Tabularis MongoDB Plugin

A Tabularis database driver plugin for MongoDB 6.0+, written in Rust on top of the official mongodb driver. It speaks Tabularis' JSON-RPC-over-stdio protocol.

Unlike SQL drivers, the query editor accepts native MongoDB shell calls (db.collection.find(...), aggregate, CRUD, ...) rather than SQL.

Requirements

• A reachable MongoDB 6.0+ server (local, self-hosted, or Atlas).
• A Rust toolchain to build from source. No system libraries needed — the driver is pure Rust with bundled TLS (rustls), so builds are clean on Windows, Linux and macOS.

Building

cargo test                # unit tests (no database needed)
cargo build --release     # -> target/release/tabularis-mongodb-plugin(.exe)

Installing

Copy the manifest and binary into a folder named mongodb inside the Tabularis plugins directory, then restart Tabularis (or enable it under Settings → Plugins → Installed):

• Windows: %APPDATA%\debba\tabularis\data\plugins\mongodb\
• Linux: ~/.local/share/tabularis/plugins/mongodb/
• macOS: ~/Library/Application Support/tabularis/plugins/mongodb/

mongodb/
├── manifest.json
└── tabularis-mongodb-plugin(.exe)

Connecting

Fill the connection form (host, port, user, password). MongoDB is multi-database: you connect once and browse every database on the server; collections appear as "tables". Use the plugin settings (gear icon) for anything beyond host/port:

Setting	Purpose
Connection String	Full mongodb:// / mongodb+srv:// URI, used verbatim (overrides host/port).
Use SRV	DNS seedlist scheme (mongodb+srv://), typical for Atlas.
Auth Source	authSource, e.g. admin.
Replica Set	replicaSet name.
Enable TLS	Connect over TLS.
Schema Inference Sample Size	Documents sampled per collection to infer columns (default 100).
Extra Connection Options	Verbatim query params, e.g. retryWrites=true&w=majority.

You can also paste a full connection string into the connection-string import field.

Querying (native MongoDB)

The query editor accepts shell-style calls. Argument bodies are JSON / Extended JSON; the shell helpers ObjectId(".."), ISODate(".."), NumberLong(..), NumberDecimal("..") and NumberInt(..) are understood.

db.users.find({ "age": { "$gt": 30 } }, { "name": 1, "_id": 0 })
db.users.find({ "_id": ObjectId("507f1f77bcf86cd799439011") })
db.sales.aggregate([{ "$match": { "region": "MX" } }, { "$group": { "_id": "$city", "n": { "$sum": 1 } } }])
db.users.countDocuments({ "active": true })
db.users.distinct("country", {})
db.users.insertOne({ "name": "Ada", "active": true })
db.users.updateMany({ "active": false }, { "$set": { "archived": true } })
db.users.deleteOne({ "_id": ObjectId("507f1f77bcf86cd799439011") })
db.createCollection("audit_log")
db.runCommand({ "dbStats": 1 })

find and aggregate are paginated by the grid automatically. Clicking a collection in the explorer runs the equivalent of db.coll.find({}).

Feature coverage

Area	Status
Connect / ping / test_connection	✅
List databases, collections (as tables)	✅
Column inference by sampling (nested fields → JSON)	✅
Indexes (list, create, drop)	✅
Views (list + definition)	✅
Native queries: find / aggregate / count / distinct	✅
Document CRUD: insert / update / delete / replace	✅
Inline grid editing (by _id, ObjectId-aware)	✅
Create / drop collection	✅ (via db.createCollection(...) / db.coll.drop())

Known limitations / mapping notes

• Schemaless ⇒ no column management. manage_tables is off; there are no fixed columns to ALTER. Create/drop collections via native commands.
• No foreign keys or stored routines — those metadata calls return empty.
• Column inference is sampled, so rarely-used fields may be missed; raise the sample size if needed. Conflicting scalar types across a field report as MIXED.
• Decimal128/dates/ObjectId render as strings in the grid to stay readable.

License

Apache-2.0.