如何提取SQL语句中的表名

Question

RT. 讨论一下思路
String getTableNames(String sqlStmt){
     .... //该如和设计解析逻辑
}

例如：
sqlStmt = "
INSERT INTO A
SELECT B.B1, B.B2, C.C1, C.C2
FROM B INNER JOIN C ON B.B0 = C.C0
WHERE B.B0 IN (SELECT D0 FROM D,E WHERE D.D1 = E.E1)
";

要求getTableNames(sqlStmt)的输出为 "A,B,C,D,E"

需要考虑到所有DML语句包括JOIN结构以及子查询的情况

Answer 1

如果不带别名的话, 可以就正则实现.
不过java的正则有些限制, 所以对于from中多于10张表的就有问题了, 不多可以自己在添加, 考虑不是很完善

[code="java"]
String str = "INSERT INTO A\n" +
"SELECT B.B1, B.B2, C.C1, C.C2\n" +
"FROM B INNER JOIN C ON B.B0 = C.C0\n" +
"WHERE B.B0 IN (SELECT D0 FROM D,E,G,H WHERE D.D1 = E.E1) \n";
Pattern p = Pattern
.compile(
"(?i)(?<=(?:from|into|update|join)\s{1,1000}"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?" // 重复这里, 可以多个from后面的表
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ "(?:\w{1,1000}(?:\s{0,1000},\s{0,1000})?)?"
+ ")(\w+)"
);
Matcher m = p.matcher(str);
List tables = new ArrayList();
while (m.find()) {
tables.add(m.group());
}
System.out.println(tables);
[/code]

不过像楼上说的, 最好还是自己写个sql解析器, 可以采用javacc或者antlr来写, 会容易些