java:Druid工具类解析sql获取表名
1 前言
alibaba的druid连接池除了sql执行的功能外,还有sql语法解析的工具提供,参考依赖如下:
<dependency><groupId>com.alibaba</groupId><artifactId>druid</artifactId><version>1.2.15</version>
</dependency>
2 使用
参考druid的工具类:com.alibaba.druid.sql.parser.SQLParserUtils#getTables(String sql, DbType dbType)方法,可以用于获取sql的表名:
比如针对mysql的select语句:
package com.xiaoxu.parser;import com.alibaba.druid.DbType;
import com.alibaba.druid.sql.parser.SQLParserUtils;import java.util.List;/*** @author xiaoxu* @date 2024-03-11* java_demo2:com.xiaoxu.parser.SQLParserTest*/
public class SQLParserTest {public static void main(String[] args) {List<String> tables = SQLParserUtils.getTables("select * from my_fuitrs_99 where id = ?", DbType.mysql);System.out.println(tables);String nihao_99 = "select * from my_fuitrs_99 where id = ?".replace(tables.get(0), "my_fuitrs");System.out.println(nihao_99);}}
执行结果如下:
该方法的本质是根据自定义的hash值,对于每个标识符,能唯一生成一个hashCode,将标识符与存入map中的标识符来比对。比如当lexer.token获取的是com.alibaba.druid.sql.parser.Token类中的FROM(“FROM”)时,先执行lexer.nextToken()方法调用,获取到用户自定义的from后续的标识符(from关键字后续跟着的空格等字符会跳过),判断在没有获取到,则将其置为:Token.IDENTIFIER,表示用户自定义的identifier标识符。如下 keywords.getKeyword(hashLCase) 如果获取的标识符不存在,则将判断为用户自定义的标识符。
参考com.alibaba.druid.sql.parser.Lexer的scanIdentifier()方法:
public void scanIdentifier() {this.hashLCase = 0;this.hash = 0;final char first = ch;if (ch == '`') {mark = pos;bufPos = 1;char ch;int startPos = pos + 1;int quoteIndex = text.indexOf('`', startPos);if (quoteIndex == -1) {throw new ParserException("illegal identifier. " + info());}hashLCase = 0xcbf29ce484222325L;hash = 0xcbf29ce484222325L;for (int i = startPos; i < quoteIndex; ++i) {ch = text.charAt(i);hashLCase ^= ((ch >= 'A' && ch <= 'Z') ? (ch + 32) : ch);hashLCase *= 0x100000001b3L;hash ^= ch;hash *= 0x100000001b3L;}stringVal = MySqlLexer.quoteTable.addSymbol(text, pos, quoteIndex + 1 - pos, hash);//stringVal = text.substring(mark, pos);pos = quoteIndex + 1;this.ch = charAt(pos);token = Token.IDENTIFIER;return;}final boolean firstFlag = isFirstIdentifierChar(first);if (!firstFlag) {throw new ParserException("illegal identifier. " + info());}hashLCase = 0xcbf29ce484222325L;hash = 0xcbf29ce484222325L;hashLCase ^= ((ch >= 'A' && ch <= 'Z') ? (ch + 32) : ch);hashLCase *= 0x100000001b3L;hash ^= ch;hash *= 0x100000001b3L;mark = pos;bufPos = 1;char ch = 0;for (; ; ) {char c0 = ch;ch = charAt(++pos);if (!isIdentifierChar(ch)) {if ((ch == '(' || ch == ')') && c0 > 256) {bufPos++;continue;}break;}hashLCase ^= ((ch >= 'A' && ch <= 'Z') ? (ch + 32) : ch);hashLCase *= 0x100000001b3L;hash ^= ch;hash *= 0x100000001b3L;bufPos++;continue;}this.ch = charAt(pos);if (bufPos == 1) {switch (first) {case '(':token = Token.LPAREN;return;case ')':token = Token.RPAREN;return;default:break;}token = Token.IDENTIFIER;stringVal = CharTypes.valueOf(first);if (stringVal == null) {stringVal = Character.toString(first);}return;}Token tok = keywords.getKeyword(hashLCase);if (tok != null) {token = tok;if (token == Token.IDENTIFIER) {stringVal = SymbolTable.global.addSymbol(text, mark, bufPos, hash);} else {stringVal = null;}} else {token = Token.IDENTIFIER;stringVal = SymbolTable.global.addSymbol(text, mark, bufPos, hash);}
}
获取表名的工具类方法:
public static List<String> getTables(String sql, DbType dbType) {Set<String> tables = new LinkedHashSet<>();boolean set = false;Lexer lexer = createLexer(sql, dbType);lexer.nextToken();SQLExprParser exprParser;switch (dbType) {case odps:exprParser = new OdpsExprParser(lexer);break;case mysql:exprParser = new MySqlExprParser(lexer);break;default:exprParser = new SQLExprParser(lexer);break;}for_:for (; lexer.token != Token.EOF; ) {switch (lexer.token) {case CREATE:case DROP:case ALTER:set = false;lexer.nextToken();if (lexer.token == Token.TABLE) {lexer.nextToken();if (lexer.token == Token.IF) {lexer.nextToken();if (lexer.token == Token.NOT) {lexer.nextToken();}if (lexer.token == Token.EXISTS) {lexer.nextToken();}}SQLName name = exprParser.name();tables.add(name.toString());if (lexer.token == Token.AS) {lexer.nextToken();}}continue for_;case FROM:case JOIN:lexer.nextToken();if (lexer.token != Token.LPAREN&& lexer.token != Token.VALUES) {SQLName name = exprParser.name();tables.add(name.toString());}continue for_;case SEMI:set = false;break;case SET:set = true;break;case EQ:if (set && dbType == DbType.odps) {lexer.nextTokenForSet();continue for_;}break;default:break;}lexer.nextToken();}return new ArrayList<>(tables);
}
比如上述的from关键字,当执行完lexer.nextToken()方法后,lexer.stringVal()方法即可以通过字符串的头尾下标切割字符串并返回该标识符,比如表名,也就是我们自定义的标识符。同时关注源码逻辑可知,数据库的表名,druid工具类会将大写字符、小写字符(大写字母A-Z的ASCII码值范围是65-90,而小写字母的ASCII码值范围是97-122,在大写字母的ASCII码值上+32即可转换成小写字母)判定为标识符,同时druid在处理时,除了大小写字母外,下划线(_)、美元符号($)、数字(0-9,ASCII码值范围是48-57)等等,均可判定为标识符。
修改表名含有大写字母,如下获取sql的表名:
List<String> tables = SQLParserUtils.getTables("select * from my_fuitrs_99 where id = ?", DbType.mysql);
System.out.println(tables);
String nihao_99 = "select * from my_fuitrs_99 where id = ?".replace(tables.get(0), "my_fuitrs");
System.out.println(nihao_99);
List<String> tables2 = SQLParserUtils.getTables("select * from MY_fuitrs_99 where id = ?", DbType.mysql);
System.out.println(tables2);
重新执行执行结果如下:
[my_fuitrs_99]
select * from my_fuitrs where id = ?
[MY_fuitrs_99]
3 举一反三
那么我们可以根据上面的工具,简单自定义实现一个替换sql表名的工具,工具类如下:
SQLParserUtil:
package com.xiaoxu.parser;import com.alibaba.druid.DbType;
import com.alibaba.druid.sql.ast.SQLName;
import com.alibaba.druid.sql.dialect.mysql.parser.MySqlExprParser;
import com.alibaba.druid.sql.dialect.mysql.parser.MySqlLexer;
import com.alibaba.druid.sql.parser.Lexer;
import com.alibaba.druid.sql.parser.SQLExprParser;
import com.alibaba.druid.sql.parser.SQLParserFeature;
import com.alibaba.druid.sql.parser.Token;
import org.springframework.lang.Nullable;
import org.springframework.util.StringUtils;/*** @author xiaoxu* @date 2024-03-12* java_demo2:com.xiaoxu.parser.SQLParserUtil*/
@SuppressWarnings("all")
public class SQLParserUtil {public static String getSqlFromReplaceNameIfNeccessary(String sql, DbType dbType, @Nullable String replaceTableName) {String table = null;String newSql = sql;Lexer lexer = createLexer(sql, dbType);lexer.nextToken();SQLExprParser exprParser;switch (dbType) {case mysql:exprParser = new MySqlExprParser(lexer);break;default:exprParser = new SQLExprParser(lexer);break;}for_:for (; lexer.token() != Token.EOF; ) {switch (lexer.token()) {case FROM:case INTO:lexer.nextToken();if (lexer.token() != Token.LPAREN && lexer.token() != Token.VALUES) {if (StringUtils.hasText(replaceTableName)) {int mark = ((MLexer) lexer).getMark();int bufPoss = ((MLexer) lexer).getBufPos();StringBuilder sbd = new StringBuilder();sbd.append(sql.substring(0, mark));sbd.append(replaceTableName);sbd.append(sql.substring(mark + bufPoss));newSql = sbd.toString();}SQLName name = exprParser.name();table = name.toString();}break for_;default:break;}lexer.nextToken();}System.out.println("原本的表名:" + table);System.out.println("替换表名为:" + replaceTableName + "后的sql:" + newSql);return newSql;}public static Lexer createLexer(String sql, DbType dbType) {return createLexer(sql, dbType, new SQLParserFeature[0]);}public static Lexer createLexer(String sql, DbType dbType, SQLParserFeature... features) {if (dbType == null) {dbType = DbType.mysql;}switch (dbType) {case mysql:return new MLexer(sql);default:return new Lexer(sql, null, dbType);}}private static class MLexer extends MySqlLexer {public MLexer(char[] input, int inputLength, boolean skipComment) {super(input, inputLength, skipComment);}public MLexer(String input) {super(input);}public MLexer(String input, SQLParserFeature... features) {super(input, features);}public MLexer(String input, boolean skipComment, boolean keepComments) {super(input, skipComment, keepComments);}public int getBufPos() {return this.bufPos;}public int getMark() {return this.mark;}}}
测试下我们自定义的SQLParserUtil工具类:
package com.xiaoxu.parser;import com.alibaba.druid.DbType;/*** @author xiaoxu* @date 2024-03-12* java_demo2:com.xiaoxu.parser.SQLParserTest2*/
public class SQLParserTest2 {public static void main(String[] args) {String sql = SQLParserUtil.getSqlFromReplaceNameIfNeccessary("select * from my_fruits_99 where id = ?",DbType.mysql, "xiaoxu_88");System.out.println(sql);System.out.println("\n");String sql2 = SQLParserUtil.getSqlFromReplaceNameIfNeccessary("insert into apple_$66 values()",DbType.mysql, "Pear$_88");System.out.println(sql2);}}
执行结果如下:
可以看到,上面工具针对扫描到标识符为FROM(比如select * from语句)或者INTO(比如insert into语句)时,可以实现替换sql的表名功能,其余类似功能参考druid的工具类自行实现即可。
再来举个栗子,新增方法getSqlInHoldCountIfNeccessary如下:
package com.xiaoxu.parser;import com.alibaba.druid.DbType;
import com.alibaba.druid.sql.ast.SQLName;
import com.alibaba.druid.sql.dialect.mysql.parser.MySqlExprParser;
import com.alibaba.druid.sql.dialect.mysql.parser.MySqlLexer;
import com.alibaba.druid.sql.parser.Lexer;
import com.alibaba.druid.sql.parser.SQLExprParser;
import com.alibaba.druid.sql.parser.SQLParserFeature;
import com.alibaba.druid.sql.parser.Token;
import com.google.common.collect.Lists;
import org.springframework.lang.Nullable;
import org.springframework.util.StringUtils;import java.util.List;/*** @author xiaoxu* @date 2024-03-12* java_demo2:com.xiaoxu.parser.SQLParserUtil*/
@SuppressWarnings("all")
public class SQLParserUtil {public static String getSqlInHoldCountIfNeccessary(String sql, DbType dbType, @Nullable Integer[] count) {StringBuilder tempSql = new StringBuilder();Lexer lexer = createLexer(sql, dbType);lexer.nextToken();int posInc = 0;int startPos = 0;int endPos = 0;int subStartPos = 0;for_:for (; lexer.token() != Token.EOF; ) {switch (lexer.token()) {case IN:lexer.nextToken();startPos = endPos;endPos = ((MLexer) lexer).getPos();if (lexer.token() == Token.LPAREN) {tempSql.append(sql, startPos, endPos);subStartPos = endPos;do {endPos = ((MLexer) lexer).getPos();lexer.nextToken();} while (lexer.token() != Token.RPAREN);String replaceMent = sql.substring(subStartPos, endPos);if (count != null && count.length > posInc && count[posInc] != null && count[posInc] > 0) {List<String> incStrs = Lists.newArrayList();for (int i = 0; i < count[posInc]; i++) {incStrs.add("?");}tempSql.append(String.join(",", incStrs));} else {tempSql.append(replaceMent);}posInc++;}continue for_;default:break;}lexer.nextToken();}tempSql.append(sql.substring(endPos));return tempSql.toString();}public static String getSqlFromReplaceNameIfNeccessary(String sql, DbType dbType, @Nullable String replaceTableName) {String table = null;String newSql = sql;Lexer lexer = createLexer(sql, dbType);lexer.nextToken();SQLExprParser exprParser;switch (dbType) {case mysql:exprParser = new MySqlExprParser(lexer);break;default:exprParser = new SQLExprParser(lexer);break;}for_:for (; lexer.token() != Token.EOF; ) {switch (lexer.token()) {case FROM:case INTO:lexer.nextToken();if (lexer.token() != Token.LPAREN && lexer.token() != Token.VALUES) {if (StringUtils.hasText(replaceTableName)) {int mark = ((MLexer) lexer).getMark();int bufPoss = ((MLexer) lexer).getBufPos();StringBuilder sbd = new StringBuilder();sbd.append(sql.substring(0, mark));sbd.append(replaceTableName);sbd.append(sql.substring(mark + bufPoss));newSql = sbd.toString();}SQLName name = exprParser.name();table = name.toString();}break for_;default:break;}lexer.nextToken();}System.out.println("原本的表名:" + table);System.out.println("替换表名为:" + replaceTableName + "后的sql:" + newSql);return newSql;}public static Lexer createLexer(String sql, DbType dbType) {return createLexer(sql, dbType, new SQLParserFeature[0]);}public static Lexer createLexer(String sql, DbType dbType, SQLParserFeature... features) {if (dbType == null) {dbType = DbType.mysql;}switch (dbType) {case mysql:return new MLexer(sql);default:return new Lexer(sql, null, dbType);}}private static class MLexer extends MySqlLexer {public MLexer(char[] input, int inputLength, boolean skipComment) {super(input, inputLength, skipComment);}public MLexer(String input) {super(input);}public MLexer(String input, SQLParserFeature... features) {super(input, features);}public MLexer(String input, boolean skipComment, boolean keepComments) {super(input, skipComment, keepComments);}public int getBufPos() {return this.bufPos;}public int getMark() {return this.mark;}public int getPos() {return this.pos;}}}
getSqlInHoldCountIfNeccessary方法的效果是,我们知道在mysql的子查询IN中,假设一个sql有多处具有子查询IN,假定为:in (?)。但是我们需要自定义IN子查询后续的参数个数,意即类似更新sql的子查询参数数目为in (?,?,?),同时不改变原有sql的语句,那么通过我们自定义的该方法可以达到效果,测试如下:
package com.xiaoxu.parser;import com.alibaba.druid.DbType;/*** @author xiaoxu* @date 2024-03-12* java_demo2:com.xiaoxu.parser.SQLParserTest2*/
public class SQLParserTest2 {public static void main(String[] args) {String sql = SQLParserUtil.getSqlInHoldCountIfNeccessary("select * from my where id in( ?,?) and status is not null and name in (?) and ot ='N' and pr in ()",DbType.mysql, new Integer[]{null, 4});System.out.println("最终结果是:");System.out.println(sql);String sql2 = SQLParserUtil.getSqlInHoldCountIfNeccessary("select * from my where id in (?)",DbType.mysql, new Integer[]{3});System.out.println(sql2);String sql3 = SQLParserUtil.getSqlInHoldCountIfNeccessary("select * from my where id = ?",DbType.mysql, new Integer[]{3});System.out.println(sql3);}}
执行结果如下:
其中参数new Integer[]{null, 4}的效果是,第一个IN子查询不变,第二个子查询更新为in (?,?,?,?)。该逻辑是按照顺序更新IN的后续参数数目,同时不改变原有的sql。