sdic

Asynchronous soft constraints executed against you databases


Keywords
sdic, sql, mysql, postgresql, sqlalchemy, data, integrity, constraints
License
GPL-3.0
Install
pip install sdic==0.1

Documentation

sdic

A.K.A. SQL Data Integrity Checker

CircleCI

One line purpose

sdic executes all the SQL queries found in a folder and displays its output.

More detailed purpose: Soft Constraints

Usual Constraints

In any RDBMS, you can set constraints to prevent the application to save the data in a way that's not consistent. E.g. if you want all your users to have an email, you can set the email column to NOT NULL.

This works for simple constraints:

  1. It's easy to implement
  2. It's cheap for the database to check on every change

But for more complex constraints that you'd like to set, it'd be either very expensinve to check on every write, or even impossible to write as a constraint.

Soft Constraints

With sdic, you can write you complex constraints as simple queries, and have the database run them asynchronously at the occurrence you want.

We call them "soft constraints".

Example

Let's say that you have a users table, defined like this:

  • id Primary Key NOT NULL
  • firstname NULL
  • lastname NULL
  • email NOT NULL

Now, let's suppose your application allows users to register just with their email but can fill in their firstname and lastname later on, but we don't want our users to have only a firstname or a lastname.

Simply put, our constraint is: Make sure every users has either a firstname and a lastname set, or both set to NULL.

With sdic, you can add this enforce_fullname.sql file and let sdic check that every user comply nightly.

-- Make sure every user with a name has both a firstname and a lastname
SELECT id, firstname, lastname
FROM users
WHERE
    (firstname IS NULL AND lastname IS NOT NULL) OR
    (firstname IS NOT NULL AND lastname IS NULL)
LIMIT 10
;
-- Could also be written as firstname IS NULL <> lastname IS NULL but this is
-- for people to understand the use case.

Put this file in your-environment/your-server/enforce_fullname.sql.

Edit the your-environment/servers.ini file to tell sdic how to connect to your server.

Now run sdic your-environment and it will output any user that do no comply with your soft constraint.

You can have as many soft constraints on as many servers and as many environments as you need.

Constraints that can be temporary violated

Another use case is that there are times that certain business rules can be violated for short periods. For example, you may want every department to have a head, but also to allow the creation of a new department without or NULLing out that field when somebody quits.

In those sorts of cases, a reporting tool like sdic is likely just what the doctor ordered.

Install as a cron

If you want to get an email every night to give you a list of all the soft constraints that have been broken during the last day, just add it to you crontab. We like to have it run daily, so we can fix any bug generating bad data before it becomes a real problem.

Example crontab:

MAILTO="dba@acme.com"
@daily sdic live

dba@acme.com is the email that will get the soft constraints broken every day. Make sure your local MTA is well configured on your system. You can test it by doing date | mail -s test dba@acme.com.

Databases supported

Any database supported by SQLAlchemy should be supported, including PostgreSQL and MySQL.

Install

pip install sdic

Configuration

An example configuration is given in the example-environment folder.

The script reads from a designated folder, whose path you pass as an argument. This folder should consist of the following:

  1. A servers.ini file, which contains the Database URLs (see the example-environment folder)
  2. A sub-folder, which contains the actual queries in a .sql file format

Usage

A directory argument is mandatory:

sdic path/to/your/folder

If you have e.g more than one server in a folder, but you want to only run one of them, an optional server argument can be passed as well:

sdic path/to/your/folder server1

If a query produces an output, it will look something like this:

-----===== /!\ INCOMING BAD DATA /!\ =====-----

Server: big-database
File: test_query.sql

SQL Query:
-- This is a query that returns current time.
Select now();

+---------------------+
|        now()        |
+---------------------+
| 2016-06-03 19:27:14 |
+---------------------+

FAQ

Pretty sure that's what CHECK constraints are for

Right but there are 2 main differences:

  • CHECK constraints are checked on every write, for expensive checks (e.g. one that require scanning a whole big table) it's not an option. The point here is to choose how often you run the checks, and to not have to run them on writes. For us it's once, nightly: That's when our servers are doing less.
  • MySQL simply ignores the CHECK statement.