{"@attributes":{"version":"2.0"},"channel":{"title":"DEV Community","description":"The most recent home feed on DEV Community.","link":"https:\/\/dev.to","language":"en","item":[{"title":"I built a state management library for Rust called GLoC","pubDate":"Sun, 07 Jun 2026 19:11:58 +0000","link":"https:\/\/dev.to\/godwin_jk\/i-built-a-state-management-library-for-rust-called-gloc-3o1h","guid":"https:\/\/dev.to\/godwin_jk\/i-built-a-state-management-library-for-rust-called-gloc-3o1h","description":"

Hey,
\nI built a state management library for Rust called GLoC<\/strong>(Global Logic Component) \u2014 it's basically Bloc from Flutter but for Rust.<\/p>\n\n

Just published it on crates.io, would love some feedback.<\/p>\n\n

Crate<\/a>
\nRepo<\/a><\/p>\n\n

If you're not familiar with Bloc, I have a doc in the repo explaining it. Also wrote a few Medium articles walking through it with Dioxus \u2014 links in the repo.<\/p>\n\n

Medium tutorial<\/a><\/p>\n\n

Still early, any thoughts welcome!<\/p>\n\n","category":["rust","gloc","bloc"]},{"title":"SQLite in Production: Why We Chose It Over Postgres for a Multi-Tenant SaaS","pubDate":"Sun, 07 Jun 2026 19:11:31 +0000","link":"https:\/\/dev.to\/helperx\/sqlite-in-production-why-we-chose-it-over-postgres-for-a-multi-tenant-saas-44n8","guid":"https:\/\/dev.to\/helperx\/sqlite-in-production-why-we-chose-it-over-postgres-for-a-multi-tenant-saas-44n8","description":"

\"You're using SQLite in production? For a SaaS?\" \u2014 every developer who hears about our stack.<\/p>\n\n

Yes. And after a year of running it with hundreds of active accounts, I'd make the same choice again. Here's why \u2014 and the specific patterns that make it work for multi-tenant workloads.<\/p>\n\n

\n \n \n The architecture decision\n<\/h2>\n\n

HelperX manages multiple X accounts. Each account is an isolated \"slot\" with its own configuration, auth tokens, audit logs, and operational state. When we designed the data layer, we had two options:<\/p>\n\n

Option A: Shared Postgres<\/strong>
\nOne PostgreSQL database. All slots share tables. Isolation enforced by WHERE slot_id = ?<\/code> on every query.<\/p>\n\n

Option B: SQLite per slot<\/strong>
\nOne SQLite database file per slot. Complete physical isolation. No shared state.<\/p>\n\n

We chose Option B. Here's the decision matrix:<\/p>\n\n

\n\n\n\n\n\n\n\n\n\n\n\n
Factor<\/th>\nShared Postgres<\/th>\nSQLite per slot<\/th>\n<\/tr>\n<\/thead>\n
Data isolation<\/td>\nLogical (WHERE clause)<\/td>\nPhysical (separate files)<\/td>\n<\/tr>\n
Deployment complexity<\/td>\nRequires Postgres server<\/td>\nZero \u2014 embedded<\/td>\n<\/tr>\n
Backup granularity<\/td>\nFull DB or table-level<\/td>\nPer-slot file copy<\/td>\n<\/tr>\n
Slot deletion<\/td>\nDELETE + potential orphans<\/td>\nrm slot_abc.db<\/code><\/td>\n<\/tr>\n
Concurrent writes<\/td>\nExcellent<\/td>\nLimited (WAL mode helps)<\/td>\n<\/tr>\n
Read performance<\/td>\nNetwork round-trip<\/td>\nLocal disk, sub-ms<\/td>\n<\/tr>\n
Operational overhead<\/td>\nMonitoring, upgrades, HA<\/td>\nNone<\/td>\n<\/tr>\n
Cost<\/td>\n$20-100\/mo (managed DB)<\/td>\n$0<\/td>\n<\/tr>\n<\/tbody>\n<\/table><\/div>\n\n

For our workload \u2014 lots of reads, moderate writes, strong isolation requirements \u2014 SQLite won on every axis except concurrent writes. And our write pattern (one slot writes sequentially, never concurrently) means that limitation doesn't apply.<\/p>\n

\n \n \n The implementation\n<\/h2>\n

\n \n \n Database-per-slot pattern\n<\/h3>\n\n\n
\n
import<\/span> Database<\/span> from<\/span> '<\/span>better-sqlite3<\/span>'<\/span>;<\/span>\nimport<\/span> path<\/span> from<\/span> '<\/span>path<\/span>'<\/span>;<\/span>\nimport<\/span> fs<\/span> from<\/span> '<\/span>fs<\/span>'<\/span>;<\/span>\n\nconst<\/span> DB_DIR<\/span> =<\/span> path<\/span>.<\/span>join<\/span>(<\/span>process<\/span>.<\/span>cwd<\/span>(),<\/span> '<\/span>data<\/span>'<\/span>,<\/span> '<\/span>slots<\/span>'<\/span>);<\/span>\n\nfunction<\/span> getSlotDb<\/span>(<\/span>slotId<\/span>)<\/span> {<\/span>\n  const<\/span> dbPath<\/span> =<\/span> path<\/span>.<\/span>join<\/span>(<\/span>DB_DIR<\/span>,<\/span> `<\/span>${<\/span>slotId<\/span>}<\/span>.db`<\/span>);<\/span>\n  const<\/span> db<\/span> =<\/span> new<\/span> Database<\/span>(<\/span>dbPath<\/span>);<\/span>\n\n  db<\/span>.<\/span>pragma<\/span>(<\/span>'<\/span>journal_mode = WAL<\/span>'<\/span>);<\/span>\n  db<\/span>.<\/span>pragma<\/span>(<\/span>'<\/span>synchronous = NORMAL<\/span>'<\/span>);<\/span>\n  db<\/span>.<\/span>pragma<\/span>(<\/span>'<\/span>foreign_keys = ON<\/span>'<\/span>);<\/span>\n  db<\/span>.<\/span>pragma<\/span>(<\/span>'<\/span>busy_timeout = 5000<\/span>'<\/span>);<\/span>\n\n  return<\/span> db<\/span>;<\/span>\n}<\/span>\n<\/code><\/pre>\n\n<\/div>\n\n\n
Each slot gets its own .db<\/code> file. The pragmas are critical:<\/p>\n\n