InsertOrUpdate.UpdateColumns 生成的sql 只保留 主键+更新列 ，无关列不要生成

[zhujiancc]

运行环境

1. postgres
2. net 6
3. FreeSql.Provider.Postgresql:3.2.813

演示代码

[Table(Name = "test_artwork_info")]
public class TestArtworkInfo
{
    [Column(Name = "datecode", IsPrimary = true)] public int DateCode { get; set; }
    [Column(Name = "platform", IsPrimary = true)] public int Platform { get; set; }
    [Column(Name = "artwork_id", IsPrimary = true)] public string ArtworkId { get; set; }
    [Column(Name = "monthcode", IsPrimary = true)] public int MonthCode { get; set; }

    [Column(Name = "user_id")] public string UserId { get; set; }

    [Column(Name = "tags")] public string[] Tags { get; set; }

    [Column(Name = "auth", DbType = "jsonb")] public string Auth { get; set; } // 如果需要更复杂类型可以用 JObject 或自定义 class

    [Column(Name = "title")] public string Title { get; set; }

    [Column(Name = "addtime")] public DateTime AddTime { get; set; }
}

var model = new TestArtworkInfo()
        {
            DateCode = 20251008,
            Platform = 8,
            ArtworkId = "ac46b2ef0f4244c08bbf9e2769d74ac0",
            MonthCode = 202510,
            UserId = string.Empty
        };

        model.Tags = new[] { "48d7f37033" };
        model.Title = "test";


        //1. 定义一个完整的model , 差量更新 ，全量更新
        await context.MergeIntoAsync(
            source: new List<TestArtworkInfo> { model },
            monthCodeSelector: s => s.MonthCode, //分区键
            updateColumns: s => new { s.Tags, s.Title });

public static async Task MergeIntoAsync<T1>(
        this IFreeSql<HoloFlag> context,
        IEnumerable<T1> source,
        Expression<Func<T1, int>> monthCodeSelector,
        Expression<Func<T1, object>> updateColumns = null) where T1 : class
    {
        if (source == null || !source.Any())
        {
            return;
        }

        var tableName = GetTableName<T1>();

        var keySelector = monthCodeSelector.Compile();

        var groups = source.GroupBy(keySelector);

        foreach (var group in groups)
        {
            try
            {
                var insert = context
                    .InsertOrUpdate<T1>()
                    .AsTable($"{tableName}_{group.Key}")
                    .SetSource(group);

                if (updateColumns != null)
                    insert.UpdateColumns(updateColumns);
                
                await insert.ExecuteAffrowsAsync();
            }
            catch (Exception e)
            {
                if (!e.Message.Contains("#Temp_"))
                {
                    throw new Exception("无法删除Temp ，请检查class和表是否一致");
                }
            }
        }
    }

生成的sql

INSERT INTO "yunlue_bussiness"."test_artwork_info_202510"("datecode", "platform", "artwork_id", "monthcode", "user_id", "tags", "auth", "title", "addtime") VALUES(20251008, 8, 'ac46b2ef0f4244c08bbf9e2769d74ac0', 202510, '', ARRAY['48d7f37033']::varchar[], NULL, 'test', '0001-01-01 00:00:00.000000')
ON CONFLICT("datecode", "platform", "artwork_id", "monthcode") DO UPDATE SET
"tags" = EXCLUDED."tags", 
"title" = EXCLUDED."title"

这里的sql能不能优化下 ，自动去掉user_id，auth，addtime， 只留下 主键和更新字段

INSERT INTO "yunlue_bussiness"."test_artwork_info_202510"("datecode", "platform", "artwork_id", "monthcode","tags", "title") VALUES(20251008, 8, 'ac46b2ef0f4244c08bbf9e2769d74ac0', 202510, ARRAY['48d7f37033']::varchar[], 'test')
ON CONFLICT("datecode", "platform", "artwork_id", "monthcode") DO UPDATE SET
"tags" = EXCLUDED."tags", 
"title" = EXCLUDED."title"

**大宽表能减少大量的生成sql **