python – 如何将这个复杂的SQL转换为Django模型查询？

我有两个非常简单的模型,如下所示：

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')

这是我填充的虚拟数据：

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)

我想找到过去一周内年度低点的所有股票.

但是为了使这个问题更简单,只要假设我想找到自“2017-05-04”发生在“2018-04-30”之后或之后的最低点的所有股票.下面是我写的SQL来找到它.有用.

但是我需要帮助找出要写的Django Query以获得与此SQL相同的结果.我该怎么做？

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)

我们可以使用Django的aggregates with SubQuery expressions将查询翻译成Django ORM：

>创建子查询以检索每个符号的最低关闭：

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')

>细分：

>过滤查询集以仅获取具有trading_date> =’2017-05-04’的股票.
>“GROUP BY”stock__symbol(在Djnago中分组的例子：GROUP BY ... MIN/MAX,GROUP BY ... COUNT/SUM).
>为每个元素注释最低(低)价格.
>再次过滤查询集以仅获取在trading_date> =’2018-04-30’上发生低字段的对象.

>中级结果：

虽然我们无法在此阶段获得结果,但子查询将如下所示：

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

我们错过了trading_date.

>利用子查询检索特定的StockHistory对象：

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')

>细分：

> lows.values(‘stock__symbol’)和lows.values(‘low’)从子查询中检索相应的值.
>根据lows子查询值过滤查询集.同时过滤指定日期,以消除在该日期之前发生的低收盘价.
>获取指定的值.
>按stock__symbol排序结果(默认为升序).

>结果：

[
    {
        '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'
    }
]