SED replace value in prisma schema

Assuming I have a file called schema.prisma that looks like this

generator client {
 provider = "prisma-client-js"
}

datasource db {
  provider = somevalue
  url      = env("DATABASE_URL")
}

model User {
  id        Int     @id @default(autoincrement())
  email     String      @unique
  name      String?
  todos     Todo[]
  createdAt DateTime    @default(now())
  updatedAt DateTime    @updatedAt
}

model Todo {
  id        Int     @id @default(autoincrement())
  createdAt DateTime    @default(now())
  updatedAt DateTime    @updatedAt
  title     String      @db.VarChar(255)
  completed Boolean     @default(false)
  user      User        @relation(fields: [userId], references: [id])
  userId    Int
}

And I want to use SED to replace the second provider instance

I have crafted the following regex that gets me the line I want

\bprovider(?!.*prisma).*/

But I just can't seem to get SED to work with it

For example

sed 's/\bprovider\(?!.*prisma\)/blue/' ./prisma/schema.prisma

Should produce

generator client {
 provider = "prisma-client-js"
}

datasource db {
  provider = blue
... rest of file

I know sed reads line by line so this approach to me seems like it should work.

What am I missing with this regex?

You could try matching a known pattern before the second occurrence: Using sed

$ sed -E '/^datasource db \{/{n;s/(= ).*/\1blue/}' input_file
generator client {
 provider = "prisma-client-js"
}

datasource db {
  provider = blue
  url      = env("DATABASE_URL")
}

model User {
  id        Int     @id @default(autoincrement())
  email     String      @unique
  name      String?
  todos     Todo[]
  createdAt DateTime    @default(now())
  updatedAt DateTime    @updatedAt
}

model Todo {
  id        Int     @id @default(autoincrement())
  createdAt DateTime    @default(now())
  updatedAt DateTime    @updatedAt
  title     String      @db.VarChar(255)
  completed Boolean     @default(false)
  user      User        @relation(fields: [userId], references: [id])
  userId    Int

Or you could null data the entire file with the -z flag matching the second occurrence of provider.

$ sed -Ez 's/(([^=]*provider = ){2})[^\n]*/\1blue/' input_file