slate

Start here

gleam add slate

slate targets Erlang, where DETS ships with OTP. There is no database service or JavaScript-target dependency to add.

What it looks like

Open a table, insert, look up. Every value passes through the decoders you choose, so what comes off disk is typed — not dynamic data you have to trust.

import gleam/dynamic/decode
import gleam/result
import slate/set

pub fn main() {
  use users <- set.with_table(
    "data/users.dets",
    key_decoder: decode.string,
    value_decoder: decode.int,
  )
  use Nil <- result.try(
    set.insert(users, "alice", 42),
  )
  set.lookup(users, key: "alice")
}
// Ok(42), and the table is closed

Gleam's use value <- function(...) syntax passes the rest of the block as a callback. Here, with_table supplies users and closes the table when that callback returns. The decoders define the key and value types expected from disk; result.try stops early if the insert fails.

Values persist across node restarts. After an abrupt process termination, DETS may repair the file when it next opens. The Quick Start walks through this pattern step by step, and Safe Resource Management covers long-lived tables and repair behavior.

When to use DETS

Storage approaches at a glance

Approach	Complexity	Persistence	Query capability
JSON file	Low	Yes	None
DETS (slate)	Low	Yes	Key lookup, fold
SQLite/Postgres	High	Yes	Full SQL
Mnesia	High	Yes	Transactions, distribution

If a config file is too little and a database is too much, DETS is the middle layer you already have. It has real limits — 2 GB per table, disk I/O on every operation — and they are documented plainly in Limitations.

Pick a table type

Every table type gets the same typed API; they differ in what happens when a key repeats.

Set tables One value per key — inserting again overwrites. For caches, configuration, and general lookups.

Bag tables Multiple distinct values per key. For tags, categories, and one-to-many relationships.

Duplicate bag tables Duplicate key-value pairs allowed. For event logs, audit trails, and append-only patterns.

Ready to try it?

slate is 1.0, with a public API covered by semver guarantees. Continue with the Quick Start to read and write a table in about two minutes.