Mapping: stringcase conversion on introspection

[vmolero]

Problem

Manual mapping using @Map() and @@Map()

I've found a little bit off-putting the fact of having to map manually all columns and tables for a database of 110 tables, where the database is snakecase'd and the code is camelcase'd.

I think it would be beneficial if an option to automatically map table names and columns were in place

Suggested solution

I see a possible solution by just using a library like "knex-stringcase" (https://github.com/Kequc/knex-stringcase#readme). It would be used when generating the prisma schema via introspection command.

A new option in prisma.schema could be added such as (same way as knex-stringcase does)

generator client {
  provider = "prisma-client-js"
  stringcase = ['snakecase', 'camelcase']
}

So that, a prisma model name could be generated on introspection with the corresponding @Map() clauses

Example, after running $ npx prisma introspect I get the following model definition for my explicit n:m relationship

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

i get

model APARTMENT_IMAGE {
  APARTMENT_ID         Int
  IMAGE_ID             Int
  SORT_NO              Int                  @default(0)
  APARTMENT_PROPERTIES APARTMENT_PROPERTIES @relation(fields: [APARTMENT_ID], references: [APARTMENT_ID])
  IMAGE                IMAGE                @relation(fields: [IMAGE_ID], references: [ID])

  @@id([APARTMENT_ID, IMAGE_ID])
  @@index([IMAGE_ID], name: "FK_APARTMENT_IMAGE_IMAGE")
  @@unique([APARTMENT_ID, IMAGE_ID], name: "APARTMENT_ID_IMAGE_ID")
}

However, if I ran

generator client {
  provider = "prisma-client-js"
  stringcase = ['snakecase', 'camelcase']
}

The result, automatically, would be

model apartmentImage {
  apartmentId         Int		    					    @map(APARTMENT_ID)
  imageId             Int		    					    @map(IMAGE_ID)
  sortNo              Int                  @default(0)				    @map(SORT_NO)
  apartmentProperties @relation(fields: [APARTMENT_ID], references: [APARTMENT_ID]) @map(APARTMENT_PROPERTIES)
  image               @relation(fields: [IMAGE_ID], references: [ID])		    @map(IMAGE)

  @@id([APARTMENT_ID, IMAGE_ID])
  @@index([IMAGE_ID], name: "FK_APARTMENT_IMAGE_IMAGE")
  @@unique([APARTMENT_ID, IMAGE_ID], name: "APARTMENT_ID_IMAGE_ID")
 
  @@map(APARTMENT_IMAGE)
}

In a next iteration, the ability to distinguish between model name and model attributes would be ideal since you might want to have your model name capitalized but not the attribute.

I hope my feature suggestion is clear enough,

Thanks!

[nikolasburk]

This feature request seems to be included in this one: Introspection Configuration File (#1184 ) Should we close it and point people to #1184 in the future?

[nikolasburk]

Also, for everyone who is interested in this, check out this tool as a workaround that can be used to automatically transform snake_case to camelCase: https://github.com/IBM/prisma-schema-transformer