Sql server SQL Server内联视图错误？

我使用的是SQL Server 2012，我认为内联视图存在一个非常奇怪的问题

首先，我有一个问题：

select 
    parcelnumber, 
    sectionname,
    value.Value CalculatedValue,
    overriddenvalue.Value OverriddenValue,
    homk.Value CalculatedHomeownersValue,
    cv.CharacteristicValue,
    pdt.PropertyDataTypeID
from landsections ls
join lands l 
  on ls.landid = ls.landid
join parcels p
  on l.ParcelID = p.ParcelID
join CharacteristicValueGroups cvg
  on cvg.ParcelObjectID = ls.ParcelObjectID
join CharacteristicValues cv 
  on cvg.CharacteristicValueGroupID = cv.CharacteristicValueGroupID
join Characteristics c
  on cv.CharacteristicID = c.CharacteristicID
join CharacteristicDetails cd
  on c.CharacteristicDetailsID = cd.CharacteristicDetailsID
 and cd.name = '09 Composite Adjustment'
join PropertyDataTypes pdt
  on cv.PropertyDataTypeID = pdt.PropertyDataTypeID
 and pdt.Label = 'Type'
join parcelobjectvalues value
  on value.ParcelObjectID = ls.ParcelObjectID
 and value.valuetype = 'CalculatedAdjustedValue'
left join parcelobjectvalues overriddenvalue
  on overriddenvalue.ParcelObjectID = ls.ParcelObjectID
 and overriddenvalue.valuetype = 'OverriddenValue'
join parcelobjectvalues homk
  on homk.ParcelObjectID = ls.ParcelObjectID
 and homk.valuetype = 'CalculatedHomeownersValue'
where year(appraisaldate) = 2016
  and CharacteristicValue like '%F%'

select *
from
(
    select 
        *,
        case 
            when CharacteristicValue like '%A' then convert(int, substring(characteristicvalue, 1, len(characteristicvalue) - 1))
            when CharacteristicValue like '%K' then convert(int, substring(characteristicvalue, 1, len(characteristicvalue) - 1))
            when isnumeric(CharacteristicValue) = 1 then convert(int, characteristicvalue)
            else 0
        end Quantity
    from
    (
        select 
            parcelnumber, 
            sectionname,
            value.Value CalculatedValue,
            overriddenvalue.Value OverriddenValue,
            homk.Value CalculatedHomeownersValue,
            cv.CharacteristicValue,
            pdt.PropertyDataTypeID
        from
            landsections ls
            join lands l on ls.landid = ls.landid
            join parcels p on l.ParcelID = p.ParcelID
            join CharacteristicValueGroups cvg on cvg.ParcelObjectID = ls.ParcelObjectID
            join CharacteristicValues cv on cvg.CharacteristicValueGroupID = cv.CharacteristicValueGroupID
            join Characteristics c on cv.CharacteristicID = c.CharacteristicID
            join CharacteristicDetails cd on c.CharacteristicDetailsID = cd.CharacteristicDetailsID and cd.name = '09 Composite Adjustment'
            join PropertyDataTypes pdt on cv.PropertyDataTypeID = pdt.PropertyDataTypeID and pdt.Label = 'Type'
            join parcelobjectvalues value on value.ParcelObjectID = ls.ParcelObjectID and value.valuetype = 'CalculatedAdjustedValue'
            left join parcelobjectvalues overriddenvalue on overriddenvalue.ParcelObjectID = ls.ParcelObjectID and overriddenvalue.valuetype = 'OverriddenValue'
            join parcelobjectvalues homk on homk.ParcelObjectID = ls.ParcelObjectID and homk.valuetype = 'CalculatedHomeownersValue'
        where
            year(appraisaldate) = 2016
    ) Tbl
) Tbl2
where
    CharacteristicValue like '%A' and Quantity > 10

---

注意where子句中的最后一项。此查询返回0个结果，这将证明不存在包含F的

CharacteristicValue

列的行

现在让我们运行此查询：

select 
    parcelnumber, 
    sectionname,
    value.Value CalculatedValue,
    overriddenvalue.Value OverriddenValue,
    homk.Value CalculatedHomeownersValue,
    cv.CharacteristicValue,
    pdt.PropertyDataTypeID
from landsections ls
join lands l 
  on ls.landid = ls.landid
join parcels p
  on l.ParcelID = p.ParcelID
join CharacteristicValueGroups cvg
  on cvg.ParcelObjectID = ls.ParcelObjectID
join CharacteristicValues cv 
  on cvg.CharacteristicValueGroupID = cv.CharacteristicValueGroupID
join Characteristics c
  on cv.CharacteristicID = c.CharacteristicID
join CharacteristicDetails cd
  on c.CharacteristicDetailsID = cd.CharacteristicDetailsID
 and cd.name = '09 Composite Adjustment'
join PropertyDataTypes pdt
  on cv.PropertyDataTypeID = pdt.PropertyDataTypeID
 and pdt.Label = 'Type'
join parcelobjectvalues value
  on value.ParcelObjectID = ls.ParcelObjectID
 and value.valuetype = 'CalculatedAdjustedValue'
left join parcelobjectvalues overriddenvalue
  on overriddenvalue.ParcelObjectID = ls.ParcelObjectID
 and overriddenvalue.valuetype = 'OverriddenValue'
join parcelobjectvalues homk
  on homk.ParcelObjectID = ls.ParcelObjectID
 and homk.valuetype = 'CalculatedHomeownersValue'
where year(appraisaldate) = 2016
  and CharacteristicValue like '%F%'

select *
from
(
    select 
        *,
        case 
            when CharacteristicValue like '%A' then convert(int, substring(characteristicvalue, 1, len(characteristicvalue) - 1))
            when CharacteristicValue like '%K' then convert(int, substring(characteristicvalue, 1, len(characteristicvalue) - 1))
            when isnumeric(CharacteristicValue) = 1 then convert(int, characteristicvalue)
            else 0
        end Quantity
    from
    (
        select 
            parcelnumber, 
            sectionname,
            value.Value CalculatedValue,
            overriddenvalue.Value OverriddenValue,
            homk.Value CalculatedHomeownersValue,
            cv.CharacteristicValue,
            pdt.PropertyDataTypeID
        from
            landsections ls
            join lands l on ls.landid = ls.landid
            join parcels p on l.ParcelID = p.ParcelID
            join CharacteristicValueGroups cvg on cvg.ParcelObjectID = ls.ParcelObjectID
            join CharacteristicValues cv on cvg.CharacteristicValueGroupID = cv.CharacteristicValueGroupID
            join Characteristics c on cv.CharacteristicID = c.CharacteristicID
            join CharacteristicDetails cd on c.CharacteristicDetailsID = cd.CharacteristicDetailsID and cd.name = '09 Composite Adjustment'
            join PropertyDataTypes pdt on cv.PropertyDataTypeID = pdt.PropertyDataTypeID and pdt.Label = 'Type'
            join parcelobjectvalues value on value.ParcelObjectID = ls.ParcelObjectID and value.valuetype = 'CalculatedAdjustedValue'
            left join parcelobjectvalues overriddenvalue on overriddenvalue.ParcelObjectID = ls.ParcelObjectID and overriddenvalue.valuetype = 'OverriddenValue'
            join parcelobjectvalues homk on homk.ParcelObjectID = ls.ParcelObjectID and homk.valuetype = 'CalculatedHomeownersValue'
        where
            year(appraisaldate) = 2016
    ) Tbl
) Tbl2
where
    CharacteristicValue like '%A' and Quantity > 10

---

请注意，我已删除%F%检查。由于第一个查询证明

CharacteristicValue

列中没有带F的行，因此我假设外部查询可以依赖于此

然而，我明白了：

将varchar值“F”转换为数据类型int时，转换失败

---

我认为唯一可能发生这种情况的方法是忽略pdt.Label='Type'，但就我个人而言，我不明白为什么

答案似乎是，由于相关性，内部查询无法在外部查询之前执行。在外部查询中添加筛选器解决了此问题。

您的视图筛选器位于：

where year(appraisaldate) = 2016 and CharacteristicValue like '%F%'

这意味着可以有只满足

特征值的行，如“%F%”

T-SQL中未指定执行顺序。这是一个问题，当试图保护计算时，可能会抛出一些检查

通常，

case。。。当

是可靠的防护装置时。在你的情况下，我看没有理由不这样做。因此，当CharacteristicValue像“%A”而CharacteristicValue不像“%F%”时，可以执行

或者，您可以使用始终安全的

TRY\u CONVERT

---

其中

不是可靠的防护装置。在你的另一个回答中，你写道这是可行的，但这只是巧合。如果使用不同的查询计划，则可能会失败。计划因多种原因而更改。

如果将%F%筛选器添加到第二个查询中，是否会返回相同的错误？

CharacteristicValue

这是一个

INT

数据类型列？？？请尝试“从CharacteristicValue中选择*，其中CharacteristicValue类似“%F%”，然后查看是否返回任何内容。

isnumeric（）

是一个糟糕的函数选择，用于检查它是否真的是一个数值，因为对于像

123e12

或

456d1

这样的值，它将返回true，您应该使用像

这样的like运算符，其中CharacteristicValue不象“[^0-9]”

您有很多与多个表的联接，您说过表中有“%F%”的数据，pdt.Lavel和cd.Name会将它们过滤掉。我对此不确定，因为您是通过另一个表加入的。