Data Fetching and store

Overview

In context of dApp, data fetching from RPC is typically managed by [wagmi which use Tanstack Query under the hood] (https://wagmi.sh/react/guides/tanstack-query). It is generally recommend to reuse that for other queries.

graphql-request is used with @tanstack query as lightweight fetching. Data are not normalized
- it is optional to combine with client from the graph with block tracking, multi-chain subgraphs handling etc.
@graphprotocol/client-cli is used for artifact generating
it’s challenging to manage schemas & related rate limit for multiple servers
https://docs.ens.domains/web/subgraph#the-graph
currently we consolidate graphql generation and types under packages/graphql instead of workspace level
opted for string documentMode for smaller bundle size and not using AST on the client (tanstack)
we will try to modularize and support cli to generate under @/graphql

For state management, we explore component-based options (and make React Context optional when possible).
In context of Web3, it is typical to listen to price feed or latest events. We find it performant and intuitive to use reactive pattern chaining compute graph from Observables and pass to components
We prefer nanostore over jotai over zustand
- Nanostore is tiny, tree-shakable, reactive and framework agnostic, recommended by astro, will be used for more basic use cases
- Jotai is more mature and is preferred to zustand although zustand is being used in scaffold. Firstly it is used by shadcn, and it doesn’t use a single store as in zustand