SQLite to CSV Converter

Worked example: a two table database

A file called field-notes.db. It holds two tables with rows in them, one view, one table that was created and never used, and an index:

CREATE TABLE sessions    (id INTEGER, started_at TEXT, minutes REAL, notes TEXT);
CREATE TABLE attachments (id INTEGER, session_id INTEGER, thumb BLOB);
CREATE TABLE empty_log   (id INTEGER, line TEXT);
CREATE VIEW  recent_sessions AS SELECT id, started_at FROM sessions;
CREATE INDEX idx_sessions_started ON sessions(started_at);

Drop it in and the conversion has already run before you have decided anything. This is what comes back:

2 rows · from "attachments"

id,session_id,thumb
1,1,iVBORw0KGgo=
2,2,

attachments, not sessions. The list is alphabetical rather than in creation order, and the first name in it converts by default. Above the table sits a dropdown holding both names, attachments and sessions. It is the same control the Excel pages use for worksheets, so it carries the label Sheet; the names inside it are your tables. Choosing sessions reconverts immediately:

2 rows · from "sessions"

id,started_at,minutes,notes
1,2026-01-14T09:02:11Z,42.5,first run
2,2026-01-14T18:40:03Z,7,

Three things in those two outputs are worth naming. The view is not in the dropdown at all, and neither is empty_log, because a table with no rows is dropped before the list is built. The 8 byte blob in attachments came back as base64. And the REAL value stored as 7.0 prints as 7, since SQLite kept a number and CSV writes the number it was given.

The downloads are named field-notes-attachments.csv and field-notes-sessions.csv. That only happens when the database has more than one table to offer; a single table database keeps the file's own name, so vault.sqlite comes back as vault.csv.

Binary columns, and other things worth checking

• BLOBs become base64, up to a point. Under 1 MB the bytes are encoded into the cell. Over it, the cell reads <BLOB 1200000 bytes> or whatever the size happens to be, which keeps a table of stored photos from turning into a CSV nobody can open. Either way the row is still there with its other columns intact.
• Nulls and empty strings look the same. Both arrive as an empty cell, which is CSV behaving as CSV. In a database where a null answer and a blank answer mean different things, note the columns that matter before you send the file on.
• SQLite already converted some of your values. Column affinity happens at insert time, not on the way out. Text that looks like a number, put into a NUMERIC column, was stored as a number long before this page saw it, so 1e3 in the file reads as 1000. A TEXT column keeps 007 as 007.
• Only one file goes in. A database in write ahead logging mode keeps recent changes in a sidecar .db-wal file, and there is nowhere to attach one here. Check point or close the database properly before exporting if the last few writes matter.
• The name has to end in .sqlite, .sqlite3 or .db. Anything else is turned away unless the browser reports it as a generic binary file. Plenty of apps ship databases with no extension at all or their own invention, and renaming a copy is enough.
• Not every .db file is SQLite. A file that does not start with the SQLite header comes back saying it is not a database, which is also what a half finished download or an encrypted SQLCipher database looks like from the outside.
• The preview shows the first 20 rows. The count beside it is the real total, and the download holds all of them.

Related