All Downloads are FREE. Search and download functionalities are using the official Maven repository.
Please wait. This can take some minutes ...
Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
scouter.util.SQLSimpleParser Maven / Gradle / Ivy
/*
* Copyright 2015 the original author or authors.
* @https://github.com/scouter-project/scouter
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package scouter.util;
import java.util.ArrayList;
import java.util.Arrays;
public class SQLSimpleParser {
SQLNode sqlNode = null;
SQLNode currentNode = null;
boolean inParenthesis = false;
int parenthesisCount = 0;
// Stack sqlStack = new Stack();
public class SQLNode {
SQLTypeEnum type;
ArrayList tableList;
int depth;
SQLNode nextNode = null;
public SQLNode() {
tableList = new ArrayList();
depth = 0;
}
public SQLNode(SQLTypeEnum type) {
tableList = new ArrayList();
this.type = type;
}
public SQLNode(SQLTypeEnum type, int depth) {
this.type = type;
tableList = new ArrayList();
this.depth = depth;
}
};
enum SQLTypeEnum {
SELECT, DELETE, UPDATE, INSERT, MERGE;
@Override
public String toString() {
switch (this) {
case SELECT:
return "R";
case DELETE:
return "D";
case UPDATE:
return "U";
case INSERT:
return "C";
case MERGE:
return "C";
default:
return "R";
}
}
};
private ArrayList spliter = new ArrayList(
Arrays.asList('=', '<', '>', '!', ',', ' ', '(', ')','\r','\n'));
int depth = 0;
private void createOrAppendNode(SQLTypeEnum type) {
depth++;
if (sqlNode == null) {
sqlNode = new SQLNode(type, depth);
sqlNode.nextNode = null;
} else {
appendNode(sqlNode, new SQLNode(type, depth));
}
/*
* SQLNode node = new SQLNode(type); node.nextNode = null;
*
* sqlStack.push(sqlNode);
*/
}
private void appendNode(SQLNode head, SQLNode newNode) {
while (head.nextNode != null) {
head = head.nextNode;
}
head.nextNode = newNode;
}
private void release() {
//clearNode(sqlNode);
clearNode();
sqlNode = null;
depth = 0;
parenthesisCount = 0;
}
private void clearNode() {
if(sqlNode == null) {
return;
}
SQLNode currNode = null;
if(sqlNode.nextNode != null) {
currNode = sqlNode;
sqlNode = sqlNode.nextNode;
currNode = null;
} else {
sqlNode = null;
}
clearNode();
}
public String getCrudInfo(String value) {
try {
/*String[] lines = value.split("[\r\n]");
for (int i = 0; i < lines.length; i++) {
String[] tokens = tokenize(lines[i], false);
crudInfo(tokens);
}*/
String[] tokens = tokenize(value, false);
crudInfo(tokens);
} catch (Exception ex) {
return "";
}
StringBuffer sb = new StringBuffer();
ArrayList tempList = new ArrayList();
SQLNode node = sqlNode;
do {
for (int i = 0; i < node.tableList.size(); i++) {
StringBuffer s = new StringBuffer();
String tblInfo = s.append(node.tableList.get(i)).append("(").append(node.type.toString())
.append(")").toString();
if (!tempList.contains(tblInfo)) {
if (sb.length() > 0) {
sb.append(",");
}
sb.append(tblInfo);
tempList.add(tblInfo);
}
}
node = node.nextNode;
} while (node != null);
// printCRUD() ;
release();
return sb.toString();
}
public void printCRUD() {
do {
System.out.print("type:" + sqlNode.type.toString() + "-->");
for (int i = 0; i < sqlNode.tableList.size(); i++) {
System.out.print(sqlNode.tableList.get(i) + " ");
}
System.out.println();
sqlNode = sqlNode.nextNode;
} while (sqlNode != null);
}
public void crudInfo(String[] tokens) throws Exception {
try {
for (int i = 0; i < tokens.length; i++) {
String token = tokens[i].toUpperCase();
if(token.equals("(")){
parenthesisCount++;
inParenthesis = true;
} else if(token.equals(")")) {
parenthesisCount--;
if (parenthesisCount == 0) {
inParenthesis = false;
}
} else if(token.equals("SELECT")) {
if (inParenthesis) {
for (int j = i + 1; j < tokens.length; j++) {
if (tokens[j].toLowerCase().equals("from")) {
createOrAppendNode(SQLTypeEnum.SELECT);
j = applyNode(j, tokens);
}
if (tokens[j].equals("(")) {
parenthesisCount++;
inParenthesis = true;
}
if (tokens[j].equals(")")) {
parenthesisCount--;
if (parenthesisCount == 0) {
inParenthesis = false;
i = j;
break;
}
}
}
} else {
createOrAppendNode(SQLTypeEnum.SELECT);
}
} else if(token.equals("DELETE")) {
createOrAppendNode(SQLTypeEnum.DELETE);
} else if(token.equals("INSERT")) {
createOrAppendNode(SQLTypeEnum.INSERT);
} else if(token.equals("UPDATE")) {
createOrAppendNode(SQLTypeEnum.UPDATE);
i = applyNode(i, tokens);
} else if(token.equals("MERGE")) {
if (tokens[i + 1].toUpperCase().equals("INTO")) {
createOrAppendNode(SQLTypeEnum.MERGE);
}
} else if(token.equals("FROM")) {
i = applyNode(i, tokens);
} else if(token.equals("INTO")) {
i = applyNode(i, tokens);
} else if(token.equals("JOIN")) {
createOrAppendNode(SQLTypeEnum.SELECT);
i = applyNode(i, tokens);
}
}
} catch (Exception e) {
throw e;
}
}
private int applyNode(int index, String[] tokens) throws Exception {
int returnIndex = index;
try {
if(index == tokens.length -1 ) {
throw new RuntimeException(index + " is the last index of tokens.");
}
if(depth == 0) {
return index;
}
SQLNode node = findNode(depth);
if(node == null) {
throw new RuntimeException("Can't find node which has proper depth.");
}
if (node.type == SQLTypeEnum.SELECT) {
if (!tokens[index + 1].equals("(")) {
node.tableList.add((tokens[index + 1]));
returnIndex = index + 1;
int step = 0;
// select
// from
// table1, table2
if (tokens.length > (index + 2) && tokens[index + 2].equals(",")) {
step = 2;
}
// select
// from
// table1 a, table2 b
if (tokens.length > (index + 3) && tokens[index + 3].equals(",")) {
step = 3;
}
if (step > 0) {
for (int i = index + 1; i < tokens.length; i += step) {
if (tokens.length >= i + step) {
if (tokens[i + step - 1].equals(",")) {
if (!tokens[i + step].equals("(")) {
node.tableList.add(tokens[i + step]);
returnIndex = i + step;
}
} else {
break;
}
} else {
break;
}
}
}
}
} else { // except select clause.
node.tableList.add((tokens[index + 1]));
returnIndex = index + 1;
}
node.depth = -1; // set node.depth = -1 if node used.
depth--;
} catch (Exception ex) {
throw ex;
}
return returnIndex;
}
private SQLNode findNode(int depth) {
SQLNode node = sqlNode;
if (node.nextNode == null) {
return node;
} else {
while (node.nextNode != null) {
node = node.nextNode;
if (node.depth == depth) {
break;
}
}
}
return node;
}
/**
* tokenize sql with given spliter.
*
* @param value
* @param keepComments
* @return string array
* @throws Exception
*/
private String[] tokenize(String value, boolean keepComments) throws Exception {
char[] arrays = value.toCharArray();
boolean hasQuotation = false;
boolean hasSingleLineComments = false;
boolean hasMultiLineComments = false;
int start = 0;
ArrayList tokenList = new ArrayList();
int len = arrays.length;
for (int i = 0; i < len; i++) {
if (i == (len - 1)) {
if (spliter.contains(arrays[i])) {
char[] token = Arrays.copyOfRange(arrays, start, len - 1);
tokenList.add(new String(token));
if(arrays[i] != ' ') {
tokenList.add(String.valueOf(arrays[i]));
}
} else {
char[] token = Arrays.copyOfRange(arrays, start, len);
tokenList.add(new String(token));
}
break;
}
if (arrays[i] == '\'') { // single quotation(')
if (hasQuotation == false) {
hasQuotation = true;
} else {
hasQuotation = false;
}
continue;
}
if (arrays[i] == '-') {
if (!hasQuotation) {
if (i < len - 1 && arrays[i + 1] == '-') { // sql comment
hasSingleLineComments = true;
}
}
continue;
}
if (arrays[i] == '#') {
if (!hasQuotation) { // sql comments for mysql
hasSingleLineComments = true;
}
continue;
}
if(arrays[i] == '\n') {
if(hasSingleLineComments) {
hasSingleLineComments = false;
}
}
if (arrays[i] == '/') {
if (i < len - 1 && arrays[i + 1] == '*') { // sql comments or hint
hasMultiLineComments = true;
}
if (i > 0 && arrays[i - 1] == '*' && hasMultiLineComments) {
hasMultiLineComments = false;
if (!keepComments) {
start = i + 1;
}
}
continue;
}
if (i > 0 && spliter.contains(arrays[i])) {
if (hasQuotation || hasSingleLineComments || hasMultiLineComments) { // ignore comments, literal
continue;
}
if (i >= start) {
char[] token = Arrays.copyOfRange(arrays, start, i);
/*
* 구분자가 연속으로 있는 경우 -> ex: 공백 + ( ==> select * from id in (...)
* start 와 index 는 동일값. 이러한 경우 token.length = 0
*/
if (token.length > 0) {
tokenList.add(new String(token));
}
if(arrays[i] ==' ') {
start = i+1;
continue;
}
if(arrays[i] == '\r') {
if(arrays[i+1] == '\n') {
start = i + 2;
i++;
continue;
} else {
start = i+1;
}
continue;
}
if(arrays[i] == '\n') {
start = i+1;
continue;
}
//other case => '=' '<' '>' '!' ',' '(' ')'
if (spliter.contains(arrays[i+1])) { // ex) >=, <=, <>...
String temp = "" + arrays[i] + arrays[i + 1];
if (temp.equals("<>") || temp.equals("!=") || temp.equals("<=") || temp.equals(">=")) {
tokenList.add(temp);
start = i + 1;
i++;
} else {
tokenList.add(String.valueOf(arrays[i]));
}
} else {
tokenList.add(String.valueOf(arrays[i]));
}
}
start = i + 1;
continue;
}
}
return tokenList.toArray(new String[tokenList.size()]);
}
}