如何使用 GraphQL 突变进行更新（热巧克力）

Question

We recently introduced GraphQL to our project, I've never used it before, however I like it a lot. We decided to go with the HotChocolate library for .NET Core and Apollo for client side.

One thing I am not quite sure is about mutations, specifically peforming updates and partial updates with mutations. I read somewhere that the practice is and that I should stick with creating specific mutation for each update , for instance updateUsername(), updateAddress(), updateCity() all of them should have specific mutation.

Issue with that is that my codebase will grow enormously if I decide to go in that direction, as we are very much data driven, with a lot of tables and columns per table.

Another question is, how to handle nullable properties, I can create a mutation which accepts some input object, but I'll end up with my entity being overwritten and all nullable properties not provided on the calling end will be set to null. Is there a way to handle this update partially or I should go with specific update mutation for each property I want updated?

Answer 1

I think you understood the best practice around specific mutations wrong. It's less "have one mutation to update one field" and more "have specific mutations that encapsulate actions in your domain". A concrete example would be creating an "addItemToBasket" mutation, instead of having 3 mutations that update the individual tables related to your shopping basket, etc.

GraphQL is very much centered around front-end development, so your mutations should, in general, closely resemble actions a user can perform in your front-end. Eg your front-end has an "Add to basket" button and your backend has an "addItemToBasket" mutation that resembles the action of placing an item in the user's basket.

If you design your mutations with this in mind, most of the time you shouldn't be having an issue with partial updates, since the mutation knows exactly what's to do and you are not just letting the user of your schema update fields at will.

If for whatever reason you need to have these partial updates, you likely won't get around implementing the patching yourself, unless your data provider supports it. Meaning you will have to have an input object type with nullable properties and your mutation that decides which fields have been changed and changing them using your data provider.

That being said, there's also a proposal for patching types in the Hot Chocolate repository that should simplify the patching part: https://github.com/ChilliCream/hotchocolate/issues/1326

Answer 2

for instance updateUsername(), updateAddress(), updateCity() all of them should have specific mutation.

Issue with that is that my codebase will grow enormously if I decide to go in that direction, as we are very much data driven, with a lot of tables and columns per table.

Correct. That's practically impossible to follow that way for more or less big data-driven applications. Consider how we implement the patching in our API here . Also consider following the discussion about the patching feature in HotChocolate github thread. Hope, that helps!