Skip to content

armgit / dub

Public Imported from GitHub
0

The modern link attribution platform. Loved by world-class engineering teams at Twilio, Zillow, Jobber, Vercel, Raycast, and more.

Update seed script to add filter values

Rafiebe4e2ccb630
diff --git a/apps/web/scripts/dev/seed-100k-partners.ts b/apps/web/scripts/dev/seed-100k-partners.ts
index 59d8c40cf4..e067a4abaa 100644
--- a/apps/web/scripts/dev/seed-100k-partners.ts
+++ b/apps/web/scripts/dev/seed-100k-partners.ts
@@ -1,9 +1,13 @@
/**
* Bulk-seeds partners for local development and partner-search benchmarking.
*
- * Writes User, Partner, PartnerUser, ProgramEnrollment, PartnerPlatform, and Link
- * rows — every field the search index reads — in atomic chunks of CHUNK_SIZE
- * partners, so 100K partners lands ~400K rows in seconds.
+ * Writes User, Partner, PartnerUser, ProgramEnrollment, PartnerPlatform, Link,
+ * and ProgramPartnerTag rows — every field the search index reads — in atomic
+ * chunks of CHUNK_SIZE partners, so 100K partners lands ~500K rows in seconds.
+ *
+ * Statuses, groups, and tags are spread across each partner so the filters have
+ * something to select on: seeding everything as approved and ungrouped makes a
+ * filtered search look correct while matching every row.
*
* Re-running with the same `--seed` collides. `seedFingerprint` is a deterministic
* hash of the program ID and `--seed`, so identical arguments regenerate identical
@@ -26,7 +30,7 @@ import { createId } from "@/lib/api/create-id";
import { hashPassword } from "@/lib/auth/password";
import { prisma } from "@/lib/prisma";
import { parsePositiveInteger } from "@/scripts/utils/parse-cli-number";
-import { PlatformType, Prisma } from "@prisma/client";
+import { PlatformType, Prisma, ProgramEnrollmentStatus } from "@prisma/client";
import { createHash } from "crypto";
import "dotenv-flow/config";
@@ -132,6 +136,25 @@ const DESCRIPTION_SPECIALTIES = [
"normalization", "partitioning", "replication",
];
+// Weighted so most partners are approved, which is what the table shows by
+// default, while leaving enough of every other status to filter on.
+// prettier-ignore
+const STATUS_WEIGHTS: ProgramEnrollmentStatus[] = [
+ ...Array<ProgramEnrollmentStatus>(84).fill("approved"),
+ ...Array<ProgramEnrollmentStatus>(6).fill("pending"),
+ ...Array<ProgramEnrollmentStatus>(3).fill("rejected"),
+ ...Array<ProgramEnrollmentStatus>(3).fill("banned"),
+ ...Array<ProgramEnrollmentStatus>(2).fill("archived"),
+ ...Array<ProgramEnrollmentStatus>(2).fill("deactivated"),
+];
+
+// prettier-ignore
+const TAG_NAMES = [
+ "VIP", "Newsletter", "YouTube", "Enterprise", "Agency", "Beta", "Inactive", "High Intent",
+];
+
+const SEEDED_GROUP_NAMES = ["Creators", "Agencies", "Enterprise", "Affiliates"];
+
const PLATFORM_TYPES: PlatformType[] = [
PlatformType.website,
PlatformType.youtube,
@@ -155,6 +178,7 @@ type PartnerChunk = {
enrollments: Prisma.ProgramEnrollmentCreateManyInput[];
platforms: Prisma.PartnerPlatformCreateManyInput[];
links: Prisma.LinkCreateManyInput[];
+ programPartnerTags: Prisma.ProgramPartnerTagCreateManyInput[];
};
type GeneratePartnerChunkOptions = {
@@ -164,9 +188,10 @@ type GeneratePartnerChunkOptions = {
passwordHash: string;
runStartedAt: Date;
programId: string;
- defaultGroupId: string | null;
programDomain: string | null;
workspaceId: string;
+ groupIds: (string | null)[];
+ tagIds: string[];
};
// FNV-1a with murmur3's fmix32 finalizer. Cheap enough to call several times
@@ -338,9 +363,10 @@ const generatePartnerChunk = ({
passwordHash,
runStartedAt,
programId,
- defaultGroupId,
programDomain,
workspaceId,
+ groupIds,
+ tagIds,
}: GeneratePartnerChunkOptions): PartnerChunk => {
const users: Prisma.UserCreateManyInput[] = [];
const partners: Prisma.PartnerCreateManyInput[] = [];
@@ -348,6 +374,7 @@ const generatePartnerChunk = ({
const enrollments: Prisma.ProgramEnrollmentCreateManyInput[] = [];
const platforms: Prisma.PartnerPlatformCreateManyInput[] = [];
const links: Prisma.LinkCreateManyInput[] = [];
+ const programPartnerTags: Prisma.ProgramPartnerTagCreateManyInput[] = [];
// Pick from a pool by hashing rather than `index % pool.length`. The benchmark
// samples every Nth partner, so a stride sharing a factor with the pool size
@@ -429,12 +456,24 @@ const generatePartnerChunk = ({
id: enrollmentId,
partnerId,
programId,
- groupId: defaultGroupId,
- status: "approved",
+ groupId: pick(groupIds, "groupId", i),
+ status: pick(STATUS_WEIGHTS, "status", i),
...generatePartnerMetrics(i),
createdAt,
});
+ // 0-3 tags per partner, so the array filter has partners to include and
+ // partners to exclude rather than everything matching.
+ const tagCount = hashToUint32(`${seedFingerprint}:tagCount:${i}`) % 4;
+ const firstTag = hashToUint32(`${seedFingerprint}:tag:${i}`);
+ for (let t = 0; t < tagCount; t++) {
+ programPartnerTags.push({
+ programId,
+ partnerId,
+ partnerTagId: tagIds[(firstTag + t) % tagIds.length],
+ });
+ }
+
// Offsetting a hashed start keeps each partner's platform types distinct,
// which PartnerPlatform requires per partner.
const platformCount =
@@ -474,7 +513,15 @@ const generatePartnerChunk = ({
});
}
- return { users, partners, partnerUsers, enrollments, platforms, links };
+ return {
+ users,
+ partners,
+ partnerUsers,
+ enrollments,
+ platforms,
+ links,
+ programPartnerTags,
+ };
};
const insertPartnerChunk = async ({
@@ -484,6 +531,7 @@ const insertPartnerChunk = async ({
enrollments,
platforms,
links,
+ programPartnerTags,
}: PartnerChunk) => {
// Keep every chunk atomic so a failed write cannot leave partial relations.
const [, partnerResult] = await prisma.$transaction([
@@ -493,6 +541,7 @@ const insertPartnerChunk = async ({
prisma.programEnrollment.createMany({ data: enrollments }),
prisma.partnerPlatform.createMany({ data: platforms }),
prisma.link.createMany({ data: links }),
+ prisma.programPartnerTag.createMany({ data: programPartnerTags }),
]);
return partnerResult.count;
@@ -572,6 +621,42 @@ async function main() {
.digest("hex")
.slice(0, 16);
const runStartedAt = new Date();
+
+ // Tags and groups are created once per run and reused, so filters have a
+ // small stable set of values to select on rather than one value per partner.
+ const tagIds = TAG_NAMES.map((_, index) => createId({ prefix: "ptag_" }));
+ await prisma.partnerTag.createMany({
+ data: TAG_NAMES.map((name, index) => ({
+ id: tagIds[index],
+ programId: program.id,
+ name: `${name} ${seed}`,
+ })),
+ });
+
+ const seededGroupIds = SEEDED_GROUP_NAMES.map(() =>
+ createId({ prefix: "grp_" }),
+ );
+ await prisma.partnerGroup.createMany({
+ data: SEEDED_GROUP_NAMES.map((name, index) => ({
+ id: seededGroupIds[index],
+ programId: program.id,
+ name: `${name} ${seed}`,
+ slug: `${name.toLowerCase()}-${seedFingerprint.slice(0, 6)}`,
+ })),
+ });
+
+ // The program's own default group, the seeded ones, and null — so "no group"
+ // is represented too, which is what a negated filter has to include.
+ const groupIds: (string | null)[] = [
+ program.defaultGroupId,
+ ...seededGroupIds,
+ null,
+ ];
+
+ console.log(
+ `Created ${tagIds.length} tags and ${seededGroupIds.length} groups for filtering\n`,
+ );
+
const totalChunks = Math.ceil(totalCount / CHUNK_SIZE);
const startTime = Date.now();
let insertedPartners = 0;
@@ -586,9 +671,10 @@ async function main() {
passwordHash,
runStartedAt,
programId: program.id,
- defaultGroupId: program.defaultGroupId,
programDomain: program.domain,
workspaceId: workspace.id,
+ groupIds,
+ tagIds,
});
if (chunk === 0) {