Django app abstract from data model

Question

I am currently working on a small data ‘reporting’ application. The idea behind it is to be able to import data from a specific location (be it a flat file or a remote database) and report on that data (e.g. Sales) following specific ‘dimensions’ (e.g. Sales by region, by product, by customer). At first when I built it, I created a ‘Sales’ model with these dimensions as FK’s to other models (a Region model, a Product model, a Customer model). The reporting aspect of the application is simply a bunch of requests on that table ending up in charts---this works fine.

Now let’s say I want to deploy/scale this application to multiple users; each user would have its own dataset, e.g. User A wants to report on Sales (by region, product and customer) while User B wants to report on Production volumes (by product, by plant and by month). What would be the best way to handle this? Is there a way to make the data model abstract from the underlying data sets?

I’ve had a quick look at various solutions such as:

1. Django dynamic models

   This could be an interesting approach however looking at the code it seems like it's only possible to add/remove fields dynamically---what I would need is to add/remove models dynamically.

2. NoSQL or any unstructured database backend

   This could be an option from what I have seen but I have no experience with those so it's difficult to assume.

3. Any EAV type of data model

   Same as for (1) I am under the impression that this could work for fields, but rather not for entire tables.

4. Serialization using JSONField to store data

I’m struggling to see how I could take this forward using Django (at some point I feel like I’m (wrongly) trying to build a DBMS all over again).

Any help is appreciated !

Cheers, J–

Answer 1

For example, something like:

models.py:

class Sale(models.Model):
    seller = models.ForeignKey(Vendor)
    buyer = models.ForeignKey(Customer)
    product = models.ForeignKey(Product)
    region = models.ForeignKey(Region)

and views.py:

def sales_by_product(request, product_id):
    product_sales = Sale.objects.filter(product=product_id)
    return render_to_response('sales_report.html', {'sales':product_sales,}, context_instance=RequestContext(request)))

def sales_by_region(request, region_id):
    regional_sales = Sale.objects.filter(region=region_id)
    return render_to_response('sales_report.html', {'sales':regional_sales,}, context_instance=RequestContext(request)))

def all_sales(request):
    return render_to_response('sales_report.html', {'sales':Sale.objects.all(),}, context_instance=RequestContext(request)))

And if you want, in models.py, something like:

class UserProfile(models.Model):
    user = models.OneToOneField(User)
    preferred_report = models.CharField(max_length=1, choices=(('P','Product'),('R','Region'),))
    region = models.ForeignKey(Region)
    favorite_product = models.ForeignKey(Product)

Then in views.py:

def user_sales_report(request):
    profile = request.user.get_profile()
    if profile.preferred_report == "P":
        HttpResponseRedirct('/sales/product/%i/' % profile.favorite_product_id)
    elif profile.preferred_report == "R":
        HttpResponseRedirct('/sales/region/%i/' % profile.region_id)
    else:
        HttpResponseRedirect('/sales/all/')

The above forces the user to view either a region's sales or a product's sales depending on the value of preferred_report. If it were me, however, and there were no reason not to, I'd just allow the user to select the criteria for how to view sales. Like use django-filter to allow him/her to pick as many criteria from your Sale model as they'd like to view the data for.