Your database schema, as readable code.

A readable schema language

Tables, logical types, primary and foreign keys (->), indexes (partial and expression indexes), checks and enums are written in .tabliva files. Dialect-specific details are added with annotations (@postgres(type: "jsonb")); views, triggers and functions with raw sql blocks. The editor has syntax highlighting, completion and find.

Two-way ER diagram

The diagram updates as you type; on a syntax error the last valid state stays and the error is shown with its line. Adding, renaming or deleting tables, columns, relationships and indexes, or changing a type in the diagram rewrites only the relevant line of code; comments are kept. The layout is kept in a separate file (diagram.tabliva-layout.json).

Problems: 55 rules

Keys, missing foreign key suggestions with a confidence level, types (e.g. float for money), indexes, duplicated data, naming, structure and dialect compatibility. Every problem has a plain explanation and a fix; safe fixes can be applied all at once. Ignoring a problem or turning a rule off is written to the kurallar.tabliva file.

Normalization

Functional dependencies are declared in code with a depends line or suggested from column names. For each table, the steps from 1NF to BCNF, fixes with a before / after preview (move to another table, new table, child table, deliberate copy) and data migration SQL for the dialect are shown.

Change plan and migration files

The code is compared with a target (migrations folder, snapshot; on the desktop, a live database). Renames are detected; each step is labelled with risks such as data loss, table rewrite or locking. SQL is generated for PostgreSQL, MySQL / MariaDB and SQLite; the migration file is written to the migrations/ folder with a dated name and its checksum is added to tabliva.sum.

Import an existing schema

SQL dumps (pg_dump, mysqldump, SQLite .schema or hand-written CREATE TABLE), DBML and another Tabliva project are read; one file per table or a single file can be chosen. Only the schema is taken; INSERT and COPY lines are skipped. On the desktop, the schema can also be read directly from a database.

All features