Database & SQL tools
Read CREATE TABLE definitions as a graph, field inventory and exportable structure. Choose the dialect used by your DDL.
Enter SQL or choose an example.
Processed locally in your browser. No database connection or saved history.
CREATE TABLE and simple named-column CREATE INDEX only. Relationships require explicit foreign keys; SQL is never executed.
Up to 50,000 UTF-16 units, 50 tables, 1,000 columns and 200 keys/indexes/checks/foreign keys combined. Fields appear 50 at a time.
TypeScript describes read rows: exact large numbers and decimals use string; unknown and SQLite types stay unknown. Review your driver mapping.
Result is out of date. Process the current input before exporting.
A SQL schema visualizer is useful when a migration review starts with a file rather than a running database. Seeing table names is easy; checking which columns form a composite key or which references leave the supplied schema takes more care.
Use the sample to compare the diagram with its field inventory. A column called customer_id is not evidence of a relationship on its own. Only an explicit foreign-key definition creates an arrow.
Keep declared structure separate from assumptions about the application.
The SQL schema visualizer draws the foreign key that actually exists and leaves an unrelated table separate. A missing referenced table becomes a dashed external node, so a partial schema does not look complete.
A diagram helps with orientation, JSON keeps the structural inventory, and TypeScript gives a starting row shape. These outputs share one parsed result, so switching views cannot quietly reinterpret the SQL.
The selector changes parsing and identifier rules; it does not translate SQL. MySQL uses table-level FOREIGN KEY constraints, while PostgreSQL and SQLite also accept column references.
An arrow can name a table that is not included in the input. Inspect the dashed node and compare its target columns before treating the diagram as a complete database schema diagram.
Exact decimals and large integers export as string; nullable known types get a null union. Defaults do not make properties optional, because the output describes a read row.
The DDL viewer reads a supported subset of MySQL, PostgreSQL and SQLite. Quoted identifiers and qualified table names survive normalization, with PostgreSQL unquoted names folded to lowercase.
Primary and unique keys retain their ordered column lists. Foreign key relationships retain both sides of each mapping; bridge tables remain real tables instead of becoming an invented many-to-many shortcut.
Select a table to review type spelling, nullability and primary-key membership. Original column definitions preserve DEFAULT, CHECK and generated clauses as text. Explicit indexes retain their column order and direction.
The JSON output contains table identities, columns, keys, indexes, checks and references, including unresolved targets. It is a schema inventory for documentation or scripts, not a JSON Schema validator.
Generated interfaces use safe numbered names and quoted property keys. PostgreSQL booleans become boolean, binary data becomes Uint8Array, and JSON or unrecognized types stay unknown. The mapping is a starting convention, not a driver guarantee.
SQL to ER diagram output is available as an SVG with table cards, field previews and directed links. The complete field inventory stays in JSON even when a diagram card abbreviates long names or shows only five fields.
Select MySQL, PostgreSQL or SQLite before analysis. Support is limited to the accepted CREATE TABLE and simple CREATE INDEX grammar; a database vendor supporting a construct does not mean this parser supports it. The error position helps identify a statement to simplify.
Remove data and migration commands first. INSERT, ALTER and other unsupported statements stop the entire batch. This avoids showing a plausible diagram that quietly omits columns or foreign keys introduced later in a dump.
They are not inferred. A field name ending in _id remains just a field until the DDL contains an explicit constraint. If a declared target is missing from the supplied text, the reference is kept and marked external or unresolved.
Yes. A PRIMARY KEY(a,b) remains one key with two ordered columns, and a matching composite foreign key preserves the same positional mapping. Two separate primary-key declarations on one table are rejected; individual composite members are not presented as independently unique.
Download SVG for the diagram, JSON for the complete structural inventory or a .ts file for row interfaces. JSON and TypeScript also support copying. SVG cards preview five fields; use the field view or JSON for the rest. No PNG, DBML or live database export is offered.
Found a formatting problem?
Bring a small, self-contained DDL sample to the SQL schema visualizer and check its declared keys before reusing the outputs.
Edit the DDL to change the structure; every result remains a read-only view.