returns back to my webhook with the response.

I'm able to use this helper function to find and connect a product to this theoretical Order Item data object for my app's database. Using await Promise.all()

{
	orderItem: {
		product: {
			connect: { id: await findProductId(stripe.product.id)  },
		}
	}
}

Conclusion

As to refrain from rambling on, I'm left with a deep "There must be a better way" feeling to all of this.

I apologies If there is a simple solution I'm not seeing. Perhaps I've fiddled for too long in this code base and am reaching out to fresh eyes and experienced devs who have a better idea of how to handles this 1 to 1 setup between stand alone NextJS app and Stripe API integration.

[gautamsi]

you want to use context.prisma for updating anything in afterOperation to avoid any loop.

Also in case you want to make something unique but still want to not provide this, use resolveInput to set a temporary value like stripe-temp-${nanoid(25)} and then overwrite this field with stripe product id.

[marekryb]

Hi,

afterOperation is not good fit for what you are doing. You should create custom mutation for creating product. Main reason being that if stripeProductCreate fails, you should probably actually rollback creating product (correct me if I am wrong). And good luck doing that in the hook.

Regarding stipeProductId uniqueness... Either I do not understand what you want to do or you missed the fact that a unique column in SQL table can have (multiple) null values and that's totally fine.