dev notes

software development craftsmanship

/ home [ / blog / letter / springer link / brain / relaunch / time for money, money for time / first steps with rasmus / connect the dots / get in touch with vue / Alternative content management system approach / A database gate keeper / Generate a ERM from a PostgreSQL database schema / Working with immutable data in Postgres / Automatically update MIT licenses / Moving ORM Mapping towards the database / providing test data for databases / using pandoc filters to create graphs with hakyll / get in touch with react / six months in new job / days left / minimum viable product / Repository ownership / getting better / git cleanup ]


Alternative content management system approach

In the last months, I had to manage a bigger Wordpress instance with many plugins installed. It is a kind of mess. Plugins affecting each other. Different look and feel. Different UX. Plugins which must not be plugins, solving caching issues for example. The site CMS Garden shows several approaches.

I would like to think of a CMS like an application platform, let's call it rasmus, where there plugins can be understood as a separate application. Accepting the challenge, I start something new.

At first, let's start small and build a scaffold for the application platform. You can imagine this like a fast food restaurant.


You put your request at the counter and behind the counter the magic happens. The worker behind the counter must not know how a burger is made, and how many fries must be in a fryer. He sees the current orders on a screen, and put the meals together, when the separate parts are ready. When the order is ready, the customer can be served.

Architectural overview

As you can see the database take a place in the middle.

As mentioned in the database architecture series the web backend can put a request in the database transfer table. This is the only table, which can be accessed by the web backend. After a request is processed, the result is put back into the transfer entity. The database sends a notification to the web backend, which can pull the response from the database. This represents a very thin interface to database. This approach leads to less security issues, because more database related functions are hidden from the web backend.

The database

Every application's heart is it's database. rasmus is no exception. An rasmus application has it's own schema within the database. The schemes look as following.

The core

The different rasmus applications share a core. Here you can find functions, used across all applications, like user and role management.

First start with a simple user / role management approach.

CREATE TABLE user_account(
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    first_name VARCHAR(254),
    last_name VARCHAR(254),
    email_address VARCHAR(254),
    password VARCHAR(254),
    login VARCHAR(254),
    signature VARCHAR(254)

For the time being, we store the password with the user. This field can be moved later on, when different authorization methods will be introduced.

Starting with a simple role definition,

CREATE TYPE role_level AS ENUM ('admin','user','guest');

    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    name VARCHAR(254) NOT NULL,
    description VARCHAR(254),
    role_level role_level NOT NULL DEFAULT 'guest'

users can have several roles.

CREATE TABLE user_in_role(
    id_user_account UUID NOT NULL REFERENCES user_account(id),
    id_role UUID NOT NULL REFERENCES role(id),
    PRIMARY KEY(id_user_account, id_role)

Every part of an rasmus application can be associated with a privilege

CREATE TABLE privilege(
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    name VARCHAR(80),
    description VARCHAR(254)

Each privilege can be assigned to a role.

CREATE TABLE role_privilege(
    id_role UUID NOT NULL REFERENCES role(id),
    id_privilege UUID NOT NULL REFERENCES privilege(id),
    PRIMARY KEY (id_role, id_privilege)

These are the minimal relations for user and role management storage.


The transfer relation is located in the core schema. It uses the following state definitions.

CREATE TYPE transfer_status as ENUM (

This is not carved into stone. It will fit the first requirements to a stateful exchange between the web backend and the database.

The transfer relation itself looks like

CREATE TABLE transfer(
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    status transfer_status NOT NULL DEFAULT 'pending',
    request JSONB NOT NULL,
    result JSONB

Per default, every request is a pending request. A transfer entity must have at least a request body.

At this point conventions have to be made. A request must contain

After a request is inserted into transfer,

    PERFORM pg_notify(NEW.request->>'schema',; 
$$ LANGUAGE plpgsql;

CREATE TRIGGER transfer_before_trigger BEFORE INSERT ON transfer
    FOR EACH ROW EXECUTE PROCEDURE transfer_trigger();

a notification is send to the database backend. Pattern matching is used to find the right addressee for the request. A closer look has to be made, to decide, which operation must be done inside the database, to prevent multiple round trips.


As said before, the business logic is completely separated from the web backend.


A cms starts with an article.

    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    id_author UUID NOT NULL REFERENCES core.user_account(id),
    title VARCHAR(254),
    raw text,
    html text,
    is_visible boolean NOT NULL DEFAULT FALSE,
    is_draft boolean NOT NULL DEFAULT TRUE

The author lives in the core namespace, and can be used inside the cms schema. The raw text is markdown based, the html is the generated result. A new article is per default invisible and in draft mode.

Before uploading attachments to the database, the file_type has to be set.


CREATE TABLE attachment(
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    raw bytea NOT NULL,
    file_type file_type NOT NULL DEFAULT 'binary'

As for now, the binary content is stored within the database.

Every article can have multiple attachments.

CREATE TABLE article_attachment(
    id_article UUID NOT NULL REFERENCES article(id),
    id_attachment UUID NOT NULL REFERENCES attachment(id),
    PRIMARY KEY(id_article, id_attachment)

The attachments can be addressed within the articles markdown content.

The cms provides a tree of categorys.

CREATE TABLE category(
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    name VARCHAR(254) NOT NULL,
    description VARCHAR(512),
    icon VARCHAR(1), -- font awesome
    is_active BOOLEAN NOT NULL DEFAULT false,
    is_visible BOOLEAN NOT NULL DEFAULT false,

The tree is described as a nested set.

An article can be assigned to multiple categories.

CREATE TABLE article_in_category(
    id_article UUID NOT NULL REFERENCES article(id),
    id_category UUID NOT NULL REFERENCES category(id),
    PRIMARY KEY(id_article, id_category)


Having set up the database, it comes for choosing the technology for the backend. There are several possibilities choosing the 'right' backend technology.

This will be covered with the next article.

Fell free to browse through the sources.