实现的功能为:
1:从csv文件中的第一行提取列名生成数据库表
2:将原始的csv文件转换成sqlloder能识别的csv(增加列),根据时间列生成一个整形主键列
3:生成ctl控制文件
4:启动Sqlloder并导入
这里做记录主要是导入时出现了莫名其妙的问题
第一行第一列总是会报错误
第一列为数字列,则SQLLoader报 ORA-01722
第一列调整成时间列并将数字列放到最后,则SQLLoader报时间列转换错误。
参考各种文章不能解决,参考的解决方案包括
1:在ctl文件中指定字段类型
https://blog.csdn.net/qq_25798961/article/details/88534104
CHAR 字符型
INTEGER EXTERNAL 整型
DECIMAL EXTERNAL 浮点型
DATE 'YYYYMMDD' 日期型
数字列增加 选项 INTEGER EXTERNAL 没有效果
2:数据列做转换
https://community.oracle.com/tech/developers/discussion/3684710/ora-01722-invalid-number-in-sql-loader
Options(errors=10000,skip=3)
LOAD DATA
INFILE '/home/appltest/Sheet1.csv'
BADFILE 'sample.bad'
DISCARDFILE 'sample.dsc'
REPLACE
INTO TABLE sampl_table
fields terminated by ','
optionally enclosed by '"'
trailing nullcols
(
customer_name CHAR NULLIF customer_name = BLANKS "RTRIM(:customer_name)",
customer_no CHAR NULLIF customer_no = BLANKS "RTRIM(:customer_no)",
item_no CHAR NULLIF item_no = BLANKS "RTRIM(:item_no )",
field1 FILLER,
field2 FILLER,
valid_amount_month1 "to_number(:valid_amount_month1 , '999,999,999.99')" ,
valid_units_month1 "to_number(substr(:valid_units_month1,instr(:valid_units_month1,'$')+1),'999,999,999.99')",
field3 FILLER,
field4 FILLER,
valid_budget_units_month1 "to_number(:valid_budget_units_month1, '999,999,999.99')" ,
valid_budget_amount_month1 "to_number(trim(substr(:valid_budget_amount_month1,instr(:valid_budget_amount_month1,'$')+1)),'999,999,999.99')" ,
valid_rate CHAR NULLIF valid_rate = BLANKS "RTRIM(:valid_rate)"
)
3:ctl文件增加 "str '\r\n'" 项没有效果
3:ctl文件第一列加上"replace(replace(:CHANNELKEY, CHR(13), ''), CHR(10), '')" 配置项,没有效果
最终通过以下方案解决
1:设置生成的csv中包含一行头信息
2:ctl中包含一个OPTIONS(skip=1)配置,跳过头信息行的导入
究其根本仍不知道是 文件编码 还是 windows下换行 等等原因导致的。
核心代码如下:
public void Process()
{
ClearFiles();
ReBuildDataFile(); //解析-1:将输入文件转换成Sqlloder可识别的文件,拼接时间,生成主键
CreateDataTable(); //解析-2:生成表结构
BuildCtlFile(); //解析-3:生成Sqlloder控制文件
StartSqlloader(); //解析-4:启动Sqlloder,执行导入
}
//生成sqloader可识别的控制文件
private string BuildCtlFile()
{
string strError = string.Empty;
string strColumns = AttachKeyColumn();
strColumns = strColumns.Replace("DATETIMES", "DATETIMES \"to_timestamp(:DATETIMES,'yyyy-mm-dd hh24:mi:ss.ff')\"");
StringBuilder sb = new StringBuilder();
sb.AppendLine("OPTIONS(skip=1)");
sb.AppendLine("load data");
sb.AppendLine(string.Format("infile '{0}' \"str '\\r\\n'\"", GetCsvFileFullName()));
sb.AppendLine(string.Format("append into table {0}", GetTableName()));
sb.AppendLine(string.Format("fields terminated by '{0}'", this.splitChar.ToString()));
sb.AppendLine("optionally enclosed by '\"'");
sb.AppendLine("trailing nullcols");
sb.AppendLine(string.Format("({0})", strColumns));
GenerateFile(GetCtlFileFullName(), sb.ToString());
return strError;
}
private string StartSqlloader()
{
string strError = string.Empty;
string output = string.Empty;
bool processOk = true;
using (System.Diagnostics.Process p = new System.Diagnostics.Process())
{
p.StartInfo.FileName = this.SqlLoaderConfig;
p.StartInfo.Arguments = string.Format(" {0} control={1} log={2}", this.OracleDBConfig, GetCtlFileFullName(), GetCtlFileFullName().Replace("ctl", "log"));
p.StartInfo.CreateNoWindow = true; //不再新窗口启动
p.StartInfo.UseShellExecute = false; //不使用操作系统的shell启动
p.StartInfo.RedirectStandardOutput = true; //输出重定向
p.StartInfo.RedirectStandardError = true;
p.StartInfo.RedirectStandardInput = true;
//p.Exited += P_Exited;
p.Start();
output = p.StandardOutput.ReadToEnd();
//ExitCode 11g 12c和9i不同
if (p.ExitCode == 1) //Ex_Fail
{
processOk = false;
}
if (p.ExitCode == 2) //Ex_Warn
{
processOk = false;
}
if (p.ExitCode == 4) //Ex_FTL
{
processOk = false;
}
}
if (!processOk)
{
throw new IOException(output);
}
return strError;
}
原始CSV文件为:
DATETIMES,CD_000_0,CD_000_1,CD_000_2
2022-09-19 16:11:17.720,100,200,300
2022-09-19 16:11:17.721,101,201,301
2022-09-19 16:11:17.722,102,202,302
2022-09-19 16:11:17.723,103,203,303
2022-09-19 16:11:17.724,104,204,304
2022-09-19 16:11:17.725,105,205,305
2022-09-19 16:11:17.726,106,206,306
2022-09-19 16:11:17.727,107,207,307
2022-09-19 16:11:17.728,108,208,308
2022-09-19 16:11:17.729,109,209,309
2022-09-19 16:11:17.730,110,210,310
2022-09-19 16:11:17.731,111,211,311
2022-09-19 16:11:17.732,112,212,312
2022-09-19 16:11:17.733,113,213,313
2022-09-19 16:11:17.734,114,214,314
2022-09-19 16:11:17.735,115,215,315
2022-09-19 16:11:17.736,116,216,316
2022-09-19 16:11:17.737,117,217,317
2022-09-19 16:11:17.738,118,218,318
2022-09-19 16:11:17.739,119,219,319
2022-09-19 16:11:17.740,120,220,320
转换成的CSV文件为:
CHANNELKEY INTEGER EXTERNAL,DATETIMES,CD_000_0,CD_000_1,CD_000_2
4295477720,2022-09-19 17:11:17.720,100,200,300
4295477721,2022-09-19 17:11:17.721,101,201,301
4295477722,2022-09-19 17:11:17.722,102,202,302
4295477723,2022-09-19 17:11:17.723,103,203,303
4295477724,2022-09-19 17:11:17.724,104,204,304
4295477725,2022-09-19 17:11:17.725,105,205,305
4295477726,2022-09-19 17:11:17.726,106,206,306
4295477727,2022-09-19 17:11:17.727,107,207,307
4295477728,2022-09-19 17:11:17.728,108,208,308
4295477729,2022-09-19 17:11:17.729,109,209,309
4295477730,2022-09-19 17:11:17.730,110,210,310
4295477731,2022-09-19 17:11:17.731,111,211,311
4295477732,2022-09-19 17:11:17.732,112,212,312
4295477733,2022-09-19 17:11:17.733,113,213,313
4295477734,2022-09-19 17:11:17.734,114,214,314
4295477735,2022-09-19 17:11:17.735,115,215,315
4295477736,2022-09-19 17:11:17.736,116,216,316
4295477737,2022-09-19 17:11:17.737,117,217,317
4295477738,2022-09-19 17:11:17.738,118,218,318
4295477739,2022-09-19 17:11:17.739,119,219,319
4295477740,2022-09-19 17:11:17.740,120,220,320
生成的ctl文件为:
OPTIONS(skip=1)
load data
infile 'D:\******\TemplateData\1cf37b73-6727-4618-b636-73a9292ef33d.csv' "str '\r\n'"
append into table FZ_41657_01
fields terminated by ','
optionally enclosed by '"'
trailing nullcols
(CHANNELKEY INTEGER EXTERNAL,DATETIMES "to_timestamp(:DATETIMES,'yyyy-mm-dd hh24:mi:ss.ff')",CD_000_0,CD_000_1,CD_000_2)
附ctl文件说明:
https://blog.csdn.net/demonson/article/details/79712207
OPTIONS (skip=1,rows=128) -- sqlldr 命令显示的 选项可以写到这里边来,skip=1 用来跳过数据中的第一行
LOAD DATA
INFILE "users_data.csv" --指定外部数据文件,可以写多 个 INFILE "another_data_file.csv" 指定多个数据文件
--这里还可以使 用 BADFILE、DISCARDFILE 来指定坏数据和丢弃数据的文件,
truncate --操作类型,用 truncate table 来清除表中原有 记录
INTO TABLE users -- 要插入记录的表
Fields terminated by "," -- 数据中每行记录用 "," 分隔
Optionally enclosed by '"' -- 数据中每个字段用 '"' 框起,比如字段中有 "," 分隔符时
trailing nullcols --表的字段没有对应的值时允 许为空
(
virtual_column FILLER, --这是一个虚拟字段,用来跳 过由 PL/SQL Developer 生成的第一列序号
user_id number, --字段可以指定类型,否则认 为是 CHARACTER 类型, log 文件中有显示
user_name,
login_times,
last_login DATE "YYYY-MM-DD HH24:MI:SS" -- 指定接受日期的格式,相当用 to_date() 函数转换
)
主键列 CHANNELKEY,生成规则为:
private string FixKeyValue(DateTime time)
{
TimeSpan toNow = time.Subtract((new DateTime(time.Year, time.Month - 1, 1)));
//return (toNow.Ticks / 10000).ToString().Substring(3);
return (toNow.Ticks / 10000).ToString();
}
网友评论