All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.apache.tomcat.util.http.fileupload.ParameterParser Maven / Gradle / Ivy

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You 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 org.apache.tomcat.util.http.fileupload;

import java.util.HashMap;
import java.util.Locale;
import java.util.Map;

/**
 * A simple parser intended to parse sequences of name/value pairs.
 * Parameter values are exptected to be enclosed in quotes if they
 * contain unsafe characters, such as '=' characters or separators.
 * Parameter values are optional and can be omitted.
 *
 * 

* param1 = value; param2 = "anything goes; really"; param3 *

* * @author Oleg Kalnichevski */ public class ParameterParser { /** * String to be parsed. */ private char[] chars = null; /** * Current position in the string. */ private int pos = 0; /** * Maximum position in the string. */ private int len = 0; /** * Start of a token. */ private int i1 = 0; /** * End of a token. */ private int i2 = 0; /** * Whether names stored in the map should be converted to lower case. */ private boolean lowerCaseNames = false; /** * Default ParameterParser constructor. */ public ParameterParser() { super(); } /** * Are there any characters left to parse? * * @return true if there are unparsed characters, * false otherwise. */ private boolean hasChar() { return this.pos < this.len; } /** * A helper method to process the parsed token. This method removes * leading and trailing blanks as well as enclosing quotation marks, * when necessary. * * @param quoted true if quotation marks are expected, * false otherwise. * @return the token */ private String getToken(boolean quoted) { // Trim leading white spaces while ((i1 < i2) && (Character.isWhitespace(chars[i1]))) { i1++; } // Trim trailing white spaces while ((i2 > i1) && (Character.isWhitespace(chars[i2 - 1]))) { i2--; } // Strip away quotation marks if necessary if (quoted) { if (((i2 - i1) >= 2) && (chars[i1] == '"') && (chars[i2 - 1] == '"')) { i1++; i2--; } } String result = null; if (i2 > i1) { result = new String(chars, i1, i2 - i1); } return result; } /** * Tests if the given character is present in the array of characters. * * @param ch the character to test for presense in the array of characters * @param charray the array of characters to test against * * @return true if the character is present in the array of * characters, false otherwise. */ private boolean isOneOf(char ch, final char[] charray) { boolean result = false; for (int i = 0; i < charray.length; i++) { if (ch == charray[i]) { result = true; break; } } return result; } /** * Parses out a token until any of the given terminators * is encountered. * * @param terminators the array of terminating characters. Any of these * characters when encountered signify the end of the token * * @return the token */ private String parseToken(final char[] terminators) { char ch; i1 = pos; i2 = pos; while (hasChar()) { ch = chars[pos]; if (isOneOf(ch, terminators)) { break; } i2++; pos++; } return getToken(false); } /** * Parses out a token until any of the given terminators * is encountered outside the quotation marks. * * @param terminators the array of terminating characters. Any of these * characters when encountered outside the quotation marks signify the end * of the token * * @return the token */ private String parseQuotedToken(final char[] terminators) { char ch; i1 = pos; i2 = pos; boolean quoted = false; boolean charEscaped = false; while (hasChar()) { ch = chars[pos]; if (!quoted && isOneOf(ch, terminators)) { break; } if (!charEscaped && ch == '"') { quoted = !quoted; } charEscaped = (!charEscaped && ch == '\\'); i2++; pos++; } return getToken(true); } /** * Returns true if parameter names are to be converted to lower * case when name/value pairs are parsed. * * @return true if parameter names are to be * converted to lower case when name/value pairs are parsed. * Otherwise returns false */ public boolean isLowerCaseNames() { return this.lowerCaseNames; } /** * Sets the flag if parameter names are to be converted to lower case when * name/value pairs are parsed. * * @param b true if parameter names are to be * converted to lower case when name/value pairs are parsed. * false otherwise. */ public void setLowerCaseNames(boolean b) { this.lowerCaseNames = b; } /** * Extracts a map of name/value pairs from the given string. Names are * expected to be unique. Multiple separators may be specified and * the earliest found in the input string is used. * * @param str the string that contains a sequence of name/value pairs * @param separators the name/value pairs separators * * @return a map of name/value pairs */ public Map parse(final String str, char[] separators) { if (separators == null || separators.length == 0) { return new HashMap(); } char separator = separators[0]; if (str != null) { int idx = str.length(); for (int i = 0; i < separators.length; i++) { int tmp = str.indexOf(separators[i]); if (tmp != -1) { if (tmp < idx) { idx = tmp; separator = separators[i]; } } } } return parse(str, separator); } /** * Extracts a map of name/value pairs from the given string. Names are * expected to be unique. * * @param str the string that contains a sequence of name/value pairs * @param separator the name/value pairs separator * * @return a map of name/value pairs */ public Map parse(final String str, char separator) { if (str == null) { return new HashMap(); } return parse(str.toCharArray(), separator); } /** * Extracts a map of name/value pairs from the given array of * characters. Names are expected to be unique. * * @param inputChars the array of characters that contains a sequence of * name/value pairs * @param separator the name/value pairs separator * * @return a map of name/value pairs */ public Map parse(final char[] inputChars, char separator) { if (inputChars == null) { return new HashMap(); } return parse(inputChars, 0, inputChars.length, separator); } /** * Extracts a map of name/value pairs from the given array of * characters. Names are expected to be unique. * * @param inputChars the array of characters that contains a sequence of * name/value pairs * @param offset - the initial offset. * @param length - the length. * @param separator the name/value pairs separator * * @return a map of name/value pairs */ public Map parse( final char[] inputChars, int offset, int length, char separator) { if (inputChars == null) { return new HashMap(); } HashMap params = new HashMap(); this.chars = inputChars; this.pos = offset; this.len = length; String paramName = null; String paramValue = null; while (hasChar()) { paramName = parseToken(new char[] { '=', separator }); paramValue = null; if (hasChar() && (chars[pos] == '=')) { pos++; // skip '=' paramValue = parseQuotedToken(new char[] { separator }); } if (hasChar() && (chars[pos] == separator)) { pos++; // skip separator } if ((paramName != null) && (paramName.length() > 0)) { if (this.lowerCaseNames) { paramName = paramName.toLowerCase(Locale.ENGLISH); } params.put(paramName, paramValue); } } return params; } }




© 2015 - 2024 Weber Informatics LLC | Privacy Policy