Postgres: Prisma push/migrate with custom unique index fails

Bug description

When there is custom unique index (altered migration file generated by Prisma), Prisma cannot see it and generates the same index with default unique parameters.

Custom index:

CREATE UNIQUE INDEX "Mixture_aromaId_boosterId_baseId_nicotine_key" ON "Mixture"(hash_record_extended(("aromaId", "boosterId", "baseId", "nicotine"), 0));

Prisma wants/generates:

CREATE UNIQUE INDEX "Mixture_aromaId_boosterId_baseId_nicotine_key" ON "Mixture"(("aromaId", "boosterId", "baseId", "nicotine"));

model Mixture {
  id              String        @id @default(cuid())
  // total size of a bottle
  volume          Decimal       @db.Decimal(10, 2)
  // amount of liquid in a bottle
  content         Decimal       @db.Decimal(10, 2)
  // amount of available space to be filled
  available       Decimal       @db.Decimal(10, 2)
  // volume - content => amount of extra or missing liquid
  diff            Decimal       @db.Decimal(10, 2)
  // amount of nicotine (mg/ml)
  nicotine        Decimal       @db.Decimal(10, 2)
  nicotineToRound Int
  // vg/pg ratio
  vg              Decimal       @db.Decimal(10, 2)
  pg              Decimal       @db.Decimal(10, 2)
  // vg/pg in ml
  vgToMl          Decimal       @db.Decimal(10, 2)
  // rounded VG ratio (for example 70 instead of 76)
  vgToRound       Int
  pgToMl          Decimal       @db.Decimal(10, 2)
  // rounded PG ratio (for example 70 instead of 76)
  pgToRound       Int
  aromaId         String
  aroma           Aroma         @relation(fields: [aromaId], references: [id], onDelete: Cascade)
  boosterId       String?
  booster         Booster?      @relation(fields: [boosterId], references: [id], onDelete: Cascade)
  boosterCount    Int
  baseId          String?
  base            Base?         @relation(fields: [baseId], references: [id], onDelete: Cascade)
  baseMl          Decimal       @db.Decimal(10, 2)
  error           MixtureError?

  Liquid Liquid[]

  @@unique([aromaId, boosterId, baseId, nicotine])
}

Given error is not showstopper, but makes DX a much harder as it fails on push.

How to reproduce

• (on Postgres)
• create model with multiple unique columns
• alter generated .sql file slightly (like example provided)
• reset database to execute updated migrations (what to say, you know the best what to do 😃 )
• basically in any time execute prisma push; it will try to add already created unique index

Expected behavior

Prisma detects uniqe keys at least by name and do not try to add it again with it’s (prisma’s) defaults.

Prisma information

"@prisma/client": "^3.13.0",

Environment & setup

• OS: Docker (alpine linux)
• Database: Postgres
• Node.js version: v16.13.0

Prisma Version

prisma                  : 3.13.0
@prisma/client          : 3.13.0
Current platform        : windows
Query Engine (Node-API) : libquery-engine efdf9b1183dddfd4258cd181a72125755215ab7b (at node_modules\@prisma\engines\query_engine-windows.dll.node)
Migration Engine        : migration-engine-cli efdf9b1183dddfd4258cd181a72125755215ab7b (at node_modules\@prisma\engines\migration-engine-windows.exe)
Introspection Engine    : introspection-core efdf9b1183dddfd4258cd181a72125755215ab7b (at node_modules\@prisma\engines\introspection-engine-windows.exe)
Format Binary           : prisma-fmt efdf9b1183dddfd4258cd181a72125755215ab7b (at node_modules\@prisma\engines\prisma-fmt-windows.exe)
Default Engines Hash    : efdf9b1183dddfd4258cd181a72125755215ab7b
Studio                  : 0.459.0

[marek-hanzal]

I understand that, the problem is Prisma trying to create already existing index. I was expecting Prisma gets indexes by name (is that possible?) and does not create them if they exists.

For the hash: on new Postgres it’s possible to create unique index on fields including null values, thus this method (hash_record_extended). It’s quite edge case, but it would be nice to prevent creation of already existing indexes.

Source of the finding: https://stackoverflow.com/questions/69330676/unique-index-with-a-lots-of-columns

[janpio]

The feature “expression indexes” is tracked here: https://github.com/prisma/prisma/issues/2504 The bug is fundamentally caused by us not supporting that (yet). Upvotes on that issue would be helpful.