Skip to content
IntoSQL home
Convert a file

Migration guides

SQLite to SQLite migration guide

Convert a SQLite database or dump into a SQLite script with the same tables, keys, indexes and data: how to dump, convert, load, and every rule the converter applies.

View as Markdown

IntoSQL reads a SQLite database file, or the SQL that SQLite's .dump writes, and turns it into one SQLite script with the same tables, columns, keys, indexes, foreign keys, defaults, counters and rows. SQLite keeps any value in any column, so the declared type is a hint: each column gets the SQLite type all of its values fit, and a column holding mixed values keeps their text. Every value SQLite can't hold as it is, and every object it doesn't carry, is listed in the report. The script loads on SQLite 3.31 and later.

Convert in the browser: SQLite to SQLite opens the converter set up for this pair.

The file or a dump

Drop the database file itself (.db, .sqlite, .sqlite3): it is read page by page, in UTF-8 or UTF-16, and text with NUL characters and every digit of a double arrive as stored. Close the app that writes to it first, so no change is left behind in its -wal file.

A dump converts too, as .sql or .sql.gz:

Terminal
sqlite3 shop.db .dump > shop.sql
  • DB Browser for SQLite writes the same SQL with File, Export, Database to SQL file, and so does Python's iterdump().
  • .dump --nosys converts too. A --data-only dump has no CREATE TABLE, so its columns are named column_1, column_2 and so on, and the types come from the values.
  • Every test file and dump comes from sqlite3 3.31 to 3.54, with STRICT and WITHOUT ROWID tables, and is compared with its source cell by cell.

Convert

In the browser, open SQLite to SQLite or the converter on the home page, drop the file and choose SQLite. It reads the file from your disk in pieces, twice, and writes the script as it goes, so the file is never uploaded and its size isn't limited by memory. The desktop app does the same with the file on disk.

With the command line:

Terminal
intosql inspect shop.db
intosql convert shop.db --to sqlite --output shop.sqlite.sql --report shop.report.json

inspect lists the tables, their rows and what won't be converted, without writing anything. convert writes the script and, with --report, the report as JSON.

Through the API, a database file or a dump converts like any other file: open a conversion with its name and "output": {"dialect": "sqlite"}.

Load into SQLite

Terminal
sqlite3 -bail shop.sqlite < shop.sqlite.sql

The script creates each table with its keys, copies the rows and adds indexes and counters, all in one transaction with syncing off, so it runs at disk speed and a load that stops leaves no partial tables. Declared types, COLLATE NOCASE and RTRIM stay as they are, and sqlite_sequence continues where the source left it. STRICT is left out, so the script loads on every version from 3.31. The script ends with PRAGMA foreign_key_check, which lists any row that points to a missing row instead of failing the load.

Or load it straight in

intosql load and the desktop app load the dump into SQLite without a script, with the same tables, keys and indexes the script creates:

Terminal
intosql load shop.db --into shop.db

It stops before writing anything when one of the dump's tables already exists. It loads in one transaction, and a file the load created is removed if it fails. Loading into SQLite is free. See Loading.

What isn't converted

Views and triggers are written in SQLite's own language. They are listed in the report by kind and name, never dropped silently, so you can recreate them in SQLite. Virtual tables, such as FTS5 and R*Tree, are listed too, and their shadow tables are left out.

Data types

SQLite data types between versions lists each type that changes in the script: what it becomes, why, and the report line.

Every other rule

Generated from the engine, where each rule is decided.

In the dump In the script Why Report line
COLLATE NOCASE COLLATE NOCASE The same collation. None
generated columns generated columns, STORED or VIRTUAL as in the source SQLite computes them the same way. None
generated columns with functions IntoSQL doesn't write, and the || operator a plain column A formula that can't be written leaves a plain column and a report line; the dump has no values for generated columns. generated_not_translated
defaults with functions IntoSQL doesn't write (strftime, randomblob) no default A default that can't be written is left out and reported; CURRENT_TIMESTAMP and datetime('now') are written. default_not_translated
CHECK constraints translated The same checks, tested as each row loads. None
CHECK constraints with functions IntoSQL doesn't write not created A check that can't be written is left out and reported. check_skipped
partial and expression indexes not created IntoSQL writes indexes on whole columns; create the others again after loading. index_skipped
foreign keys (unenforced unless PRAGMA foreign_keys=ON) the same foreign keys, listed by PRAGMA foreign_key_check SQLite checks foreign keys only when told to, so rows without a parent are common: they load, and the key is left unchecked with a warning instead of stopping the load. None
foreign keys to tables the dump doesn't create not created A key can't name a table that isn't there. foreign_key_skipped
primary keys holding NULL, and NOT NULL columns with values that become NULL NOT NULL dropped SQLite lets a primary key other than the rowid hold NULL; the column can't be NOT NULL in the target, or the load would stop. not_null_relaxed
views, triggers and virtual tables (FTS, R*Tree) listed in the report Their code is written in SQLite's own language, and virtual tables need SQLite's modules; they are listed for you to recreate. Their internal tables aren't copied. not_converted_objects
dumps without CREATE TABLE (--data-only) types read from the values A data-only dump has only rows, so the types come from the values and the columns are numbered. no_schema
expressions in INSERT values other than the text functions dumps write NULL unistr(), replace(), char() and || are evaluated as the shell writes them; other expressions can't be without a database. expression_value
text decoded twice (ñ) copied as it is The dump holds exactly these characters, so they are copied; the report points them out. mojibake_suspected