{{toc}}

Below are some database related code snippets used in development.

h1. Replace the database with a backup

It is possible (for users with the role admin or manager) to make SQL dumps via the backend anytime: admin -> data -> SQL export. Also, every time the database_upgrade.py script, which is used at an upgrade to a new OpenAtlas version, is called before doing database changes. These backups are located at *files/export* and because the date of creation is included in the filename it is easy to see when they were made. Of course it is good practice to store these SQL backups on another machine too, e.g. in case of a hard disk failure, but this is outside the scope of OpenAtlas.

In case these backups are needed, lets say for example an import went wrong, the workflow on a Debian machine would be as follows. The commands are executed as *postgres* user and it is assumed the database and user are called **openatlas**.

* Delete the current database
<pre><code class="shell">dropdb openatlas</code></pre>

* Create a new, empty database with the owner openaltas
<pre><code class="shell">createdb openatlas -O openatlas</code></pre>

* Import the (unpacked) backup into the new database
<pre><code class="shell">psql openatlas < path/to/the/unpacked/backup.sql</code></pre>

h1. Export

h2. Database Structure

Export the database structure from the test database (used to avoid specialties in production databases) into install/1_structure.sql

<pre>
pg_dump -sc --if-exists -n model -n gis -n log -n web -n import openatlas_test > install/1_structure.sql
</pre>

h2. Model Data

<pre>
pg_dump openatlas --rows-per-insert 10000 -a -t model.cidoc_class -t model.cidoc_class_i18n -t model.cidoc_class_inheritance -t model.property -t model.property_i18n -t model.property_inheritance > install/2_data_model.sql
</pre> 

h2. Single Schema

<pre>
pg_dump -n web openatlas > /tmp/openatlas_web.sql
pg_dump -n model openatlas > /tmp/openatlas_web.sql
</pre>

h2. Windows import

There are mainly two ways of restoring a PostgreSQL database on Windows. For clarification of the commands and options, please read the "pg_dump":https://www.postgresql.org/docs/15/app-pgdump.html and "pg_restore":https://www.postgresql.org/docs/15/app-pgrestore.html documentation.

h3. Custom Format

Export on Debian server as @postgres@ user with following command:
<pre><code class="shell">
pg_dump -Fc openatlas > openatlas.dump
</code></pre>

This dump file can be used with pgAdmin4 to restore the database or with the following command in PowerShell (git bash doesn't work). Please adjust variables for your need:
<pre><code class="shell">
pg_restore.exe --host "localhost" --port "5432" --username "openatlas" --dbname "openatlas" --clean --verbose "path to dump file"
</code></pre>

h3. Plain text file (.sql) 

Export on Debian server as @postgres@ user with following command:
<pre><code class="shell">
pg_dump --attribute-inserts openatlas > openatlas.sql
</code></pre>
In the pgAdmin4 tool, select the corresponding database, run the _query tool_ and select the _open file_ option.
Load the provided dump file and execute (F5) it.
This method can be quite slow and if the database has a large size (over ~10 MB) loading the file in pgAdmin4 will be a challenge.
Also, the "psql":https://www.postgresql.org/docs/15/app-psql.html tool can be used.

h1. Reset Demo

<pre>
/var/lib/postgresql/reset_demo.sh
</pre>

h1. Add update trigger for modified field

Replace *schema.table* below accordingly and execute:
<pre>
CREATE TRIGGER update_modified BEFORE UPDATE ON schema.table FOR EACH ROW EXECUTE PROCEDURE model.update_modified();
</pre>

h1. Data clean up

h2. Find used but missing places

e.g. after a case study separation (not sure if this statement is now showing really missing or just needed locations)

<pre><code class="sql">
SELECT e.id, e.name, e.system_type FROM model.entity e
WHERE e.class_code = 'E53' AND e.system_type = 'place location' AND e.id IN (
   SELECT r.id FROM model.link lr JOIN model.entity r ON lr.range_id = r.id AND lr.property_code IN ('P74', 'OA8', 'OA9', 'P7'))
   AND e.id NOT IN (SELECT range_id FROM model.link WHERE property_code = 'P53');
</code></pre>

h2. Delete orphaned locations

<pre><code class="sql">
DELETE FROM model.entity WHERE id IN (
    SELECT id FROM model.entity WHERE system_type = 'place location' AND id NOT IN (
        SELECT e.id FROM model.entity e JOIN model.link l ON e.id = l.range_id AND l.property_code = 'P53' 
        WHERE e.class_code = 'E53' AND e.system_type = 'place location'));
</code></pre>

h2. Recursive Events

To get all child events of a given event the SQL below (replace ROOT_EVENT_ID at bottom). It works but is slow and could be improved.

<pre><code class="sql">
WITH RECURSIVE tree AS (
  SELECT e.id, ARRAY[]::INTEGER[] AS ancestors
  FROM model.entity e
  WHERE (SELECT s.id FROM model.entity s JOIN model.link l ON s.id = l.range_id AND l.domain_id = e.id AND l.property_id = (SELECT id FROM model.property WHERE code = 'P117')) IS NULL 
  UNION ALL 
  SELECT e.id, tree.ancestors || 
    (SELECT s.id FROM model.entity s JOIN model.link l ON s.id = l.range_id AND l.domain_id = e.id AND l.property_id = (SELECT id FROM model.property WHERE code = 'P117'))
  FROM model.entity e, tree
  WHERE (SELECT s.id FROM model.entity s JOIN model.link l ON s.id = l.range_id AND l.domain_id = e.id AND l.property_id = (SELECT id FROM model.property WHERE code = 'P117')) = tree.id
)
SELECT * FROM tree WHERE ROOT_EVENT_ID = ANY(tree.ancestors);

</code></pre>