Convert JSONL to DuckDB
Upload a .json or .ndjson file and get back a DuckDB file. Types are preserved rather than flattened to text. Free, no registration, up to 10MB.
Click to upload or drag and drop
CSV, Excel, SQLite, SQL, DBF (max 10 MB)
How your data is typed in DuckDB
Generated from the converter's own DuckDB type mapping — this is what it actually emits, not an example.
| Your data | DuckDB type |
|---|---|
| Whole number | BIGINT |
| Decimal (money) | DECIMAL(10,2) |
| Floating point | DOUBLE |
| True / false | BOOLEAN |
| Date | DATE |
| Date and time | TIMESTAMP |
| Short text | VARCHAR |
| Long text | VARCHAR |
| Binary data | BLOB |
Working from a .jsonl file
JSON Lines and NDJSON are the same format under two names, and .jsonl is the spelling used by jq pipelines, Spark and machine-learning tooling. The content converts identically; only the file name causes trouble.
- →The .jsonl extension is not one the uploader recognises. Rename the file to .ndjson or .json - nothing inside it has to change.
- →Tools that write JSON Lines usually emit a directory of shards, part-00000.jsonl and so on. Only one convertible file per upload is accepted, and with no enclosing array to merge, cat part-*.jsonl > all.ndjson is the whole job.
- →Relaxed JSON is not JSON: trailing commas, single quotes and unquoted keys fail to parse, which is what a file written in a JSON5-ish mode will do. Running it through jq normalises it.
Decimals and text in the .duckdb file
DuckDB takes the source's declared types more nearly unchanged than any other target here, which makes the two places where it cannot the things worth reading: the width of a decimal, and the length of a string.
- →DECIMAL tops out at 38 digits, so a wider source column is clamped rather than refused: an Oracle NUMBER(50,10) is created here as DECIMAL(38,10), and the scale is never allowed to grow past the precision holding it.
- →A decimal that declared no scale gets six places for its fraction rather than none, inside that same rule: a source precision of 10 becomes DECIMAL(10,6), one of 5 becomes DECIMAL(5,5), and one that declared no precision either becomes DECIMAL(18,6).
- →Strings become VARCHAR with no length attached to them. DuckDB's VARCHAR is unbounded, so no column has to be sized in advance and no row can be turned away for being longer than the column it goes in.
Reading JSONL
- →JSON types are trusted: 1 and "1" stay different things.
- →Numbers with a decimal point are read as exact decimals rather than floats.
- →Both an array of objects and newline-delimited JSON work, and a wrapped array like {"data": [...]} is unwrapped automatically.
- →Nested objects are kept as JSON rather than flattened into invented column names.
Writing DuckDB
- →A real .duckdb file, not a script — open it straight away with the DuckDB CLI.
- →DuckDB has genuine DECIMAL and BOOLEAN types, so nothing is flattened on the way in.
What you get
A ZIP containing database.duckdb. If anything about your data needed a judgement call, a _warnings.txt explains it.
Questions
- Is converting JSONL to DuckDB free?
- Yes. Files up to 10MB convert free with no account. There is a limit of five conversions per day per IP address.
- Will my JSONL data keep its types in DuckDB?
- Yes — that is the point. Whole numbers, decimals, dates and booleans are mapped to real DuckDB types rather than everything becoming text. The mapping table above shows exactly what you get.
- Do you store my file?
- No. The file is written to a temporary directory, converted, and the directory is deleted as soon as your download has been sent.