[Docs] migrator: shadow schema

[indeyets]

I'm trying to migrate to pgkit/migrator from older slonik-based version. I'm mostly good except for understanding the usage of shadow database.

• Is it a requirement or there's some way to work without it?
• Our database servers are configured in a way where accounts do not have privileges to create new schemas in cluster. We can allow some hardcoded name for the "staging" probably. But then we need a way to specify this name via config.
• Would be nice if migrator could detect this "misconfiguration" of database server and show friendly error instead of "scary" QueryError :)

[mmkal]

Thanks for trying it out. The shadow database is not a requirement, it's only used when you're trying to autogenerate changes. It's still possible to write everything by hand like in the previous version.

My setup is to use the shadow database features on a local db in order to speed up writing migrations etc. but only run up in prod.

I agree it should detect the misconfiguration and probably point to some docs about how you might want to change your workflow. I'll look into this. In the meantime which query exactly is your database setup rejecting?

[indeyets]

It fails if I try "goto" or "check" commands (the second one is more interesting)

query: {
    parse: [Function: parse],
    name: 'create_d867d5c',
    sql: '\n' +
      '      create table if not exists $1:name(\n' +
      '        name text primary key,\n' +
      '        content text not null,\n' +
      '        status text,\n' +
      '        date timestamptz not null default now()\n' +
      '      )\n' +
      '    ',
    token: 'sql',
    values: [ 'migration' ],
    templateArgs: [Function: templateArgs]
  },
  result: undefined,
  [cause]: error: database "(redacted)-shadow_1729545545884_473647" does not exist

[indeyets]

@mmkal some of my issues were related to the fact that I use database name with dashes in it. #435 adds compatibility with such cases