schema manager

[pounard]

The plan

• First, this is going to add a schema read API, for introspecting schema.
• Then, and API for altering it.
• Third, normalizing reading and writing to be ISO, and able to diff.
• Maybe a visitor pattern for users, to do whatever they want to do with it, for navigating through the schema as a graph.

What's working

• Full MySQL support ✔️
• Full MariaDB support ✔️
• Full PostgreSQL support ✔️
• Partial SQLite support 🐗

What's missing

• SQLServer support (I have a 6.7 linux kernel, and SQL Server crashes with it)
• SQLite constraints ADD, DROP and ALTER, because they require
• Allow arbitrary SQL queries during schema transaction, lets say ->query(fn (QueryBuilder): void), which does not directly executes the SQL but stacks it into the change log (after all, schema transactions are basically upgrade procedures, and the user probably wants to be able to add data alteration along the way) Done it !
• ALTER TABLE x MODIFY COLUMN y OK done it!
• ALTER CONSTRAINT x RENAME TO y
• Reading indexes
• Reading ISO to what we are able to write
• More tests
• AaAAAAAahhhh COLLATIONS !!! Terrible thing to handle... Some SQL vendors don't want to you to read column collation, some others have terrible names, some other are so dynamic that from one container image to another collation names are not the same...
• Better column default values handling
• A method for not sending queries on transaction apply, but a dry-run result instead.
• Logging !

Future plans

• Types normalization, for identifying sequences, serials, identity, etc...
• Ability for read the schema, then recreate it identically from the definition, required for:
  • Emulation of ADD / DROP / RENAME in SQLite
  • Schema diff

[SimonMellerin]

This API seems really nice to use!

congrats!