EF Core 迁移从新的复合索引中删除第一个外键索引

EF Core migrations drops first foreign key index from a new composite index

这是我的模型:

public class Period
{
    [Key]
    public int Id { get; set; }
}

public class Company
{
    [Key]
    public int Id { get; set; }
}

public class Invoice
{
    [Key]
    public int Id { get; set; }

    public int PeriodId { get; set; }
    public Period Period { get; set; }

    public int CompanyId { get; set; }
    public Company Company { get; set; }
}

没什么特别的。一张包含时间和公司的简单发票。

这是它们的配置方式:

modelBuilder.Entity<Invoice>()
    .HasOne(p => p.Period)
    .WithMany()
    .OnDelete(DeleteBehavior.Restrict)
;

modelBuilder.Entity<Invoice>()
    .HasOne(p => p.Company)
    .WithMany()
    .OnDelete(DeleteBehavior.Restrict)
;

这是由此生成的迁移文件:

migrationBuilder.CreateTable(
    name: "Companies",
    columns: table => new
    {
        Id = table.Column<int>(nullable: false)
            .Annotation("SqlServer:Identity", "1, 1")
    },
    constraints: table =>
    {
        table.PrimaryKey("PK_Companies", x => x.Id);
    });

migrationBuilder.CreateTable(
    name: "Periods",
    columns: table => new
    {
        Id = table.Column<int>(nullable: false)
            .Annotation("SqlServer:Identity", "1, 1")
    },
    constraints: table =>
    {
        table.PrimaryKey("PK_Periods", x => x.Id);
    });

migrationBuilder.CreateTable(
    name: "Invoices",
    columns: table => new
    {
        Id = table.Column<int>(nullable: false)
            .Annotation("SqlServer:Identity", "1, 1"),
        PeriodId = table.Column<int>(nullable: false),
        CompanyId = table.Column<int>(nullable: false)
    },
    constraints: table =>
    {
        table.PrimaryKey("PK_Invoices", x => x.Id);
        table.ForeignKey(
            name: "FK_Invoices_Companies_CompanyId",
            column: x => x.CompanyId,
            principalTable: "Companies",
            principalColumn: "Id",
            onDelete: ReferentialAction.Restrict);
        table.ForeignKey(
            name: "FK_Invoices_Periods_PeriodId",
            column: x => x.PeriodId,
            principalTable: "Periods",
            principalColumn: "Id",
            onDelete: ReferentialAction.Restrict);
    });

migrationBuilder.CreateIndex(
    name: "IX_Invoices_CompanyId",
    table: "Invoices",
    column: "CompanyId");

migrationBuilder.CreateIndex(
    name: "IX_Invoices_PeriodId",
    table: "Invoices",
    column: "PeriodId");

一切似乎都是合法的。现在我需要创建一个唯一索引来控制一段时间内公司发票的唯一性:

modelBuilder.Entity<Invoice>()
    .HasIndex(p => new
    {
        p.PeriodId,
        p.CompanyId,
    })
    .IsUnique()
;

这就是迁移现在的样子:

migrationBuilder.DropIndex(
    name: "IX_Invoices_PeriodId",
    table: "Invoices");

migrationBuilder.CreateIndex(
    name: "IX_Invoices_PeriodId_CompanyId",
    table: "Invoices",
    columns: new[] { "PeriodId", "CompanyId" },
    unique: true);

注意“DropIndex”。它从我的复合索引中获取第一列。如果我改变顺序,它仍然会掉落第一个:

modelBuilder.Entity<Invoice>()
    .HasIndex(p => new
    {
        p.CompanyId,
        p.PeriodId,
    })
    .IsUnique()
;

migrationBuilder.DropIndex(
    name: "IX_Invoices_CompanyId",
    table: "Invoices");

migrationBuilder.CreateIndex(
    name: "IX_Invoices_CompanyId_PeriodId",
    table: "Invoices",
    columns: new[] { "CompanyId", "PeriodId" },
    unique: true);

我不确定我是否理解这背后的逻辑。复合索引中有什么特殊之处使单独的索引无用吗?或者它可能是某种错误?

我发现在 SO 上唯一提到这个问题有点过时 post: EF Migrations drops index when adding compsite index

P.S。使用 SQL 服务器作为数据库提供程序在 3.1.8 上测试。

应该更深入地研究这个问题。 创建复合索引时,它的第一列可以在查询中单独使用。 https://github.com/dotnet/efcore/issues/22513