Reputation: 88
I am accessing an API where the number of fields can change at any time, but I must store and display the data as a table. Therefore, each record from the API is stored as a single record per field. My problem is that I am having trouble working out how I would order by multiple columns at a time. Putting all of the data into a 2D array (list of lists) before sorting is not a viable option as the number of records could be too large to feasibly hold in memory.
I've put together a simple example to explain. If anyone has an idea on how to overcome the problem, or how I could redesign my approach, I'd be very grateful.
| record_id | field | data |
| 1 | x | 2 |
| 1 | y | 1 |
| 1 | z | 3 |
| 2 | x | 30 |
| 2 | y | 42 |
| 2 | z | 7 |
| 3 | x | 53 |
| 3 | y | 2 |
| 3 | z | 7 |
If ordering by fields 'z' then 'x' (both ascending), the record order would be 1,2,3
If ordering by fields 'z' then 'y' (both ascending), the record order would be 1,3,2
I am using models in django to store and I am using QuerySets to retrieve the data. I don't have any control over the API or database from which I am originally accessing the data.
Upvotes: 1
Views: 360
Reputation: 88
After a fair amount of research I realised I was going about this all wrong - I am now using an hstore field in postgres and django-hstore to utilise it, for a schema-less approach. I now have a single row per original record and I can order_by after casting the required field in an 'extra' method.
Upvotes: 1