org.supercsv.io.Tokenizer Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of super-csv Show documentation
Show all versions of super-csv Show documentation
Super CSV is a fast, programmer-friendly, free CSV package for Java
/*
* Copyright 2007 Kasper B. Graversen
*
* 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 org.supercsv.io;
import java.io.IOException;
import java.io.Reader;
import java.util.List;
import org.supercsv.comment.CommentMatcher;
import org.supercsv.exception.SuperCsvException;
import org.supercsv.prefs.CsvPreference;
/**
* Reads the CSV file, line by line. If you want the line-reading functionality of this class, but want to define your
* own implementation of {@link #readColumns(List)}, then consider writing your own Tokenizer by extending
* AbstractTokenizer.
*
* @author Kasper B. Graversen
* @author James Bassett
*/
public class Tokenizer extends AbstractTokenizer {
private static final char NEWLINE = '\n';
private static final char SPACE = ' ';
private final StringBuilder currentColumn = new StringBuilder();
/* the raw, untokenized CSV row (may span multiple lines) */
private final StringBuilder currentRow = new StringBuilder();
private final int quoteChar;
private final int delimeterChar;
private final boolean surroundingSpacesNeedQuotes;
private final boolean ignoreEmptyLines;
private final CommentMatcher commentMatcher;
/**
* Enumeration of tokenizer states. QUOTE_MODE is activated between quotes.
*/
private enum TokenizerState {
NORMAL, QUOTE_MODE;
}
/**
* Constructs a new Tokenizer, which reads the CSV file, line by line.
*
* @param reader
* the reader
* @param preferences
* the CSV preferences
* @throws NullPointerException
* if reader or preferences is null
*/
public Tokenizer(final Reader reader, final CsvPreference preferences) {
super(reader, preferences);
this.quoteChar = preferences.getQuoteChar();
this.delimeterChar = preferences.getDelimiterChar();
this.surroundingSpacesNeedQuotes = preferences.isSurroundingSpacesNeedQuotes();
this.ignoreEmptyLines = preferences.isIgnoreEmptyLines();
this.commentMatcher = preferences.getCommentMatcher();
}
/**
* {@inheritDoc}
*/
public boolean readColumns(final List columns) throws IOException {
if( columns == null ) {
throw new NullPointerException("columns should not be null");
}
// clear the reusable List and StringBuilders
columns.clear();
currentColumn.setLength(0);
currentRow.setLength(0);
// read a line (ignoring empty lines/comments if necessary)
String line;
do {
line = readLine();
if( line == null ) {
return false; // EOF
}
}
while( ignoreEmptyLines && line.length() == 0 || (commentMatcher != null && commentMatcher.isComment(line)) );
// update the untokenized CSV row
currentRow.append(line);
// process each character in the line, catering for surrounding quotes (QUOTE_MODE)
TokenizerState state = TokenizerState.NORMAL;
int quoteScopeStartingLine = -1; // the line number where a potential multi-line cell starts
int potentialSpaces = 0; // keep track of spaces (so leading/trailing space can be removed if required)
int charIndex = 0;
while( true ) {
boolean endOfLineReached = charIndex == line.length();
if( endOfLineReached )
{
if( TokenizerState.NORMAL.equals(state) ) {
/*
* Newline. Add any required spaces (if surrounding spaces don't need quotes) and return (we've read
* a line!).
*/
if( !surroundingSpacesNeedQuotes ) {
appendSpaces(currentColumn, potentialSpaces);
}
columns.add(currentColumn.length() > 0 ? currentColumn.toString() : null); // "" -> null
return true;
}
else
{
/*
* Newline. Doesn't count as newline while in QUOTESCOPE. Add the newline char, reset the charIndex
* (will update to 0 for next iteration), read in the next line, then then continue to next
* character. For a large file with an unterminated quoted section (no trailing quote), this could
* cause memory issues as it will keep reading lines looking for the trailing quote. Maybe there
* should be a configurable limit on max lines to read in quoted mode?
*/
currentColumn.append(NEWLINE);
currentRow.append(NEWLINE); // specific line terminator lost, \n will have to suffice
charIndex = 0;
line = readLine();
if( line == null ) {
throw new SuperCsvException(
String
.format(
"unexpected end of file while reading quoted column beginning on line %d and ending on line %d",
quoteScopeStartingLine, getLineNumber()));
}
currentRow.append(line); // update untokenized CSV row
if (line.length() == 0){
// consecutive newlines
continue;
}
}
}
final char c = line.charAt(charIndex);
if( TokenizerState.NORMAL.equals(state) ) {
/*
* NORMAL mode (not within quotes).
*/
if( c == delimeterChar ) {
/*
* Delimiter. Save the column (trim trailing space if required) then continue to next character.
*/
if( !surroundingSpacesNeedQuotes ) {
appendSpaces(currentColumn, potentialSpaces);
}
columns.add(currentColumn.length() > 0 ? currentColumn.toString() : null); // "" -> null
potentialSpaces = 0;
currentColumn.setLength(0);
} else if( c == SPACE ) {
/*
* Space. Remember it, then continue to next character.
*/
potentialSpaces++;
}
else if( c == quoteChar ) {
/*
* A single quote ("). Update to QUOTESCOPE (but don't save quote), then continue to next character.
*/
state = TokenizerState.QUOTE_MODE;
quoteScopeStartingLine = getLineNumber();
// cater for spaces before a quoted section (be lenient!)
if( !surroundingSpacesNeedQuotes || currentColumn.length() > 0 ) {
appendSpaces(currentColumn, potentialSpaces);
}
potentialSpaces = 0;
} else {
/*
* Just a normal character. Add any required spaces (but trim any leading spaces if surrounding
* spaces need quotes), add the character, then continue to next character.
*/
if( !surroundingSpacesNeedQuotes || currentColumn.length() > 0 ) {
appendSpaces(currentColumn, potentialSpaces);
}
potentialSpaces = 0;
currentColumn.append(c);
}
} else {
/*
* QUOTE_MODE (within quotes).
*/
if( c == quoteChar ) {
int nextCharIndex = charIndex + 1;
boolean availableCharacters = nextCharIndex < line.length();
boolean nextCharIsQuote = availableCharacters && line.charAt(nextCharIndex) == quoteChar;
if( nextCharIsQuote ) {
/*
* An escaped quote (""). Add a single quote, then move the cursor so the next iteration of the
* loop will read the character following the escaped quote.
*/
currentColumn.append(c);
charIndex++;
} else {
/*
* A single quote ("). Update to NORMAL (but don't save quote), then continue to next character.
*/
state = TokenizerState.NORMAL;
quoteScopeStartingLine = -1; // reset ready for next multi-line cell
}
} else {
/*
* Just a normal character, delimiter (they don't count in QUOTESCOPE) or space. Add the character,
* then continue to next character.
*/
currentColumn.append(c);
}
}
charIndex++; // read next char of the line
}
}
/**
* Appends the required number of spaces to the StringBuilder.
*
* @param sb
* the StringBuilder
* @param spaces
* the required number of spaces to append
*/
private static void appendSpaces(final StringBuilder sb, final int spaces) {
for( int i = 0; i < spaces; i++ ) {
sb.append(SPACE);
}
}
/**
* {@inheritDoc}
*/
public String getUntokenizedRow() {
return currentRow.toString();
}
}