組裝動態LINQ條件的利器－LINQKit

昨天提到實務上查詢條件多半由使用者動態決定，往往得靠SqlQuery或Dapper配合WHERE條件字串動態組裝搞定。（提醒：SQL語法可以用串接的，使用者輸入內容務必使用參數嵌入）但如此就得放棄LINQ .Where()的強型別優點。介紹一個好物，讓你可以兩者兼顧－LINQKit。

老樣子，用NuGet就可安裝，查詢關鍵字請填"LinqKit"：

最新版的LINQKit需要EntityFramework 6.1.3以上，但依其運作原理，EF5應該也適用，等遇到再說。

LINQKit在Github有詳細說明，這裡只簡單介紹如何用它取代自組SQL WHERE指令。LINQKit靠PredicateBuilder建立動態條件，起手式有兩種：

var pred = PredicateBuilder.False<MyEntity>();
pred = pred.Or(o => o.Col1 == 1);
pred = pred.Or(o => o.Col2 == "X");
ctx.MyEntity.AsExpandable().Where(pred);
等同於ctx.MyEntity.Where(o => o.Col1 == 1 || o.Clo2 == "X")

var pred = PredicateBuilder.True<MyEntity>();
pred = pred.And(o => o.Col1 == 1);
pred = pred.And(o => o.Col2 == "X");
ctx.MyEntity.AsExpandable().Where(pred);
等同於ctx.MyEntity.Where(o => o.Col1 == 1 && o.Clo2 == "X")

注意，OR比對一開始用False<T>，AND則是True<T>，而EF的DbSet<T>要先.AsExpandable()才能在Where()傳入Predicate，否則會出現 The LINQ expression node type 'Invoke' is not supported in LINQ to Entities. 錯誤。另一種做法是不加.AsExpandable()，而是.Where(pred.Expand())，二者皆可。

如果WHERE條件又有AND又有OR怎麼辦？昨天的案例剛好拿來當範例：WHERE PeriodId='2021A' AND (CostCenter='100' AND JobType='A' OR CostCenter='101' AND JobType='S')。秘訣在於Predicate的And()或Or()的傳入參數也可接受其他Predicate，故前述需求可寫成：
var pred2 = PredicateBuilder.False<MyEntity>();
pred2 = pred2.Or(o => o.CostCenter == "100" && o.JobType == "A");
pred2 = pred2.Or(o => o.CostCenter == "101" && o.JobType == "S");
var pred = PredicateBuilder.True<MyEntity>();
pred = pred.And(o => o.PeriodId == "2021A");
pred = pred.And(pred2);
ctx.DataGroup.AsExpandable().Where(pred);

很簡單吧？最後來看看昨天的案例如何用LINQKit改寫：

staticstring[] keys = newstring[] { "100-A", "101-S" };

staticvoid Test3()

{

using (var ctx = new XDBEntities())

    {

        var pOr = PredicateBuilder.False<DataGroup>();

        keys.ForEach(o =>

        {

            var p = o.Split('-');

            var costCenter = p[0];

            var jobType = p[1];

            pOr = pOr.Or(q => q.CostCenter == costCenter && q.JobType == jobType);

        });

        var pWhere = PredicateBuilder.True<DataGroup>();

        pWhere = pWhere.And(o => o.PeriodId == "2021A");

        pWhere = pWhere.And(pOr);

        ctx.Database.Log = (m) =>

        {

            Console.WriteLine("**** EF Log ****");

            Console.WriteLine(m);

        };

        var dataGrps = ctx.DataGroup.AsExpandable().Where(pWhere).ToList();

//var dataGrps = ctx.DataGroup.Where(pWhere.Expand()).ToList();

foreach (var dg in dataGrps)

        {

            Console.WriteLine("DataGroup:{0}-{1}", dg.CostCenter, dg.JobType);

        }

    }

}

順便示範如何觀察EF6產生的SQL指令，ctx.Database有個Action<string> Log屬性，覆寫成自訂函式，就能接收到EF6的Log，其中包含SQL指令，測試結果如下：

**** EF Log ****
Opened connection at 2015/10/23 上午 05:33:11 +08:00

**** EF Log ****
SELECT
    [Extent1].[PeriodId] AS [PeriodId],
    [Extent1].[CostCenter] AS [CostCenter],
    [Extent1].[JobType] AS [JobType],
    [Extent1].[Status] AS [Status],
    [Extent1].[Tag] AS [Tag]
    FROM [dbo].[DataGroup] AS [Extent1]
    WHERE ('2021A' = [Extent1].[PeriodId]) AND ((([Extent1].[CostCenter] = @p__l
inq__0) AND ([Extent1].[JobType] = @p__linq__1)) OR (([Extent1].[CostCenter] = @
p__linq__2) AND ([Extent1].[JobType] = @p__linq__3)))
**** EF Log ****


**** EF Log ****
-- p__linq__0: '100' (Type = AnsiString, Size = 8000)

**** EF Log ****
-- p__linq__1: 'A' (Type = AnsiString, Size = 8000)

**** EF Log ****
-- p__linq__2: '101' (Type = AnsiString, Size = 8000)

**** EF Log ****
-- p__linq__3: 'S' (Type = AnsiString, Size = 8000)

**** EF Log ****
-- Executing at 2015/10/23 上午 05:33:11 +08:00

**** EF Log ****
-- Completed in 11 ms with result: SqlDataReader

**** EF Log ****


**** EF Log ****
Closed connection at 2015/10/23 上午 05:33:11 +08:00

DataGroup:100-A
DataGroup:101-S

Predicate產生的WHERE條件符合我們的期望：('2021A' = [Extent1].[PeriodId]) AND ((([Extent1].[CostCenter] = @p__l inq__0) AND ([Extent1].[JobType] = @p__linq__1)) OR (([Extent1].[CostCenter] = @p__linq__2) AND ([Extent1].[JobType] = @p__linq__3)))

與自組SQL指令相比，使用LINQKit Predicate寫法相對簡潔，省去為嵌入參數命名、管理參數值陣列的瑣碎手工，又可享受LINQ強型別優勢，不用擔心「打錯字」（呃，被刺傷了），決定將LINQKit收入常備工具箱，多加利用。