Automatically pascal case model names during introspection #1934

hcharley · 2020-03-25T20:06:01Z

Problem

Having to manually update model names to conform to Prisma's naming conventions using the introspection tool is 1) annoying to do for many snake cased table names, and 2) prevents an introspection-only schema approach to using prisma2.

Solution

Automatically change model names from this format:

model my_table {
  id   String @id @default(cuid())
  name String
}

To this:

model MyTable {
  id   String @id @default(cuid())
  name String
  @@map("my_table")
}

Alternatives

I could write my own script.

Additional context

Started in discussion at:

#1925

Just going to dump some regex here that might help write that script:

model ([a-z_]+) {

const result = `
model ${pascalCase(name)} {
  @@map(${name})
}
`

giautm · 2020-04-22T05:13:11Z

I also want fields must mapping to camelCase, because I introspect from a database with ~ 100 tables & 10~20 column per table. I don't want to map field by manual. I hope has a flag, when introspect used to switching between modes.

TLadd · 2020-04-22T22:01:48Z

I wanted to follow standard postgres naming practices (using snake case for column and field names so I don't have to quote everything when using raw sql), but still end up with a generated client that follows normal Typescript/prisma conventions (Pascal case for models, camelcase for field names, plurals for array relationships).

I wrote a script that handles all of my personal naming issues. Might be easier with access to the AST and there could be cases I'm missing since I'm starting on a new app and so my schema is fairly simple, but works for my use case and shouldn't be too hard to add additional cases as needed. Didn't bother trying to handle spacing. Saving the file in vscode with the prisma plugin fixes it. It handles:

change model name to pascal case in both the model declaration and when used as types for relation fields. Adds the appropriate @@Map annotation.
change field names from snake case to pascal case and updates references to those fields in relations and indexes. Add @Map annotation to end of field line.
Add and s to field names for one to many and many to many relationships.

import fs from "fs";
import path from "path";

const PRISMA_FILE_PATH = path.join(__dirname, "..", "prisma", "schema.prisma");

function snakeToCamel(str: string) {
  return str.replace(/([-_]\w)/g, (g) => g[1].toUpperCase());
}
function snakeToPascal(str: string) {
  let camelCase = snakeToCamel(str);
  return `${camelCase[0].toUpperCase()}${camelCase.substr(1)}`;
}

const PRISMA_PRIMITIVES = ["String", "Boolean", "Int", "Float", "DateTime"];

function isPrimitiveType(typeName: string) {
  return PRISMA_PRIMITIVES.includes(typeName);
}

function fixFieldsArrayString(fields: string) {
  return fields
    .split(", ")
    .map((field) => snakeToCamel(field))
    .join(", ");
}

async function fixPrismaFile() {
  const text = fs.readFileSync(path.join(PRISMA_FILE_PATH), "utf8");

  const textAsArray = text.split("\n");

  const fixedText = [];
  let currentModelName: string | null = null;
  let hasAddedModelMap = false;

  for (let line of textAsArray) {
    // Are we at the start of a model definition
    const modelMatch = line.match(/^model (\w+) {$/);
    if (modelMatch) {
      currentModelName = modelMatch[1];
      hasAddedModelMap = false;
      const pascalModelName = snakeToPascal(currentModelName);
      fixedText.push(`model ${pascalModelName} {`);
      continue;
    }

    // We don't need to change anything if we aren't in a model body
    if (!currentModelName) {
      fixedText.push(line);
      continue;
    }

    // Add the @@map to the table name for the model
    if (!hasAddedModelMap && (line.match(/\s+@@/) || line === "}")) {
      if (line === "}") {
        fixedText.push("");
      }
      fixedText.push(`  @@map("${currentModelName}")`);
      hasAddedModelMap = true;
    }

    // Renames field and applies a @map to the field name if it is snake case
    // Adds an s to the field name if the type is an array relation
    const fieldMatch = line.match(/\s\s(\w+)\s+(\w+)(\[\])?/);
    let fixedLine = line;
    if (fieldMatch) {
      const [, currentFieldName, currentFieldType, isArrayType] = fieldMatch;

      let fixedFieldName = snakeToCamel(currentFieldName);
      if (isArrayType && !fixedFieldName.endsWith("s")) {
        fixedFieldName = `${fixedFieldName}s`;
      }

      fixedLine = fixedLine.replace(currentFieldName, fixedFieldName);

      // Add map if we needed to convert the field name and the field is not a relational type
      // If it's relational, the field type will be a non-primitive, hence the isPrimitiveType check
      if (currentFieldName.includes("_") && isPrimitiveType(currentFieldType)) {
        fixedLine = `${fixedLine} @map("${currentFieldName}")`;
      }
    }

    // Capitalizes model names in field types
    const fieldTypeMatch = fixedLine.match(/\s\s\w+\s+(\w+)/);
    if (fieldTypeMatch) {
      const currentFieldType = fieldTypeMatch[1];
      const fieldTypeIndex = fieldTypeMatch[0].lastIndexOf(currentFieldType);
      const fixedFieldType = snakeToPascal(currentFieldType);
      const startOfLine = fixedLine.substr(0, fieldTypeIndex);
      const restOfLine = fixedLine.substr(
        fieldTypeIndex + currentFieldType.length
      );
      fixedLine = `${startOfLine}${fixedFieldType}${restOfLine}`;
    }

    // Changes `fields: [relation_id]` in @relation to camel case
    const relationFieldsMatch = fixedLine.match(/fields:\s\[([\w,\s]+)\]/);
    if (relationFieldsMatch) {
      const fields = relationFieldsMatch[1];
      fixedLine = fixedLine.replace(fields, fixFieldsArrayString(fields));
    }

    // Changes fields listed in @@index or @@unique to camel case
    const indexUniqueFieldsMatch = fixedLine.match(/@@\w+\(\[([\w,\s]+)\]/);
    if (indexUniqueFieldsMatch) {
      const fields = indexUniqueFieldsMatch[1];
      fixedLine = fixedLine.replace(fields, fixFieldsArrayString(fields));
    }

    fixedText.push(fixedLine);
  }

  fs.writeFileSync(PRISMA_FILE_PATH, fixedText.join("\n"));
}

fixPrismaFile();

Edit: Added check to avoid adding @Map to relation fields.

hcharley · 2020-04-22T22:33:44Z

Didn't bother trying to handle spacing. Saving the file in vscode with the prisma plugin fixes it

Just as an aside @TLadd, you can use this snippet for prettier:

import prettier from 'prettier';

let PRETTIER_OPTS = {};

if (!process.env._HAS_RESOLVED_PRETTIER) {
  const prettierConfigPath = prettier.resolveConfigFile.sync();
  if (prettierConfigPath) {
    const o = prettier.resolveConfig.sync(prettierConfigPath);
    if (o) {
      PRETTIER_OPTS = o;
    }
  }
  process.env._HAS_RESOLVED_PRETTIER = 'true';
}

export const formatWithPrettier = (output: string) => {
  return prettier.format(output, {
    parser: 'typescript', // or X parser
    ...PRETTIER_OPTS,
  });
};

ryands17 · 2020-04-23T12:57:02Z

For those using knex to migrate, I have updated the above script to remove the knex models that occur in the schema.prisma during introspection.

import * as fs from 'fs'
import * as path from 'path'

const PRISMA_FILE_PATH = path.join(__dirname, 'schema.prisma')

function snakeToCamel(str: string) {
  return str.replace(/([-_]\w)/g, (g) => g[1].toUpperCase())
}
function snakeToPascal(str: string) {
  let camelCase = snakeToCamel(str)
  return `${camelCase[0].toUpperCase()}${camelCase.substr(1)}`
}

const PRISMA_PRIMITIVES = ['String', 'Boolean', 'Int', 'Float', 'DateTime']
const KNEX_INTERNAL_MODELS = ['knex_migrations', 'knex_migrations_lock']

function isKnexInternalModel(typeName: string) {
  return KNEX_INTERNAL_MODELS.includes(typeName)
}

function isPrimitiveType(typeName: string) {
  return PRISMA_PRIMITIVES.includes(typeName)
}

function fixFieldsArrayString(fields: string) {
  return fields
    .split(', ')
    .map((field) => snakeToCamel(field))
    .join(', ')
}

async function fixPrismaFile() {
  const text = await fs.promises.readFile(PRISMA_FILE_PATH, 'utf8')

  const textAsArray = text.split('\n')

  const fixedText = []
  let currentModelName: string | null = null
  let hasAddedModelMap = false

  for (let line of textAsArray) {
    // Are we at the start of a model definition
    const modelMatch = line.match(/^model (\w+) {$/)
    if (modelMatch) {
      currentModelName = modelMatch[1]
      if (isKnexInternalModel(currentModelName)) {
        continue
      }
      hasAddedModelMap = false
      const pascalModelName = snakeToPascal(currentModelName)
      fixedText.push(`model ${pascalModelName} {`)
      continue
    }

    if (currentModelName && isKnexInternalModel(currentModelName)) {
      continue
    }

    // We don't need to change anything if we aren't in a model body
    if (!currentModelName) {
      fixedText.push(line)
      continue
    }

    // Add the @@map to the table name for the model
    if (!hasAddedModelMap && (line.match(/\s+@@/) || line === '}')) {
      if (line === '}') {
        fixedText.push('')
      }
      fixedText.push(`  @@map("${currentModelName}")`)
      hasAddedModelMap = true
    }

    // Renames field and applies a @map to the field name if it is snake case
    // Adds an s to the field name if the type is an array relation
    const fieldMatch = line.match(/\s\s(\w+)\s+(\w+)(\[\])?/)
    let fixedLine = line
    if (fieldMatch) {
      const [, currentFieldName, currentFieldType, isArrayType] = fieldMatch

      let fixedFieldName = snakeToCamel(currentFieldName)
      if (isArrayType && !fixedFieldName.endsWith('s')) {
        fixedFieldName = `${fixedFieldName}s`
      }

      fixedLine = fixedLine.replace(currentFieldName, fixedFieldName)

      // Add map if we needed to convert the field name and the field is not a relational type
      // If it's relational, the field type will be a non-primitive, hence the isPrimitiveType check
      if (currentFieldName.includes('_') && isPrimitiveType(currentFieldType)) {
        fixedLine = `${fixedLine} @map("${currentFieldName}")`
      }
    }

    // Capitalizes model names in field types
    const fieldTypeMatch = fixedLine.match(/\s\s\w+\s+(\w+)/)
    if (fieldTypeMatch) {
      const currentFieldType = fieldTypeMatch[1]
      const fieldTypeIndex = fieldTypeMatch[0].lastIndexOf(currentFieldType)
      const fixedFieldType = snakeToPascal(currentFieldType)
      const startOfLine = fixedLine.substr(0, fieldTypeIndex)
      const restOfLine = fixedLine.substr(
        fieldTypeIndex + currentFieldType.length
      )
      fixedLine = `${startOfLine}${fixedFieldType}${restOfLine}`
    }

    // Changes `fields: [relation_id]` in @relation to camel case
    const relationFieldsMatch = fixedLine.match(/fields:\s\[([\w,\s]+)\]/)
    if (relationFieldsMatch) {
      const fields = relationFieldsMatch[1]
      fixedLine = fixedLine.replace(fields, fixFieldsArrayString(fields))
    }

    // Changes fields listed in @@index or @@unique to camel case
    const indexUniqueFieldsMatch = fixedLine.match(/@@\w+\(\[([\w,\s]+)\]/)
    if (indexUniqueFieldsMatch) {
      const fields = indexUniqueFieldsMatch[1]
      fixedLine = fixedLine.replace(fields, fixFieldsArrayString(fields))
    }

    fixedText.push(fixedLine)
  }

  await fs.promises.writeFile(PRISMA_FILE_PATH, fixedText.join('\n'))
}

fixPrismaFile()

michaellzc · 2020-05-26T21:03:57Z

For those looking for fixing the naming convention automatically, I wrote a utility program that utilizes the Prisma schema AST, provided by the internal API getDMMF from @prisma/sdk, to automatically transform the naming of model and fields while keeping the mapping correct.

https://github.com/IBM/prisma-schema-transformer

saiar · 2020-09-05T11:57:59Z

@ExiaSR Thanks for the utility!

I just found this that might help with the same thing https://paljs.com/cli/schema/ for anyone who is interested. It supposedly also helps generate the TypeScript interface classes for the models.

jjangga0214 · 2021-10-30T17:04:52Z

Does resolving this take a lot of time? I think, with AST, we can easily deal with this. There are already many libraries for changing cases(e.g. snake_case to camelCase or PascalCase or vise versa)

jedashford · 2021-11-02T06:04:54Z

@ExiaSR Thanks for the utility!

I just found this that might help with the same thing https://paljs.com/cli/schema/ for anyone who is interested. It supposedly also helps generate the TypeScript interface classes for the models.

This is what worked for us, super simple and fast. Thanks!

allyusd · 2021-12-11T16:12:46Z

I'm looking https://paljs.com/cli/schema/, it's good, but it has some issues:

feat: changing case of model name of single word · Issue #241 · paljs/prisma-tools
feat: changing case of relation · Issue #240 · paljs/prisma-tools

Then I trying https://github.com/IBM/prisma-schema-transformer, it's great, but not yet supported prisma 3 now. (have committ log but not release)

So I fork a patch for it and write my solution on README.md
https://github.com/allyusd/prisma-schema-transformer

nikolasburk · 2022-02-09T09:01:32Z

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?

rclmenezes · 2022-11-06T13:49:02Z

I wrote my own solution because:

The other libraries didn't work with pnpm (they had yarn hard-coded in)
Enums still didn't work well

So here's my solution:
https://gist.github.com/rclmenezes/b3b382f31e9df24fca6fd709057ba73d

It works as a state machine, going over the file line by line.

It will convert enum names to PascalCase
It will convert model names to PascalCase and add an @@Map and singularize
It will convert field types to camelCase and add a @Map

Basically my workflow is:

Migrate my DB using raw SQL with Postgrator.
pnpm prisma db pull
pnpm ts-node createIdiomaticPrisma.ts

Anyway, hope this helps anyone else with the same problem.

ctsstc · 2023-02-01T00:02:08Z

It would be great to see options for naming conventions, like many are saying snake case for table names and for our team we use camel case on the model names.

It's kind of crazy to see all the lengths that everyone has gone to, to handle their cases 😬

ruheni · 2023-02-01T10:28:21Z

Hey 👋

I recently came across prisma-case-format by @iiian which seems to be a good workaround for ensuring consistent naming conventions in your Prisma schema. 🙂

shellscape · 2023-04-08T03:02:55Z

@nikolasburk @janpio this is another 3 year old issue asking for the same feature. the community demand is there, your competitors already support it. this is another missing DX feature that I'm sorry to say is not a good look for Prisma. I'd hope an open source project with corporate backing would have resources enough to tackle issues with this kind of age.

jedashford · 2023-04-08T14:13:11Z

@nikolasburk @janpio this is another 3 year old issue asking for the same feature. the community demand is there, your competitors already support it. this is another missing DX feature that I'm sorry to say is not a good look for Prisma. I'd hope an open source project with corporate backing would have resources enough to tackle issues with this kind of age.

It's open source, submit a PR, and help the community.

shellscape · 2023-04-08T14:29:19Z

what's open source? that's the first I've heard of it

kevinvdburgt · 2023-04-08T20:55:44Z

what's open source?

Prisma itself is open source. Feel free to create a PR and get some credits :-)

jedashford · 2023-04-08T21:37:18Z

@shellscape you stated yourself that its open source (link to description of open source):

I'd hope an open source project with corporate backing would have resources enough to tackle issues with this kind of age.

This means the community owns it. You, me, anyone willing to help. To demean other volunteers because your feature hasn't been worked on, when you are free and enabled to build/test/deliver it yourself, feels a bit disingenuous.

shellscape · 2023-04-08T22:18:09Z

Prisma itself is open source. Feel free to create a PR and get some credits :-)

Credits! Oh man. That sounds awesome. What do I get to do with those credits?

Wait so like, I can see the code and I can change the code?! That's incredible. I wonder if there's a project out there that millions of developers use every day that I could maintain. Maybe they'll let me if I have enough credits?

This means the community owns it. You, me, anyone willing to help. To demean other volunteers because your feature hasn't been worked on, when you are free and enabled to build/test/deliver it yourself, feels a bit disingenuous.

A thousand apologies, sir. The next time I have a criticism I'll just keep it to myself for sure. I surely wouldn't want to demean anyone with criticism.

janpio · 2023-04-12T23:28:43Z

I'd hope an open source project with corporate backing would have resources enough to tackle issues with this kind of age.

No, we don't, sorry. We are busy with other things that do not have easy manual or automated workarounds (see previous comments).

(Please keep it civil, everyone. I don't want to have to start hiding or deleting comments. Thanks.)

nilsm · 2023-08-05T13:09:38Z

As mentioned above https://github.com/iiian/prisma-case-format works great. This is the script I used in package.json to pull the schema for an introspection-only workflow.

"db:pull": "prisma db pull && prisma-case-format --file prisma/schema.prisma -p && prisma generate"

janpio added topic: introspection topic: introspection-configuration kind/feature A request for a new feature. labels Mar 25, 2020

schickling self-assigned this Mar 26, 2020

janpio added the team/product label Apr 22, 2020

TLadd mentioned this issue Apr 27, 2020

Introspect should normalize naming conventions #2283

Closed

janpio unassigned schickling May 12, 2020

matthewmueller added the topic: sdk label Aug 24, 2020

albertoperdomo added the team/schema Issue for team Schema. label Nov 25, 2020

janpio removed the team/product label Dec 11, 2020

albertoperdomo mentioned this issue Apr 14, 2021

Feedback wanted: Introspection #3659

Closed

janpio added the topic: conventions label May 29, 2021

allyusd mentioned this issue Dec 11, 2021

Feature: Config/Defaults #10593

Open

janpio changed the title ~~Feature request: automatically pascal case model names during introspection~~ Automatically pascal case model names during introspection Jan 15, 2022

janpio removed the topic: sdk label Apr 13, 2022

Jolg42 mentioned this issue Oct 26, 2022

Introspection: Automatic map columns to something sane #16000

Closed

rclmenezes mentioned this issue Nov 6, 2022

Support global config for name casing (table, fields) #8283

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Automatically pascal case model names during introspection #1934

Automatically pascal case model names during introspection #1934

hcharley commented Mar 25, 2020 •

edited by Jolg42

giautm commented Apr 22, 2020 •

edited

TLadd commented Apr 22, 2020 •

edited

hcharley commented Apr 22, 2020

ryands17 commented Apr 23, 2020

michaellzc commented May 26, 2020

saiar commented Sep 5, 2020

jjangga0214 commented Oct 30, 2021 •

edited by janpio

jedashford commented Nov 2, 2021

allyusd commented Dec 11, 2021

nikolasburk commented Feb 9, 2022

rclmenezes commented Nov 6, 2022 •

edited

ctsstc commented Feb 1, 2023

ruheni commented Feb 1, 2023

shellscape commented Apr 8, 2023

jedashford commented Apr 8, 2023

shellscape commented Apr 8, 2023

kevinvdburgt commented Apr 8, 2023

jedashford commented Apr 8, 2023

shellscape commented Apr 8, 2023

janpio commented Apr 12, 2023

nilsm commented Aug 5, 2023

Automatically pascal case model names during introspection #1934

Automatically pascal case model names during introspection #1934

Comments

hcharley commented Mar 25, 2020 • edited by Jolg42

Problem

Solution

Alternatives

Additional context

giautm commented Apr 22, 2020 • edited

TLadd commented Apr 22, 2020 • edited

hcharley commented Apr 22, 2020

ryands17 commented Apr 23, 2020

michaellzc commented May 26, 2020

saiar commented Sep 5, 2020

jjangga0214 commented Oct 30, 2021 • edited by janpio

jedashford commented Nov 2, 2021

allyusd commented Dec 11, 2021

nikolasburk commented Feb 9, 2022

rclmenezes commented Nov 6, 2022 • edited

ctsstc commented Feb 1, 2023

ruheni commented Feb 1, 2023

shellscape commented Apr 8, 2023

jedashford commented Apr 8, 2023

shellscape commented Apr 8, 2023

kevinvdburgt commented Apr 8, 2023

jedashford commented Apr 8, 2023

shellscape commented Apr 8, 2023

janpio commented Apr 12, 2023

nilsm commented Aug 5, 2023

hcharley commented Mar 25, 2020 •

edited by Jolg42

giautm commented Apr 22, 2020 •

edited

TLadd commented Apr 22, 2020 •

edited

jjangga0214 commented Oct 30, 2021 •

edited by janpio

rclmenezes commented Nov 6, 2022 •

edited