为什么SELECT COUNT（）查询执行计划包括左连接表？

在SQL Server 2012中，我具有连接到另一个表的表值函数，我需要为此“表值函数”计算行数。检查执行计划时，可以看到左侧的联接表。为什么？左连接表如何影响返回的行数？我希望数据库引擎不需要在SELECT count（..）查询中评估左联合表。

Select count(realtyId) FROM [dbo].[GetFilteredRealtyFulltext]('"praha"')

执行计划：

表值函数：

CREATE FUNCTION [dbo].[GetFilteredRealtyFulltext]
(@criteria nvarchar(4000))
RETURNS TABLE
AS
RETURN (SELECT 
realty.Id AS realtyId,
realty.OwnerId,
realty.Caption AS realtyCaption,
realty.BusinessCategory,
realty.Created,
realty.LastChanged,
realty.LastChangedType,
realty.Price,
realty.Pricing,
realty.PriceCurrency,
realty.PriceNote,
realty.PricePlus,
realty.OfferState,
realty.OrderCode,
realty.PublishAddress,
realty.PublishMap,
realty.AreaLand,
realty.AreaCover,
realty.AreaFloor,
realty.Views,
realty.TopPoints,
realty.Radius,
COALESCE(realty.Wgs84X, ruian_cobce.Wgs84X, ruian_obec.Wgs84X) as Wgs84X,
COALESCE(realty.Wgs84Y, ruian_cobce.Wgs84Y, ruian_obec.Wgs84Y) as Wgs84Y,
realty.krajId,
realty.okresId,
realty.obecId,
realty.cobceId,
IsNull(CONVERT(int,realty.Ranking),0) as Ranking,

realty.energy_efficiency_rating,
realty.energy_performance_attachment,
realty.energy_performance_certificate,
realty.energy_performance_summary,

Category.Id AS CategoryId,
Category.ParentCategoryId,
Category.WholeName,
okres.nazev AS okres,
ruian_obec.nazev AS obec,
ruian_cobce.nazev AS cobce,
ExternFile.ServerPath,
Person.ParentPersonId,
( COALESCE(ftR.Rank,0) + COALESCE(ftObec.Rank,0) + COALESCE(ftOkres.Rank,0) + COALESCE(ftpobvod.Rank,0)) AS FtRank

FROM realty
JOIN Category ON realty.CategoryId = Category.Id
LEFT JOIN ruian_cobce ON realty.cobceId = ruian_cobce.cobce_kod
LEFT JOIN ruian_obec ON realty.obecId = ruian_obec.obec_kod
LEFT JOIN okres ON realty.okresId = okres.okres_kod
LEFT JOIN ExternFile ON realty.Id = ExternFile.ForeignId AND ExternFile.IsMain = 1 AND ExternFile.ForeignTable = 5
INNER JOIN Person ON realty.OwnerId = Person.Id
Left JOIN CONTAINSTABLE(Realty, *, @criteria) ftR ON realty.Id = ftR.[Key] 
Left JOIN CONTAINSTABLE(ruian_obec, *, @criteria) ftObec ON realty.obecId = ftObec.[Key] 
Left JOIN CONTAINSTABLE(Okres, *, @criteria) ftOkres ON realty.okresId = ftOkres.[Key]
Left JOIN CONTAINSTABLE(pobvod, *, @criteria) ftpobvod ON realty.pobvodId = ftpobvod.[Key]
WHERE Person.ConfirmStatus = 1
AND ( COALESCE(ftR.Rank,0) + COALESCE(ftObec.Rank,0) + COALESCE(ftOkres.Rank,0) + COALESCE(ftpobvod.Rank,0))  > 0
)

更新：

我添加唯一索引来遵循Rob Farley的想法：

 Create unique nonclustered index ExternFileIsMainUnique ON ExternFile(ForeignId) WHERE IsMain = 1 AND ForeignTable = 5

并由数据库引擎建议索引：

CREATE NONCLUSTERED INDEX [RealtyOwnerLocation] ON [dbo].[Realty]

（[OwnerId] ASC）包括（[Id]，[okresId]，[obecId]，[pobvodId]）GO

为简单起见，我删除了条件

WHERE Person.ConfirmStatus = 1

来自上面的表值函数。

现在执行计划要简单得多，但仍然可以使用表ExternFile：

也许SQL Server不够聪明？

如果ForeignId, ForeignTable, IsMain不知道*在中是唯一的ExternFile，则QO将需要包括该表以计算计数。只要多行匹配，计数都会受到影响。

在SQL Server
设计中加入简化以进行简化（SQLBits记录）

_{* 优化器当前无法将过滤的唯一索引识别为唯一}

UPDATE（通过OP）：解决方案是从LEFT JOIN（可以产生多行）中更改查询中的行：

LEFT JOIN ExternFile ON realty.Id = ExternFile.ForeignId AND ExternFile.IsMain = 1 AND ExternFile.ForeignTable = 5

使用TOP进行OUTER APPLY（产生一行，不影响COUNT）

OUTER APPLY (SELECT TOP (1) ServerPath FROM ExternFile WHERE ForeignId = realty.Id AND IsMain = 1 AND ForeignTable = 5) AS ExternFile

现在查询更加有效。无法添加唯一索引，因为值不是唯一的，它们仅对于条件中的组合才是唯一的，并且如上所述不视为唯一。