Comparing some Cloud Data options

For a recent personal project, I tried out a few of the newer crop of cloud databases. I implemented (or tried to implement) the same app across each of the following 6 platforms. I’m actually looking for a platform to fit all my future data needs, so this was a fun opportunity to play with them all.

• Firebase
• Supabase
• PlanetScale
• Xata
• Fauna
• Turso

Also, given that these each leverage a different database under the hood (MYSQL, Postgres, SQLite), it seemed like a good idea to minimize friction when moving between them by leveraging an ORM. I experimented with

• Prisma
• Drizzle

I had some very unique data requirements, so my thoughts and feedback will be unique to the issues I was trying to solve. Very briefly, the items I’m looking for and evaluating, were

• a free tier for mucking about (with approx 2GB of storage)
• a FullText search solution (or pattern, way to implement)
• can store text documents upwards of 1MB

Lastly, I’m typing this up mostly as a memory aid to myself. There are probably inaccuracies, and things may have changed in the interim.

PLANETSCALE

• FULLTEXT search works, but not (really) in conjuction w/ other queries
• good free tier
• handles all manner of data, fields
• a little clunky to work with
  • pscale CLI tool is pretty helpful, though
• supports a “branch” architecture, if that’s your thing
• any/all mysql supported, but you’ll need to roll your own sql queries

XATA

• built-in FULLTEXT search (via Elastic search)
• very generous free tier
• cannot handle large text fields (> 64K)
  • update: now it’s at 200K …?
• “branch” architecture
• tooling is quite nice
  • xata CLI is nice
  • nice table visualizer
  • schema editing is nice, easy
• only rudimentary queries, ie. no joins, many-to-many search

TURSO

• FULLTEXT search TBD
• free tier is great
• SQLite, can store whatevers, ie. large text fields, etc
• tooling is limited, very much in its infancy
  • turso CLI is okay though
• it is noticeably faster than the others on this list

SUPABASE

• FULLTEXT search works, but hard to sort by “not-relevancy” (ie createdAt)
• very limited free tier (500MB)
• handles all manner of data, fields
• easy to work with tooling
  • nice table visualizer
• nice api, easy to query
  • web api, edge functions built-in

FAUNA

• FULLTEXT… not really supported. Need to author your own via FQL
• decent free tier
• handles all manner of data, fields
• FQL is a clusterfuck, incredibly complex, vague, poorly documented
• querying is a clusterfuck
• overall was very hard to get going, seed data, and use

FIRESTORE

• FULLTEXT is very clunky – need to author your own or work w/ 3rd party integrations
• decent free tier
• handles all manner of data
• nosql, hard to manage data, very clunky tooling
• clunky querying, hard to manage
  • recommended pattern is a separate repo within your repo??
  • setting up a “schema” and rules for your data (in custom JSON) is brittle, hard to update, hard to manage

Okay, and for the ORMs:

DRIZZLE

• fast, it lives up to the hype
• concise, simple to use
• typesafe
  • generates typings for your data (and it works well!)
• uses the schema to inform data relationships
  • there appears to be redundancy between foreign keys and this “relations” data, but one is for the schema, the other is for typings and the query builder
• still not clear search vs query
  • update: query sits on search, and uses the generated typings for a more streamlined experience

PRISMA

• really wanted this to work, but
• SLOOOOOOWWWW. Had to roll virtually every query by hand, kinda obviating the need for this
• clunky, not an easy local dev experience
• honestly, not sure why this is even a thing. Once upon a time it would have been useful for defining / updating your schema, but most of the above solutions do that natively, kinda obviating this.