ian / rel Goto Github PK

Reference:

https://github.com/ulid/spec

Why? Mostly because it's lexicographically sortable, case-insensitive and URL-safe

Example:

https://gqty.dev/docs/development/logger

I suggest we have a endpoints folder where the user is able to add custom functions where GraphQL is not the answer. We might add the generated GraphQL client there for them to use it too.

Let's include an auto-generated client library to be regenerated anytime the schema changes.

Options:

• GenQL
• Zeus
• others?

Following the quickstart, go thru the CLI and set up the project.

➜  mkdir rel-playground
➜  cd rel-playground
➜  rel-playground npm init --y
Wrote to /Users/danielstarns/code/rel-playground/package.json:

{
  "name": "rel-playground",
  "version": "1.0.0",
  "description": "",
  "main": "index.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  },
  "keywords": [],
  "author": "",
  "license": "ISC"
}


➜  rel-playground npx rel-cmd@latest init

Need to install the following packages:
  rel-cmd@latest
Ok to proceed? (y) y

  ___     _
 | _ \___| |
 |   / -_) |
 |_|_\___|_| Installer

 Rel is the zero-config backend framework for Javascripters.

? What's the name of your project? api

✔ Rel packages installed

? Base directory for Rel project? .


Success! Created reljs at directory /Users/danielstarns/code/rel-playground/api
Next steps:

1. Run rel dev and visit https://localhost:4000
2. Edit ./schema.graphql to update your schema.
3. Read more documentation at https://rel.run/docs

The following files are generated:

➜  rel-playground ls
api          package.json
➜  rel-playground ls api
node_modules      package.json      schema.graphql
package-lock.json rel.config.json
➜  rel-playground

Then when I run rel dev I get an error stating:

zsh: command not found: rel

Add in support for graphql-shield and add directives / custom guards support.

All of our CYPHER is Neo4j compliant, so there's no reason we can't support Neo4j backends. This will likely cripple background jobs without a Redis server but that's easily to add.

Add a @fake directive to type definition so the user can prototype something fast. Need to investigate what library is good now that faker.js is gone

This is the server-side issue for #12

Currently the API supports only one field for sorting, should be a array and leveraging the enum we are generating now for aggregations.

This will be helpful to connect to remote endpoints, or even the custom endpoints the user creates on his side, through GraphQL.

Finish documentation - in progress.

Inspiration:

We already have streams being added data on mutation, we can intercept them with a exclusive consumer group to handle caching

At some point we want to invest in our CLI tooling.

TODO:

• rel init or npx create-rel-app whichever feels better
• rel dev

""" @model """
type Note {
  id: ID!
  """ @constraint(min: 10, max: 100) """
  title: String!
  description: String
  """ @constraint(default: 0) """
  likes: Int
}

This will be a central directive to add traditional DB checks like UNIQUE and DEFAULT, and also validate content if the user needs something outside graphql-scalars which we already support.
The custom validations will probably leverage fastest-validator and validator

Right now we are on GraphQL 15.8.

Right now the tests we create (https://github.com/rel-js/rel/tree/main/packages/runtime-redisgraph/tests) are validating only the RedisGraph runtime and the schema codegen. We need a test suite validating the whole stack.

The Graphback annotations don't make justice to the potential a graph database has. We should think about a @relation annotation suited to RedisGraph strengths.

Comparisons

• https://blitzjs.com/docs/background-processing-with-quirrel#reminding-a-user-of-an-upcoming-event
• https://repeater.dev/
• https://demo.quirrel.dev/

@ian can share his experience leveraging geo on RedisGraph

We support the rel directive for relationships but need to add add support for mutations:

• create: Create and relate entries
• connect: Connect additional existing entries by unique field
• update: Update the connected entries
• upsert: Create or update connected entries
• disconnect: Disconnect connected relations by unique field
• delete: Delete all connected entries
• set: Override all connected entries

@thelinuxlich found a cool SM library that would be really easy to integrate as a Rel extension:

https://github.com/StoneCypher/jssm

What if we expose this as a directive. Consider:

type Post {
  state: String! @stateMachine('draft -> published -> archived')
}

This would produce:

type Post {
  id: ID!
  state: String!
}

type Mutation {
  nextStateMachineForPost(id: ID!): Post
  prevStateMachineForPost(id: ID!): Post
  setStateMachineForPost(id: ID!): Post
}

Added in ece0cc2

Migrate auth code over, use a lot of the providers from next-auth.

Since it's obligatory, the user shouldn't need to care about it.

Once the graphback legacy is finally removed, we should add changesets to manage releases

Aggregations should support a optional distinct argument to remove duplicates.

Comparisons:

• https://neo4j.com/developer-blog/neo4j-graphql-aggregations/
• https://dgraph.io/docs/graphql/queries/aggregate/
• https://hasura.io/docs/latest/graphql/core/databases/postgres/queries/aggregation-queries.html
• https://www.graphqlize.org/docs/queries/query-relationships

Right now mutation events are added indefinitely to their respective streams, which means they stay there forever until a client gets the message and ACK's it.

A @lambda annotation would allow a transient and derived field available only on queries:

"""@model"""
type User {
  id: ID!
  name: String!
  lastName: String!
  """@lambda"""
  fullName: String!
}

Then Rel would scan at startup a lambda folder and use fullName.js to derive the field on queries:

// fullName.js

export default (entity) => entity.name + " " + entity.lastName

For example, Post has many Comments, therefore it could have a comments_count field automatically generated and maintained by Rel.

Examples:

https://gqty.dev/docs/client/upload
graphql-editor/graphql-zeus#150