search

Graphcool / graphcool-framework

Apache License 2.0
1.77k stars 131 forks source link

Implement Transform Request step in Request Pipeline Functions #104

Open marktani opened 6 years ago

marktani commented 6 years ago

Issue by dkh215 Monday Jun 26, 2017 at 16:11 GMT Originally opened as https://github.com/graphcool/prisma/issues/278

I have several required fields in a model that I won’t be able to populate until I fetch data from a 3rd party API which I’d like to do at the Transform Request step. Right now, I have to make these fields optional in the Model but I'd like for them to be required so I can tap into Schema Validation.

marktani commented 6 years ago

Comment by kbrandwijk Monday Jun 26, 2017 at 19:22 GMT

Idea: could this be done my being able to mark fields on a Type as 'computed'? Then they could be left out of the generated input object / mutation parameters, so the client cannot provide values for them, and then set them in the TRANSFORM_ARGUMENT hook.

marktani commented 6 years ago

Comment by tomhut Tuesday Jun 27, 2017 at 15:48 GMT

Just adding my use case for consideration: I have two related Types, A Route and a Location. Routes consist of a name and an array of points. A location has a latitude and longitude.

When a Route is created, as part of the transform request step I would like a function to be called which would return it's input but with a location dictionary added. This would then result in a related Location node being automatically created before the next step is called.

marktani commented 6 years ago

Comment by sorenbs Saturday Jul 29, 2017 at 20:56 GMT

This could also be solved by https://github.com/graphcool/feature-requests/issues/280

marktani commented 6 years ago

Comment by danmkent Monday Aug 21, 2017 at 11:10 GMT

I think this feature could also solve a use case raised on the slack group (by user dtuite):

Imagine I have a mobile application like Medium which has an Article model. I’ve released v1 of the app which orders articles based on their createdAtdatetime. Now I want to release a new version, v2, which adds a publishedAt field to Article and uses that to order posts in the UI.

First I add the field to my schema and write a script to migrate my data so that all existing articles have their publishedAt value set to that of the createdAt value. Now I can release v2 of my app which displays articles based on publishedAt.

The problem is that users on v1 of the app are still creating new articles which don’t have a publishedAt value. I can’t force them to upgrade and I can’t migrate the data on it’s way into Graphcool with a serverless function because I can’t add new fields to an incoming payload. As far as I can see I’m basically stuck running my original data migration script over and over until all legacy users have migrated.

Is there a better way?

As I understand it, implementing this feature would enable us to add such missing fields to the incoming payload before it is processed by the rest of the pipeline. Is that right?