← Back to App

Open a MongoDB .bson Dump Without MongoDB Installed

mongodump writes one .bson file per collection — not JSON, and not something a text editor or Excel can open. The usual fix is standing up a MongoDB server just to run mongorestore, or writing a script with a BSON library. This is how to query the file directly, with neither.

What's Actually In a .bson File

It is not one document: it is many BSON documents written back to back, each starting with its own length in bytes. That is the whole format — no header, no index, no schema. A reader just has to walk the file, one length-prefixed document at a time.

Doing It in the Browser

OmniSelect FileSQL reads .bson files with the same bson library MongoDB's own drivers use, inside your browser tab. Nothing is uploaded.

1. Open the app.
2. Drop the file onto the File Select panel. A collection dumped as customers.bson becomes a table named by its first letter, C, shown in the Alias column.
3. Look at it.

   SELECT * FROM C

Four BSON-specific things happened before that table appeared:

• _id was an ObjectId, MongoDB's own 12-byte identifier. It becomes the plain 24-character hex text you'd see in mongosh — queryable and joinable like any other string.
• The nested address document is flattened into address_city and address_country, the same as a nested object in JSON or Avro.
• joined was a BSON Date (stored as milliseconds since the epoch). It arrives as a real date, not a number.
• tags, an array of plain strings, is joined with commas. Delta Partners' empty array becomes a blank cell, not a missing one.

Joining by ObjectId

An orders collection dumped alongside it, orders.bson (table O), references each customer by that same ObjectId. Once it's text, it joins exactly like any foreign key:

SELECT C.name, COUNT(*) AS orders, SUM(O.amount) AS total_amount
FROM O
JOIN C ON O.customer_id = C._id
GROUP BY C.name
ORDER BY total_amount DESC

What Else Happens to BSON's Own Types

BSON has several types JSON does not. Each becomes something you can filter and sort on rather than an opaque blob:

• Long and other 64-bit integers stay exact. A value too large for an ordinary number to hold precisely arrives as text with every digit intact, rather than rounded — the same rule Avro and Parquet longs follow here.
• Decimal128, Timestamp, MinKey and MaxKey arrive as their text representation.
• Binary data becomes hex bytes for anything short, or a note of how many bytes for anything long; a UUID stored in a Binary field (subtype 4) becomes the plain UUID text instead.
• DBRef (a reference naming its own collection) becomes a small object with the collection name and the referenced id, still queryable as a path.

What's Not Supported

• Only mongodump's per-collection .bson. A single BSON document embedded in another file's binary field is not a file of this kind.
• No write-back. Export the result as CSV, JSON, Excel or Parquet — not another .bson.
• Size. Up to 50 MB and 1,000,000 rows, the same limit as every other format here.

Reasonable Questions

Do I need mongoexport instead?

mongoexport writes JSON, which this app already reads — either works, but .bson keeps every BSON type (ObjectIds, dates, longs) exactly as MongoDB stored it, where mongoexport's JSON has already converted some of them.

Can I read more than one collection at once?

Yes. Load every .bson file mongodump wrote for a database, and join across collections the way customers.bson and orders.bson are joined above — see querying up to twenty-six files at once.

What about GridFS files?

GridFS stores a file's bytes split across many small chunk documents in fs.chunks, which is not something a table of rows is the right shape for. Metadata in fs.files opens and queries normally.

Does anything leave my machine?

No. The BSON reader runs in the tab, and after the page has loaded it makes no network requests — here is how to check.

Related Guides

• Reading MessagePack files with SQL
• Reading CBOR files with SQL
• One query across seven file formats
• Full How To Use guide