How do I convert this complex SQL into a Django model query?

Question

I'm writing a Python/Django application to do some stock analysis.

I have two very simple models that look like this:

class Stock(models.Model):
    symbol = models.CharField(db_index=True, max_length=5, null=False, editable=False, unique=True)

class StockHistory(models.Model):
    stock = models.ForeignKey(Stock, related_name='StockHistory_stock', editable=False)
    trading_date = models.DateField(db_index=True, null=False, editable=False)
    close = models.DecimalField(max_digits=12, db_index=True, decimal_places=5, null=False, editable=False)

    class Meta:
        unique_together = ('stock', 'trading_date')

This is the dummy data I have populated them with:

import datetime
a = Stock.objects.create(symbol='A')
b = Stock.objects.create(symbol='B')
c = Stock.objects.create(symbol='C')
d = Stock.objects.create(symbol='D')

StockHistory.objects.create(trading_date=datetime.date(2018,1,1), close=200, stock=a)
StockHistory.objects.create(trading_date=datetime.date(2018,1,2), close=150, stock=a)
StockHistory.objects.create(trading_date=datetime.date(2018,1,3), close=120, stock=a)
StockHistory.objects.create(trading_date=datetime.date(2018,4,28), close=105, stock=a)
StockHistory.objects.create(trading_date=datetime.date(2018,5,3), close=105, stock=a)

StockHistory.objects.create(trading_date=datetime.date(2017,5,2), close=400, stock=b)
StockHistory.objects.create(trading_date=datetime.date(2017,11,11), close=200, stock=b)
StockHistory.objects.create(trading_date=datetime.date(2017,11,12), close=300, stock=b)
StockHistory.objects.create(trading_date=datetime.date(2017,11,13), close=400, stock=b)
StockHistory.objects.create(trading_date=datetime.date(2017,11,14), close=500, stock=b)

StockHistory.objects.create(trading_date=datetime.date(2018,4,28), close=105, stock=c)
StockHistory.objects.create(trading_date=datetime.date(2018,4,29), close=106, stock=c)
StockHistory.objects.create(trading_date=datetime.date(2018,4,30), close=107, stock=c)
StockHistory.objects.create(trading_date=datetime.date(2018,5,1), close=108, stock=c)
StockHistory.objects.create(trading_date=datetime.date(2018,5,2), close=109, stock=c)
StockHistory.objects.create(trading_date=datetime.date(2018,5,3), close=110, stock=c)
StockHistory.objects.create(trading_date=datetime.date(2018,5,4), close=90, stock=c)

I want to find all the stocks that made a yearly low within the past week.

But to make this question simpler, just assume that I want to find all the stocks whose lowest point since '2017-05-04' occurred on or after '2018-04-30'. Below is the SQL I wrote to find it. It works.

But I need help figuring out what Django Query to write to get the same results as this SQL. How can I do it?

mysql> select
    ->     s.symbol,
    ->     sh.trading_date,
    ->     low_table.low
    -> from
    ->     (
    ->         select
    ->             stock_id,
    ->             min(close) as low
    ->         from
    ->             stocks_stockhistory
    ->         where
    ->             trading_date >= '2017-05-04'
    ->         group by
    ->             stock_id
    ->     ) as low_table,
    ->     stocks_stockhistory as sh,
    ->     stocks_stock as s
    -> where
    ->     sh.stock_id = low_table.stock_id
    ->     and sh.stock_id = s.id
    ->     and sh.close = low_table.low
    ->     and sh.trading_date >= '2018-04-30'
    -> order by
    ->     s.symbol asc;
+--------+--------------+-----------+
| symbol | trading_date | low       |
+--------+--------------+-----------+
| A      | 2018-05-03   | 105.00000 |
| C      | 2018-05-04   |  90.00000 |
+--------+--------------+-----------+
2 rows in set (0.02 sec)

Answer 1

EDIT: I managed to reform the solution using Django subqueries.

We can translate the query to Django ORM using Django's aggregates with SubQuery expressions:

1. Create a subquery to retrieve the lowest close for every symbol:

   from django.db.models import OuterRef, Subquery, Min     
   
   lows = StockHistory.objects.filter(
       stock=OuterRef('stock'), 
       trading_date__gte='2017-05-04'
   ).values('stock__symbol')
   .annotate(low=Min('close'))
   .filter(trading_date__gte='2018-04-30')
   

   • Breakdown:

     • filter the queryset to get only the stocks with trading_date >= '2017-05-04'.
     • "GROUP BY" stock__symbol (examples of group by in Djnago: GROUP BY ... MIN/MAX, GROUP BY ... COUNT/SUM).
     • annotate the lowest (low) price to every element.
     • filter the queryset again to get only the objects with a low field occurring on trading_date >= '2018-04-30'.

   • Intermediate Result:

     Although we cannot get a result at this stage, the subquery will look like this:

     [
         {'stock__symbol': 'A', 'low': Decimal('105.00000')},            
         {'stock__symbol': 'C', 'low': Decimal('90.00000')}
     ]
     

     We are missing the trading_date.

2. Utilize the subquery to retrieve the specific StockHistory objects:

   StockHistory.objects.filter(
       stock__symbol=Subquery(lows.values('stock__symbol')),
       close=Subquery(lows.values('low')),
       trading_date__gte='2018-04-30'
   ).values('stock__symbol', 'trading_date', 'close')
   .order_by('stock__symbol')
   

   • Breakdown:

     • lows.values('stock__symbol') and lows.values('low') retrieve the respective values from the subquery.
     • filter the queryset against the lows subquery values. Also filter against the specified date in order to eliminate low close prices occurring before that date.
     • Get the specified values.
     • Order the result by stock__symbol (by default ascending).

   • Result:

     [
         {
             'close': Decimal('105.00000'), 
             'trading_date': datetime.date(2018, 5, 3), 
             'stock__symbol': 'A'
         }, 
         {
             'close': Decimal('90.00000'), 
             'trading_date': datetime.date(2018, 5, 4), 
             'stock__symbol': 'C'
         }
     ]
     

Answer 2

For newer versions of Django (1.11, 2.0):

from django.db.models import Min
low_stocks_qs = StockHistory.objects.filter(trading_date__gt='2017-05-04').annotate(low=Min('close')).filter(trading_date__gte='2018-04-30').order_by('stock__symbol')

You can iterate through the queryset to get individual values of low and stock.symbol, maybe something like this:

low_stocks_dict = {}
for inst in low_stocks_qs:
    low_stocks_dict[inst.stock.Symbol] = inst.low