dev notes

software development craftsmanship

#BlackLivesMatter

After taking a first look at the JSON columns, let's look at a few possible applications. Imagine a simple shop system with articles, prices and purchase orders.

An article can be active or inactive .

CREATE TYPE article_status AS ENUM (
    'active',
    'inactive'
);

Every article has an article_number .

CREATE TABLE article (
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    article_number VARCHAR(128) UNIQE NOT NULL DEFAULT '',
    name VARCHAR(128),
    description VARCHAR(2048),
    status article_status NOT NULL DEFAULT 'active'
);

You can see that id and article_number are unique, so both could be used as a primary key. This is not normalized in a usual way.

There are a few points, why to stick to this solution.

  • A primary key should only be used to identify a record. Not more, not less.
  • There must be no reuse for a business case, like it would be for article_number . The article_number could not be changed so easily, after being promoted to a primary key.
  • A article_number identifies an article entity not a database record.

prices with history

Every article can have a price.

CREATE TABLE price (
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    id_article UUID NOT NULL REFERENCES article(id),
    price real NOT NULL DEFAULT 0.0,
    scale_lower_limit INT NOT NULL DEFAULT 1,
    scale_upper_limit INT NOT NULL DEFAULT 2147483647,
    valid_from DATE NOT NULL DEFAULT current_date,
    valid_to DATE NOT NULL DEFAULT current_date + interval '1 year'
);

An article can have multiple prices over time. There can be multiple price scales. A price will be more likely changed than an article . Price changes may be interesting for reporting issues.

You can store these changes in a JSONB column.

ALTER TABLE price ADD COLUMN history JSONB;

Every time, a price record changes. these changes should be saved. These saved items should be immutable over time.

CREATE FUNCTION history_trigger() RETURNS TRIGGER AS $$.

BEGIN
    IF NEW.history IS NULL THEN
        NEW.history := '[]'::JSONB;
    END IF;

    NEW.history := NEW.history::JSONB || (row_to_json(OLD)::JSONB - 'history');
    RETURN NEW;
END
$$ LANGUAGE plpgsql;

CREATE TRIGGER price_history_trigger BEFORE UPDATE ON price
    FOR EACH ROW EXECUTE PROCEDURE history_trigger();

First of all, an article has to be created.

$ psql -U postgres -c "INSERT INTO test.article (article_number, name, description) \
> VALUES ('AB12345', 'Test article','Test desc')"
INSERT 0 1

For this newly inserted article

$ psql -U postgres -c "SELECT * from test.article" | cat
                  id                  | article_number |     name     | description | status |         created_at         |         updated_at         
--------------------------------------+----------------+--------------+-------------+--------+----------------------------+----------------------------
 f12def37-3de6-4985-8912-054891631499 | AB12345        | Test article | Test desc   | active | 2017-07-15 19:17:14.727931 | 2017-07-15 19:17:14.727931
(1 row)

you can add a price ,

$ psql -U postgres -c "INSERT INTO test.price (id_article, price) \
> VALUES ('f12def37-3de6-4985-8912-054891631499',50.5)"
INSERT 0 1

$ psql -U postgres -c "SELECT * FROM test.price" | cat
                  id                  |              id_article              | price | scale_lower_limit | scale_upper_limit | valid_from |  valid_to  | history |         created_at         |         updated_at         
--------------------------------------+--------------------------------------+-------+-------------------+-------------------+------------+------------+---------+----------------------------+----------------------------
 3a113796-05fd-4ff3-a33f-b08f92c01cd8 | f12def37-3de6-4985-8912-054891631499 |  50.5 |                 1 |        2147483647 | 2017-07-15 | 2018-07-15 |         | 2017-07-15 19:21:39.245331 | 2017-07-15 19:21:39.245331
(1 row)

and raise the price value for the price record.

$ psql -U postgres -c "UPDATE test.price SET price = 70.2 WHERE id = '3a113796-05fd-4ff3-a33f-b08f92c01cd8'"
UPDATE 1

$ psql -U postgres -c "SELECT * FROM test.price" | cat
                  id                  |              id_article              | price | scale_lower_limit | scale_upper_limit | valid_from |  valid_to  |                                                                                                                                                           history                                                                                                                                                            |         created_at         |         updated_at
--------------------------------------+--------------------------------------+-------+-------------------+-------------------+------------+------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------------------+----------------------------
 3a113796-05fd-4ff3-a33f-b08f92c01cd8 | f12def37-3de6-4985-8912-054891631499 |  70.2 |                 1 |        2147483647 | 2017-07-15 | 2018-07-15 | [{"id": "3a113796-05fd-4ff3-a33f-b08f92c01cd8", "price": 50.5, "valid_to": "2018-07-15", "created_at": "2017-07-15T19:21:39.245331", "id_article": "f12def37-3de6-4985-8912-054891631499", "updated_at": "2017-07-15T19:21:39.245331", "valid_from": "2017-07-15", "scale_lower_limit": 1, "scale_upper_limit": 2147483647}] | 2017-07-15 19:21:39.245331 | 2017-07-15 19:25:04.672829
(1 row)

The history is updated every time, a price record is updated.

customer

A customer is a kind of person which has a customer_number

CREATE TABLE customer (
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    id_person UUID NOT NULL REFERENCES person(id),
    customer_number VARCHAR(128) NOT NULL DEFAULT '',
    json_view JSONB
);

As you can see, a json_view column is added to the customer . The only difference between a person and a customer is the customer_number .

Analog to the persons update function we can write an update function for a customer

CREATE FUNCTION update_json_view_customer(id_customer UUID) RETURNS VOID AS $$
DECLARE
    customer_raw JSONB;
    person_id UUID;
BEGIN

    IF NOT EXISTS (SELECT 1 FROM person p 
        JOIN customer c on p.id = c.id_person 
        WHERE p.json_view IS NOT NULL AND c.id = id_customer) THEN

        SELECT id_person FROM customer WHERE id = id_customer INTO person_id;

        RAISE NOTICE 'update json_view for person %', person_id;

        perform update_json_view_person(person_id);
    END IF;

    SELECT row_to_json(c) FROM 
        (SELECT c.id, customer_number, p.json_view AS person_json_view FROM customer c
            JOIN person p on c.id_person = p.id
            WHERE c.id = id_customer LIMIT 1) c INTO customer_raw;

    customer_raw := customer_raw || jsonb_build_object('person', customer_raw->'person_json_view');
    customer_raw := customer_raw - 'person_json_view';

    UPDATE customer SET json_view = customer_raw WHERE id = id_customer;
END
$$ LANGUAGE plpgsql;

The json_view of the person is reused.

Let's take a inserted person.

$ psql -U postgres -c "SELECT id FROM test.person"
                  id
--------------------------------------
 da44de2f-aa0a-43c5-9fed-dcbb5b6c32a2
(1 row)

and insert a new customer for this person .

$ psql -U postgres -c "INSERT INTO test.customer (customer_number, id_person) \
VALUES ('AB12345', 'da44de2f-aa0a-43c5-9fed-dcbb5b6c32a2');"
INSERT 0 1

The newly inserted customer looks like

$ psql -U postgres -c "SELECT * FROM test.customer" | cat
                  id                  |              id_person               | customer_number | json_view |         created_at         |         updated_at         
--------------------------------------+--------------------------------------+-----------------+-----------+----------------------------+----------------------------
 88a99ea7-4281-496b-9c95-3625101177ca | da44de2f-aa0a-43c5-9fed-dcbb5b6c32a2 | AB12345         |           | 2017-07-15 18:41:27.811324 | 2017-07-15 18:41:27.811324
(1 row)

Now the customer 's json_view must be filled.

$ psql -U postgres -c "SET search_path TO test,public; \
>  SELECT test.update_json_view_customer('88a99ea7-4281-496b-9c95-3625101177ca');"
 update_json_view_customer 
---------------------------

(1 row)

$ psql -U postgres -c "SELECT * FROM test.customer" | cat
                  id                  |              id_person               | customer_number |                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              json_view                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              |         created_at         |         updated_at         
--------------------------------------+--------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------------------+----------------------------
 88a99ea7-4281-496b-9c95-3625101177ca | da44de2f-aa0a-43c5-9fed-dcbb5b6c32a2 | AB12345         | {"id": "88a99ea7-4281-496b-9c95-3625101177ca", "person": {"id": "da44de2f-aa0a-43c5-9fed-dcbb5b6c32a2", "notes": null, "website": null, "addresses": [{"id": "9a78ceb0-5169-4bca-bbf5-aac54fcaa95a", "city": "Dortmund", "street": "Fakestreet", "postal_code": "44339", "address_type": "private", "house_number": "123"}, {"id": "c69ec275-0a05-42ce-80ea-1ea1b5bcbd78", "city": "Bochum", "street": "Fakeroad", "postal_code": "44866", "address_type": "work", "house_number": "321"}], "last_name": "Hake", "birth_date": null, "first_name": "Jan Frederik", "phone_numbers": [{"id": "0df74379-6512-4f54-a1a7-fee1c1605342", "phone_number": "+49231123456789", "communication_type": "private", "communication_network": "landline", "is_primary_phone_number": true}, {"id": "a9d36784-7af3-47e8-b357-2f03500f7d66", "phone_number": "+49151123456789", "communication_type": "private", "communication_network": "cellular_network", "is_primary_phone_number": false}], "email_addresses": [{"id": "ff2fee9e-490f-49b2-8e0e-69d5bddd2ca0", "email_address": "jan_hake@fake.de", "communication_type": "private", "is_primary_email_address": false}]}, "customer_number": "AB12345"} | 2017-07-15 18:41:27.811324 | 2017-07-15 18:50:47.591534
(1 row)

a little bit more beautifull

{
    "id": "88a99ea7-4281-496b-9c95-3625101177ca",
    "person": {
        "id": "da44de2f-aa0a-43c5-9fed-dcbb5b6c32a2",
        "notes": null,
        "website": null,
        "addresses": [{
            "id": "9a78ceb0-5169-4bca-bbf5-aac54fcaa95a",
            "city": "Dortmund",
            "street": "Fakestreet",
            "postal_code": "44339",
            "address_type": "private",
            "house_number": "123"
        }, {
            "id": "c69ec275-0a05-42ce-80ea-1ea1b5bcbd78",
            "city": "Bochum",
            "street": "Fakeroad",
            "postal_code": "44866",
            "address_type": "work",
            "house_number": "321"
        }],
        "last_name": "Hake",
        "birth_date": null,
        "first_name": "Jan Frederik",
        "phone_numbers": [{
            "id": "0df74379-6512-4f54-a1a7-fee1c1605342",
            "phone_number": "+49231123456789",
            "communication_type": "private",
            "communication_network": "landline",
            "is_primary_phone_number": true
        }, {
            "id": "a9d36784-7af3-47e8-b357-2f03500f7d66",
            "phone_number": "+49151123456789",
            "communication_type": "private",
            "communication_network": "cellular_network",
            "is_primary_phone_number": false
        }],
        "email_addresses": [{
            "id": "ff2fee9e-490f-49b2-8e0e-69d5bddd2ca0",
            "email_address": "jan_hake@fake.de",
            "communication_type": "private",
            "is_primary_email_address": false
        }]
    },
    "customer_number": "AB12345"
}

a purchase process

Now we have a customer and articles with prices . The next step is to buy something. But first we take a look at a common workflow, when you buy something in a shop.

digraph { 
    rankdir="LR";   

    node [fontname="helvetica"];
    graph [fontname="helvetica"];
    edge [fontname="helvetica"];
    
    subgraph cluster_0 {
        style=filled;
        color=lightgrey;
        node [style="filled,rounded",color=white,shape=box];
        po -> rts;  
        po -> item [label="add"];
        po -> item [label="modify"];
        po -> item  [label="remove"];
        label = "mutable\ndata";
    } 
 
    subgraph cluster_1 {
        style=filled;
        color=lightgrey;
        node [style="filled,rounded",color=white,shape=box];
        rts -> snd -> delivered -> rfi -> inv;  
        item -> snd [label="frozen\npurchase\norder\nitems"];
        label = "immutable\ndata";
    } 


    start [ label="start", shape=Mdiamond];
    end [ label="end", shape="Msquare"];
 
    delivered [ label="delivered"];
    po [ label="purchase\norder"];
    rts [ label="ready to\nsend"];
    rfi [ label="ready for\ninvoice"];
    inv [ label="invoiced"];
    snd [ label="send"];
    item [ label="items"];
 
    start -> po;
    inv -> end;
}

When you enter a web shop, you can search for articles, and put them into a shopping cart. While you are creating your cart, you can change the amount of an article, or delete previously added ones. When you are satisfied with your selection, you finalize your requisition. After finalization, parts of your cart like the items can not be changed any more.

Article descriptions or names can be changed over time. Prices may vary.

a purchase order

A purchase order can have one of the following states

CREATE TYPE purchase_order_status AS ENUM (
    'requisition', 
    'ready_to_send', 
    'send', 
    'delivered',
    'ready_for_invoice',
    'invoiced',
    'finalized');

It is assumed, that every purchase order has a relation to a customer . For this example this is enough.

CREATE TABLE purchase_order (
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    id_customer UUID NOT NULL REFERENCES customer(id),
    purchase_order_number VARCHAR(128) NOT NULL UNIQUE DEFAULT '',
    status purchase_order_status NOT NULL DEFAULT 'requisition'
);

Every purchase order has a unique purchase_order_number .

CREATE TABLE purchase_order_item (
    id UUID NOT NULL PRIMARY KEY DEFAULT gen_random_uuid(),
    id_purchase_order UUID NOT NULL REFERENCES purchase_order(id),
    id_article UUID NOT NULL REFERENCES article(id),
    id_price UUID NOT NULL REFERENCES price(id),
    amount int NOT NULL DEFAULT 1
);

A purchase_order_item has a reference to a purchase_order .

Unless we are in a mutable state, There is no need for storing extra data. This changes, when the purchase_order_status changes to send . The purchase_order_items can't be changed any more. The purchased items are on their way to the customer . The only thing, which can change is the purchase_order_status , but only forward in the chain.

This is the point, where the items should be saved in a immutable way. There are no immutable types in Postgres, but it can be made hard for a process to change such data columns, which should not be updated.

For our example, we must store the article, with it's price at the time of purchase. We also need the customer , who must have some kind of address, to send the delivery to.

ALTER TABLE purchase_order ADD COLUMN frozen_purchase_order JSONB;

This column should be updated, when the purchase_order_status is set to send . So, we need a trigger function which listens on state changes. The scaffold looks like

CREATE FUNCTION freeze_purchase_order() RETURNS TRIGGER AS $$
BEGIN
    return NEW;
END
$$ LANGUAGE plpgsql;

CREATE TRIGGER freeze_purchase_order_trigger BEFORE UPDATE ON purchase_order
    FOR EACH ROW EXECUTE PROCEDURE freeze_purchase_order();

Now we listen to the send state. At this point, the purchase order some data have to be frozen. Everything below this state can be ignored.

IF NEW.status = ANY('{requisition,ready_to_send}'::purchase_order_status[]) THEN
    RAISE NOTICE 'nothing to do';
    RETURN NEW;
END IF;

To store some JSON objects we define some variables

DECLARE
    frozen_purchase_order JSONB;
    customer JSONB;
    purchase_order_items JSONB;
    rawItem RECORD;

The frozen_purchase_order will be the result JSON. First we get the current customer

SELECT json_view FROM customer WHERE id = NEW.id_customer INTO customer;

A purchase order item contains an article and a price. The function to get a json representation for a item can look like

CREATE FUNCTION get_json_from_item(item_id UUID) RETURNS JSONB AS $$
DECLARE
    result JSONB;
    article JSONB;
    price JSONB;
BEGIN
    SELECT row_to_json(item) FROM (SELECT id_article, id_price, amount FROM purchase_order_item WHERE id = item_id) item INTO result; 
    SELECT row_to_json(rawArticle) FROM (SELECT id, article_number, name, description FROM article WHERE id = (result->>'id_article')::UUID) rawArticle INTO article;
    SELECT row_to_json(rawPrice) FROM (SELECT id, p.price, scale_lower_limit, scale_upper_limit, valid_from, valid_to FROM price p WHERE id = (result->>'id_price')::UUID) rawPrice INTO price;

    result := result 
     || jsonb_build_object('article', article)
     || jsonb_build_object('price', price);

    result := result - 'id_article';
    result := result - 'id_price';

    RETURN result;
END
$$ LANGUAGE plpgsql;

The purchase order trigger function can consume this function as following

purchase_order_items := '[]'::JSONB;

FOR rawItem IN (SELECT get_json_from_item(id)::JSONB AS get_json FROM purchase_order_item WHERE id_purchase_order = NEW.id) 
LOOP
    purchase_order_items := purchase_order_items || rawItem.get_json;    
END LOOP;

The current NEW record must be set as root for our result JSON.

SELECT row_to_json(NEW.*) INTO frozen_purchase_order;

Then the customer and the items have to be merged into the result.

frozen_purchase_order := frozen_purchase_order 
    || jsonb_build_object('items', purchase_order_items)
    || jsonb_build_object('customer', customer);

At last, some unnecessary fields must be deleted from our frozen purchase order.

frozen_purchase_order := frozen_purchase_order - 'frozen_purchase_order';
frozen_purchase_order := frozen_purchase_order - 'id_customer';

Then we have our result.

NEW.frozen_purchase_order := frozen_purchase_order;

The complete function looks like

CREATE FUNCTION freeze_purchase_order() RETURNS TRIGGER AS $$
DECLARE
    frozen_purchase_order JSONB;
    customer JSONB;
    purchase_order_items JSONB;
    rawItem RECORD;
BEGIN
    IF NEW.status = ANY('{requisition,ready_to_send}'::purchase_order_status[]) THEN
        RAISE NOTICE 'nothing to do';
        RETURN NEW;
    END IF;
    RAISE NOTICE 'freeze';

    SELECT json_view FROM customer WHERE id = NEW.id_customer INTO customer;

    purchase_order_items := '[]'::JSONB;

    FOR rawItem IN (SELECT get_json_from_item(id)::JSONB AS get_json FROM purchase_order_item WHERE id_purchase_order = NEW.id) 
    LOOP
        purchase_order_items := purchase_order_items || rawItem.get_json;    
    END LOOP;

    SELECT row_to_json(NEW.*) INTO frozen_purchase_order;

    frozen_purchase_order := frozen_purchase_order 
        || jsonb_build_object('items', purchase_order_items)
        || jsonb_build_object('customer', customer);

    frozen_purchase_order := frozen_purchase_order - 'frozen_purchase_order';
    frozen_purchase_order := frozen_purchase_order - 'id_customer';

    NEW.frozen_purchase_order := frozen_purchase_order;

    RETURN NEW;
END
$$ LANGUAGE plpgsql;

add some data

$ psql -U postgres -c "SELECT id FROM test.customer"
                  id
--------------------------------------
 7a24ed2c-c873-4fdf-91cf-3574410acc49
(1 row)

$ psql -U postgres -c "INSERT INTO test.purchase_order (id_customer, purchase_order_number) \
> VALUES ('7a24ed2c-c873-4fdf-91cf-3574410acc49', 'PO12345');"
INSERT 0 1

$ psql -U postgres -c "SELECT * FROM test.purchase_order;"
                  id                  |             id_customer              | purchase_order_number |   status    | frozen_purchase_order |        created_at         |        updated_at         
--------------------------------------+--------------------------------------+-----------------------+-------------+-----------------------+---------------------------+---------------------------
 29e2fa06-edfc-49ed-878b-49e8ded9bb89 | 7a24ed2c-c873-4fdf-91cf-3574410acc49 | PO12345               | requisition |                       | 2017-07-16 21:15:41.81893 | 2017-07-16 21:15:41.81893
(1 row)

Now we add our article with our price .

$ psql -U postgres -c "SELECT * FROM test.article"
                  id                  | article_number |     name     | description | status |         created_at         |         updated_at         
--------------------------------------+----------------+--------------+-------------+--------+----------------------------+----------------------------
 0b177d42-368a-4cfa-bf8d-e863f4e8a1bd | AB12345        | Test article | Test desc   | active | 2017-07-16 21:06:03.668307 | 2017-07-16 21:06:03.668307
(1 row)

$ psql -U postgres -c "SELECT * FROM test.price;"
                  id                  |              id_article              | price | scale_lower_limit | scale_upper_limit | valid_from |  valid_to  | history |         created_at         |         updated_at         
--------------------------------------+--------------------------------------+-------+-------------------+-------------------+------------+------------+---------+----------------------------+----------------------------
 ac73b43d-e5ef-46dd-81e9-94291aa669c7 | 0b177d42-368a-4cfa-bf8d-e863f4e8a1bd |  50.5 |                 1 |        2147483647 | 2017-07-16 | 2018-07-16 |         | 2017-07-16 21:06:03.668307 | 2017-07-16 21:06:03.668307
(1 row)

$ psql -U postgres -c "INSERT INTO test.purchase_order_item (id_purchase_order, id_article, id_price) \
> VALUES ('29e2fa06-edfc-49ed-878b-49e8ded9bb89', '0b177d42-368a-4cfa-bf8d-e863f4e8a1bd', 'ac73b43d-e5ef-46dd-81e9-94291aa669c7')"
INSERT 0 1

Changing the state to ready_to_send will result

$ psql -U postgres -c "SET search_path TO test,public; UPDATE purchase_order SET status = 'ready_to_send' \ 
> WHERE id = '29e2fa06-edfc-49ed-878b-49e8ded9bb89'"
NOTICE:  nothing to do
UPDATE 1

Now we set the state to 'send'

$ psql -U postgres -c "SET search_path TO test,public; UPDATE purchase_order SET status = 'send' \
> WHERE id = '29e2fa06-edfc-49ed-878b-49e8ded9bb89'"
NOTICE:  freeze
UPDATE 1

$ psql -U postgres -c "SELECT * from test.purchase_order" | cat
                  id                  |             id_customer              | purchase_order_number | status |                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  frozen_purchase_order                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  |        created_at         |        updated_at         
--------------------------------------+--------------------------------------+-----------------------+--------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------+---------------------------
 29e2fa06-edfc-49ed-878b-49e8ded9bb89 | 7a24ed2c-c873-4fdf-91cf-3574410acc49 | PO12345               | send   | {"id": "29e2fa06-edfc-49ed-878b-49e8ded9bb89", "items": [{"price": {"id": "ac73b43d-e5ef-46dd-81e9-94291aa669c7", "price": 50.5, "valid_to": "2018-07-16", "valid_from": "2017-07-16", "scale_lower_limit": 1, "scale_upper_limit": 2147483647}, "amount": 1, "article": {"id": "0b177d42-368a-4cfa-bf8d-e863f4e8a1bd", "name": "Test article", "description": "Test desc", "article_number": "AB12345"}}], "status": "send", "changes": null, "customer": {"id": "7a24ed2c-c873-4fdf-91cf-3574410acc49", "person": {"id": "35b40b2f-bf40-4f71-8319-f7757de3e1f4", "notes": null, "website": null, "addresses": [{"id": "7a7f1e44-f6a6-495e-893b-e5806289ea81", "city": "Dortmund", "street": "Fakestreet", "postal_code": "44339", "address_type": "private", "house_number": "123"}, {"id": "1d8c41e5-bcd6-4842-864d-62c4da2fc506", "city": "Bochum", "street": "Fakestreet", "postal_code": "44866", "address_type": "work", "house_number": "321"}], "last_name": "Hake", "birth_date": null, "first_name": "Jan Frederik", "phone_numbers": [{"id": "618d19d6-3daf-4029-8d0f-1535272ec212", "phone_number": "+49123456789", "communication_type": "private", "communication_network": "landline", "is_primary_phone_number": true}, {"id": "81403e1d-1055-4953-8b1a-fcca9d034b1b", "phone_number": "+49151123456789", "communication_type": "private", "communication_network": "cellular_network", "is_primary_phone_number": false}], "email_addresses": [{"id": "a2f79f62-a497-4bdb-8f1b-03d6b7aacb30", "email_address": "jan_hake@fake.de", "communication_type": "private", "is_primary_email_address": false}]}, "customer_number": "AB123456"}, "created_at": "2017-07-16T21:15:41.81893", "updated_at": "2017-07-16T21:25:03.02978", "purchase_order_number": "PO12345"} | 2017-07-16 21:15:41.81893 | 2017-07-16 21:26:44.87835
(1 row)

The frozen purchase order looks like

{
    "id": "29e2fa06-edfc-49ed-878b-49e8ded9bb89",
    "items": [{
        "price": {
            "id": "ac73b43d-e5ef-46dd-81e9-94291aa669c7",
            "price": 50.5,
            "valid_to": "2018-07-16",
            "valid_from": "2017-07-16",
            "scale_lower_limit": 1,
            "scale_upper_limit": 2147483647
        },
        "amount": 1,
        "article": {
            "id": "0b177d42-368a-4cfa-bf8d-e863f4e8a1bd",
            "name": "Test article",
            "description": "Test desc",
            "article_number": "AB12345"
        }
    }],
    "status": "send",
    "changes": null,
    "customer": {
        "id": "7a24ed2c-c873-4fdf-91cf-3574410acc49",
        "person": {
            "id": "35b40b2f-bf40-4f71-8319-f7757de3e1f4",
            "notes": null,
            "website": null,
            "addresses": [{
                "id": "7a7f1e44-f6a6-495e-893b-e5806289ea81",
                "city": "Dortmund",
                "street": "Fakestreet",
                "postal_code": "44339",
                "address_type": "private",
                "house_number": "123"
            }, {
                "id": "1d8c41e5-bcd6-4842-864d-62c4da2fc506",
                "city": "Bochum",
                "street": "Fakestreet",
                "postal_code": "44866",
                "address_type": "work",
                "house_number": "321"
            }],
            "last_name": "Hake",
            "birth_date": null,
            "first_name": "Jan Frederik",
            "phone_numbers": [{
                "id": "618d19d6-3daf-4029-8d0f-1535272ec212",
                "phone_number": "+49123456789",
                "communication_type": "private",
                "communication_network": "landline",
                "is_primary_phone_number": true
            }, {
                "id": "81403e1d-1055-4953-8b1a-fcca9d034b1b",
                "phone_number": "+49151123456789",
                "communication_type": "private",
                "communication_network": "cellular_network",
                "is_primary_phone_number": false
            }],
            "email_addresses": [{
                "id": "a2f79f62-a497-4bdb-8f1b-03d6b7aacb30",
                "email_address": "jan_hake@fake.de",
                "communication_type": "private",
                "is_primary_email_address": false
            }]
        },
        "customer_number": "AB123456"
    },
    "created_at": "2017-07-16T21:15:41.81893",
    "updated_at": "2017-07-16T21:25:03.02978",
    "purchase_order_number": "PO12345"
}

This approach looks promising. In the next part , we look into updating those structures.