Files and Remote Sources

Dadbod Grip opens supported files as queryable grids without an active database connection. DuckDB reads the file directly. Enable write mode only for a supported local format when you intend to replace that file.

The duckdb client must be on your PATH before Dadbod Grip can open these sources.

GripOpen

:GripOpen is the entry point for file-based sources.

:GripOpen                      " picker: choose from recent files + connections
:GripOpen data.parquet         " open local Parquet file
:GripOpen logs.csv             " open local CSV
:GripOpen results.json         " open local JSON / NDJSON
:GripOpen book.xlsx            " open an Excel workbook

GripOpen does not save the source to .grip/connections.json. It opens a read session without polluting your connection list. Use :GripConnect if you want to persist it.

Local file types

Extension	Readable	Writable (`--write`)
`.parquet`	yes	yes
`.csv`	yes	yes
`.tsv`	yes	yes
`.json`	yes	yes
`.ndjson`	yes	yes
`.jsonl`	yes	yes
`.xlsx`	yes	no
`.orc`	yes	no
`.arrow`	yes	yes
`.ipc`	yes	yes

DuckDB infers the schema from the file, so you do not need to declare it first.

Remote HTTPS

Open any publicly accessible Parquet or CSV file directly from a URL:

:GripOpen https://raw.githubusercontent.com/some-repo/main/data.parquet
:GripOpen https://example.com/exports/report.csv

DuckDB reads the URL through httpfs. Response time depends on the file format, server, and query; remote URLs remain read-only.

S3 and object storage

:GripOpen s3://my-bucket/data/customers.parquet
:GripOpen s3://my-bucket/logs/2025/*.parquet   " glob: all matching files as one table

Authentication and any custom endpoint must be available to each DuckDB CLI invocation. Dadbod Grip does not create an S3 credential-chain secret or replay arbitrary setup SQL. Running CREATE TEMP SECRET in one query does not configure later queries: each query uses a new DuckDB process. Configure your DuckDB client for the storage provider before opening a private source.

MotherDuck

Attach MotherDuck to an active DuckDB connection:

:GripConnect duckdb::memory:
:GripAttach md:my_database cloud

Set MOTHERDUCK_TOKEN in your shell environment before attaching. Then query tables with the cloud prefix. :GripOpen md: does not provide a MotherDuck database picker.

Write mode

Write mode makes edits persist back to the source file on disk instead of running DML against a database. Supported formats are Parquet, CSV, TSV, JSON, NDJSON, JSONL, Arrow, and IPC.

:Grip data.parquet --write         " open in write mode from the start

Or toggle write mode on any open grid:

g!                                 " toggle write mode

A red WRITE badge appears in the grid winbar when write mode is active. Press a to apply staged edits back to the file. The original file is overwritten.

XLSX and ORC remain read-only. Dadbod Grip asks before overwriting a local file; remote URLs are always read-only.

Combining files in federation

Use :GripAttach for databases. Read a file with a DuckDB table function in the same query as the attached tables:

SELECT prod.customers.email, enrichment.segment
FROM prod.customers
JOIN read_parquet('/path/to/enrichment.parquet') AS enrichment
  ON enrichment.customer_id = prod.customers.id

The same pattern accepts an accessible HTTPS or S3 Parquet URL. A file path does not become an attached database with a .data table. See Federation for database attachment examples.