PostgreSQL package organisation

[qfritz]

Creating an issue because discussing in MR is painful. 😭

Opening this topic to discuss internal/infra/postgresql package organisation.

This is the current overview of the package:

Open Mermaid Schema

```mermaid graph TD subgraph cmd run end subgraph ppm c(check) s(server) end subgraph infra config subgraph postgresql pg[POSTGRESQL] bounds column connection error partition partitionconfiguration partitionsettings partitionstrategy server table end end DB[(PostgreSQL)] %% LINKS run ----> connection pg --> column & error & table & partitionconfiguration & partitionsettings partitionsettings --> partitionstrategy partitionconfiguration --> bounds & partition pg ----> bounds & partition c ----> partitionsettings s ----> server config --> partitionconfiguration pg & server & connection & column -...- DB ```

Initial notes:

• Only a few of these files actually interact with the database
• Some structures are used to represent concepts which are used in the postgresql.go file but rest of file is used by an external func
• Func that do not interact with the databases could belong to the APP in my opinion.

[qfritz]

Proposal

• migrate partition stuff to app
• postgresql.go renamed to manager and put in cmd
  • its functions interacting with the database moved to infra but coordinated with app through the manager (querier.go or something)
• table file could merge with partition file because it has not very much more information. The type table is more a full name than an actual table.

Open Updated Mermaid Schema

```mermaid graph LR subgraph cmd run pg[POSTGRESQL
renamed to manager.go
.] end subgraph app subgraph ppp[partition] table partition partitionconfiguration[configuration] partitionsettings[settings] partitionstrategy[strategy] bounds end end subgraph infra config subgraph postgresql querier connection error server column end end DB[(PostgreSQL)] %% LINKS run ----> connection pg --> column & error & table & partitionconfiguration & partitionsettings partitionsettings --> partitionstrategy partitionconfiguration --> bounds & partition pg ----> bounds & partition & querier config --> partitionconfiguration table -.->|merge with| partition querier & server & connection & column -...- DB ```

Edit: absolutely not fixed on the naming, please consider my proposal for its concepts more than the actual names used.

[dcupif]

I'm aligned with @qfritz that the postgres package contains most of the "domain knowledge". Example: the concept of partition is only meaningful in our context, but this concept does not exist physically in Postgres, partitions are just tables. Ultimately, I would suggest something along the lines of:

internal/ (or public)
  partitioning/
    package partitioning
    Main types: Partition, Bound, Key, Manager, Strategy...
infra/
  postgres/
    package postgres
    Main types: Client, Table, Column, ColumnType, Query...

Very similar to what Eve suggests 👍

[qfritz]

partitions are just tables

That's also one of the reason I suggested to merge partition and table and their very similar functions.

[vmercierfr]

As discussed, it's now implemented :)