Limitations

slate wraps Erlang's DETS. Use these limits to decide whether it fits your storage needs.

File size limit

DETS limits each table file to 2 GB. You cannot configure this limit. A write that would exceed it returns Error(FileSizeLimitExceeded(context)).

Tip

If you need more than 2 GB of storage, split data across tables or use a database such as SQLite or Postgres.

No ordered_set table type

Unlike Erlang's ETS (in-memory storage), DETS does not support ordered_set tables. Only set, bag, and duplicate_bag are available. Keys are stored in an unspecified order.

Disk-backed performance

DETS stores data on disk. Performance depends on disk I/O, buffering, and your workload. A successful insert does not mean that the data has been synchronously flushed to disk. Use sync to flush pending writes and close when you finish with a table. Handle errors from both operations.

For frequent reads with low latency, load data into ETS at startup and use DETS for persistence.

Note

The related library shelf provides ETS tables backed by DETS. Reads use memory. Persistence still depends on when writes reach disk.

Tables must be closed properly

VM or node termination and power loss can prevent a DETS table from closing properly. Pending writes may be lost, and the file may need repair on the next open. File persistence does not guarantee that all writes survive a crash.

Keep primary data files on persistent storage and maintain backups. Deleting or replacing a DETS file loses the data in that file.

An individual Erlang process exit is different. DETS tracks processes that open a table and normally closes it when the last user exits. with_table attempts to close the table after its callback returns or raises. The helper cannot run cleanup after a forced kill of the owning process. For longer-lived tables, call close yourself.

slate's default AutoRepair policy attempts repair when needed. ForceRepair repairs even a properly closed file. NoRepair returns NeedsRepair when repair is required. None of these policies guarantees full data recovery. Back up a closed file before attempting repair.

Erlang target only

DETS is a BEAM feature. slate supports only Gleam's Erlang target. It does not support the JavaScript target.

Bounded table-name pool

slate reuses a pool of 4096 DETS table names instead of creating an atom for each path. Up to 4096 distinct normalized paths can be open at once. Closing a table releases its slot for reuse. If all slots are in use, new opens for other paths return TableNamePoolExhausted.

No concurrent access from multiple OS processes

Open each DETS file from only one OS process at a time. Multiple Erlang processes within that BEAM node can share the table. Opening the same file from separate BEAM nodes or OS processes can corrupt it.

Validating DETS files

Use slate.is_dets_file to identify a DETS file before opening it:

import slate

let assert Ok(True) = slate.is_dets_file("data/cache.dets")
let assert Ok(False) = slate.is_dets_file("README.md")

Use this when scanning a directory for DETS files. A result of Ok(True) does not guarantee file integrity, successful decoding, or safe access to a user-provided path. Apply your application's path and access rules separately.

Comparison with alternatives

DETS compared with other BEAM storage options

Feature	DETS (slate)	ETS	SQLite	Mnesia
Persistence	Disk	Memory only	Disk	Memory or disk, depending on table configuration
Maximum size	2 GB per file	RAM	Depends on configuration and storage	Depends on table type, configuration, and storage
Query capability	Key lookup, fold	Key lookup, match specs	Full SQL	Match specs, QLC
Ordered keys	No	Yes (ordered_set)	Yes	Yes
Storage engine	Included in OTP	Included in OTP	Separate library	Included in OTP
Performance	Depends on disk I/O and workload	In-memory reads	Depends on workload	Depends on configuration and workload
Concurrent processes	Single node	Single node	Multiple	Distributed