Skip to main content

Discovery phase 1 - Electric SQL

Opencollect is an open source version of Epicollect.

  • In order to lower the barrier to entry for contribution as much as possible, a single programming language is used across the project: Typescript.

  • Since survey data is relational, we are going to use Postgres. Since we are trying to minimize time-to-market, we will use Supabase, which offers several advantages:

a) access to the massive Postgres ecosystem (eg geoquery support is particularly interesting to our use case). b) object a storage service out of the box. c) clear schema migration strategy. d) a user management system. e) a branching strategy for test environments.

  • Again with the goal of minimizing development effort, we are going to offload the data sync mechanism to Electric SQL, which is compatible with Supabase.

  • In order to make it easy to deploy, we are going to only focus on single-node architecture. A single node with vertical scaling should be enough for the vast majority of users to be able to collect potentially millions of datapoints.

  • In order to avoid Play Store and App Store hell and keep the solution lightweight, the frontend is build in Vue as a PWA.

Roadmap​

In order to minimize risk, we want to focus on the trickiest part of the system first, which is data sync with Electric SQL.

  1. Create a simple database schema.

  2. Create a simple Vue app that talks to Supabase.

Once these steps are developed, the core premise will be validated, and we will regroup and organise work better.

Takeaways​

  • Electric SQL is not a good fit. In fact, it does not handle writes. Also, it adds complexity in a project where simplicity is paramount.

  • Tanstack becomes then redudant. The fewer moving parts, the better.

  • Supabase confirmed.