Update
My recomendation would be to shred the XML into relations and do searches and joins on the resulted relation, in a set oriented fashion, rather than the procedural fashion of searching specific nodes in the XML. Here is a simple XML query that shreds out the nodes and attributes of interest:
select x.value(N'../../../../@stepId', N'int') as StepID
, x.value(N'../../@id', N'int') as ComponentID
, x.value(N'@nom',N'nvarchar(100)') as Nom
, x.value(N'@valeur', N'nvarchar(100)') as Valeur
from @x.nodes(N'/xml/box/components/component/variables/variable') t(x)
However, if you must use an XPath that retrieves exactly the value of interest:
select x.value(N'@valeur', N'nvarchar(100)') as Valeur
from @x.nodes(N'/xml/box[@stepId=sql:variable("@stepID")]/
components/component[@id = sql:variable("@componentID")]/
variables/variable[@nom="Enabled"]') t(x)
If the stepID and component ID are columns, not variables, the you should use sql:column() instead of sql:variable in the XPath filters. See Binding Relational Data Inside XML Data.
And finaly if all you need is to check for existance you can use the exist() XML method:
select @x.exist(
N'/xml/box[@stepId=sql:variable("@stepID")]/
components/component[@id = sql:variable("@componentID")]/
variables/variable[@nom="Enabled" and @valeur="Yes"]')
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…