Using an in-memory sqlite database

#1

Hi all,

in a recent project, using Django 2.2.7, the idea was to use an additional in-memory database, as in this configuration:

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.sqlite3',
        'NAME': os.path.join(BASE_DIR, 'db.sqlite3'),
    },
    'memory': {
        'ENGINE': 'django.db.backends.sqlite3',
        'NAME': ':memory:',
    }
}

In order to get the memory db initialized, we have this command in the config for one app called memory_db_only:

class MemoryDbOnlyConfig(AppConfig):
    name = 'memory_db_only'

    def ready(self):
        call_command('migrate',
                     app_label='memory_db_only',
                     verbosity=1,
                     interactive=False,
                     database='memory')

When starting a shell (python manage.py shell) and querying some model from the memory db, it works fine:

>>> from memory_db_only.models import MyModel
>>> MyModel.objects.using("memory").all()
<QuerySet []>

However, when we add a view to the memory_db_only app, like

def get_objects(request):
    obj = MyModel.objects.using("memory").all()
    return HttpResponse(str(obj))

an exception is raised:

OperationalError at /memory/objects
no such table: memory_db_only_mymodel

I would highly appreciate advice on how to setup Django in this particular scenario.
Thanks
Christoph

#2

Did you migrate? You’d still need to do that.

#3

Pardon my novice question, but really curious why would you use in-memory database won’t it be a bottleneck?

1 Like
#4

I guess if its a small project and speed is important you could use an in memory DB.

But I think you are right @Devil - I wouldn’t use it on anything with scale.

Having said that SQLite on a SSD is pretty fast, but maybe there are other reasons for using in memory SQLite

#5

Update: Using file::memory: for the name of the memory database fixes the problem.

@ carltongibson: The migrations are run automatically whenever the server restarts, due to the changes to MemoryDbOnlyConfig

@ Devil: In our current project, we would like to use a relational cache, in order to store relational data from an external resource in a non-permanent manner.

@ andyide: Since we only use it for caching limited amounts of data, we don’t expect to see any scalability issues.

1 Like