1
0
mirror of https://github.com/moparisthebest/k-9 synced 2024-12-25 09:08:49 -05:00
k-9/src/com/fsck/k9/helper/Utility.java
Andrew Gaul 2fd9bd5a03 Cleanup Utility.combine()
Fix spelling mistakes, simplify adding separators, use unsynchronized
StringBuilder, and fix Javadoc.
2011-05-31 15:11:36 +02:00

603 lines
21 KiB
Java

package com.fsck.k9.helper;
import android.text.Editable;
import android.util.Log;
import android.widget.EditText;
import android.widget.TextView;
import com.fsck.k9.K9;
import com.fsck.k9.mail.filter.Base64;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.io.UnsupportedEncodingException;
import java.util.Date;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
public class Utility {
// \u00A0 (non-breaking space) happens to be used by French MUA
// Note: no longer using the ^ beginning character combined with (...)+
// repetition matching as we might want to strip ML tags. Ex:
// Re: [foo] Re: RE : [foo] blah blah blah
private static final Pattern RESPONSE_PATTERN = Pattern.compile(
"((Re|Fw|Fwd|Aw|R\\u00E9f\\.)(\\[\\d+\\])?[\\u00A0 ]?: *)+", Pattern.CASE_INSENSITIVE);
/**
* Mailing-list tag pattern to match strings like "[foobar] "
*/
private static final Pattern TAG_PATTERN = Pattern.compile("\\[[-_a-z0-9]+\\] ",
Pattern.CASE_INSENSITIVE);
public static String readInputStream(InputStream in, String encoding) throws IOException {
InputStreamReader reader = new InputStreamReader(in, encoding);
StringBuffer sb = new StringBuffer();
int count;
char[] buf = new char[512];
while ((count = reader.read(buf)) != -1) {
sb.append(buf, 0, count);
}
return sb.toString();
}
public static boolean arrayContains(Object[] a, Object o) {
for (Object element : a) {
if (element.equals(o)) {
return true;
}
}
return false;
}
/**
* Combines the given array of Objects into a single String using
* each Object's toString() method and the separator character
* between each part.
*
* @param parts
* @param separator
* @return new String
*/
public static String combine(Object[] parts, char separator) {
if (parts == null) {
return null;
} else if (parts.length == 0) {
return "";
}
StringBuilder sb = new StringBuilder();
sb.append(parts[0]);
for (int i = 1; i < parts.length; ++i) {
sb.append(separator);
sb.append(parts[i]);
}
return sb.toString();
}
public static String base64Decode(String encoded) {
if (encoded == null) {
return null;
}
byte[] decoded = new Base64().decode(encoded.getBytes());
return new String(decoded);
}
public static String base64Encode(String s) {
if (s == null) {
return s;
}
byte[] encoded = new Base64().encode(s.getBytes());
return new String(encoded);
}
public static boolean requiredFieldValid(TextView view) {
return view.getText() != null && view.getText().length() > 0;
}
public static boolean requiredFieldValid(Editable s) {
return s != null && s.length() > 0;
}
public static boolean domainFieldValid(EditText view) {
if (view.getText() != null) {
String s = view.getText().toString();
if (s.matches("^([a-zA-Z0-9]([a-zA-Z0-9\\-]{0,61}[a-zA-Z0-9])?\\.)+[a-zA-Z]{2,6}$")) {
return true;
}
if (s.matches("^(?:(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\\.){3}(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)$")) {
return true;
}
if ((s.equalsIgnoreCase("localhost")) || (s.equalsIgnoreCase("localhost.localdomain"))) {
return true;
}
}
return false;
}
private static final Pattern ATOM = Pattern.compile("^(?:[a-zA-Z0-9!#$%&'*+\\-/=?^_`{|}~]|\\s)+$");
/**
* Quote a string, if necessary, based upon the definition of an "atom," as defined by RFC2822
* (http://tools.ietf.org/html/rfc2822#section-3.2.4). Strings that consist purely of atoms are
* left unquoted; anything else is returned as a quoted string.
* @param text String to quote.
* @return Possibly quoted string.
*/
public static String quoteAtoms(final String text) {
if (ATOM.matcher(text).matches()) {
return text;
} else {
return quoteString(text);
}
}
/**
* Ensures that the given string starts and ends with the double quote character. The string is not modified in any way except to add the
* double quote character to start and end if it's not already there.
* sample -> "sample"
* "sample" -> "sample"
* ""sample"" -> "sample"
* "sample"" -> "sample"
* sa"mp"le -> "sa"mp"le"
* "sa"mp"le" -> "sa"mp"le"
* (empty string) -> ""
* " -> ""
* @param s
* @return
*/
public static String quoteString(String s) {
if (s == null) {
return null;
}
if (!s.matches("^\".*\"$")) {
return "\"" + s + "\"";
} else {
return s;
}
}
/**
* A fast version of URLDecoder.decode() that works only with UTF-8 and does only two
* allocations. This version is around 3x as fast as the standard one and I'm using it
* hundreds of times in places that slow down the UI, so it helps.
*/
public static String fastUrlDecode(String s) {
try {
byte[] bytes = s.getBytes("UTF-8");
byte ch;
int length = 0;
for (int i = 0, count = bytes.length; i < count; i++) {
ch = bytes[i];
if (ch == '%') {
int h = (bytes[i + 1] - '0');
int l = (bytes[i + 2] - '0');
if (h > 9) {
h -= 7;
}
if (l > 9) {
l -= 7;
}
bytes[length] = (byte)((h << 4) | l);
i += 2;
} else if (ch == '+') {
bytes[length] = ' ';
} else {
bytes[length] = bytes[i];
}
length++;
}
return new String(bytes, 0, length, "UTF-8");
} catch (UnsupportedEncodingException uee) {
return null;
}
}
/**
* Returns true if the specified date is within 18 hours of "now". Returns false otherwise.
* @param date
* @return
*/
public static boolean isDateToday(Date date) {
Date now = new Date();
if (now.getTime() - 64800000 > date.getTime() || now.getTime() + 64800000 < date.getTime()) {
return false;
} else {
return true;
}
}
/*
* TODO disabled this method globally. It is used in all the settings screens but I just
* noticed that an unrelated icon was dimmed. Android must share drawables internally.
*/
public static void setCompoundDrawablesAlpha(TextView view, int alpha) {
// Drawable[] drawables = view.getCompoundDrawables();
// for (Drawable drawable : drawables) {
// if (drawable != null) {
// drawable.setAlpha(alpha);
// }
// }
}
/**
* <p>Wraps a multiline string of text, identifying words by <code>' '</code>.</p>
*
* <p>New lines will be separated by the system property line separator.
* Very long words, such as URLs will <i>not</i> be wrapped.</p>
*
* <p>Leading spaces on a new line are stripped.
* Trailing spaces are not stripped.</p>
*
* <pre>
* WordUtils.wrap(null, *) = null
* WordUtils.wrap("", *) = ""
* </pre>
*
* Adapted from the Apache Commons Lang library.
* http://svn.apache.org/viewvc/commons/proper/lang
* /trunk/src/main/java/org/apache/commons/lang3/text/WordUtils.java
* SVN Revision 925967, Mon Mar 22 06:16:49 2010 UTC
*
* 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.
*
* @param str the String to be word wrapped, may be null
* @param wrapLength the column to wrap the words at, less than 1 is treated as 1
* @return a line with newlines inserted, <code>null</code> if null input
*/
private static final String NEWLINE_REGEX = "(?:\\r?\\n)";
public static String wrap(String str, int wrapLength) {
StringBuilder result = new StringBuilder();
for (String piece : str.split(NEWLINE_REGEX)) {
result.append(wrap(piece, wrapLength, null, false));
result.append("\n");
}
return result.toString();
}
/**
* <p>Wraps a single line of text, identifying words by <code>' '</code>.</p>
*
* <p>Leading spaces on a new line are stripped.
* Trailing spaces are not stripped.</p>
*
* <pre>
* WordUtils.wrap(null, *, *, *) = null
* WordUtils.wrap("", *, *, *) = ""
* </pre>
*
* This is from the Apache Commons Lang library.
* http://svn.apache.org/viewvc/commons/proper/lang
* /trunk/src/main/java/org/apache/commons/lang3/text/WordUtils.java
* SVN Revision 925967, Mon Mar 22 06:16:49 2010 UTC
*
* 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.
*
* @param str the String to be word wrapped, may be null
* @param wrapLength the column to wrap the words at, less than 1 is treated as 1
* @param newLineStr the string to insert for a new line,
* <code>null</code> uses the system property line separator
* @param wrapLongWords true if long words (such as URLs) should be wrapped
* @return a line with newlines inserted, <code>null</code> if null input
*/
public static String wrap(String str, int wrapLength, String newLineStr, boolean wrapLongWords) {
if (str == null) {
return null;
}
if (newLineStr == null) {
newLineStr = "\n";
}
if (wrapLength < 1) {
wrapLength = 1;
}
int inputLineLength = str.length();
int offset = 0;
StringBuilder wrappedLine = new StringBuilder(inputLineLength + 32);
while ((inputLineLength - offset) > wrapLength) {
if (str.charAt(offset) == ' ') {
offset++;
continue;
}
int spaceToWrapAt = str.lastIndexOf(' ', wrapLength + offset);
if (spaceToWrapAt >= offset) {
// normal case
wrappedLine.append(str.substring(offset, spaceToWrapAt));
wrappedLine.append(newLineStr);
offset = spaceToWrapAt + 1;
} else {
// really long word or URL
if (wrapLongWords) {
// wrap really long word one line at a time
wrappedLine.append(str.substring(offset, wrapLength + offset));
wrappedLine.append(newLineStr);
offset += wrapLength;
} else {
// do not wrap really long word, just extend beyond limit
spaceToWrapAt = str.indexOf(' ', wrapLength + offset);
if (spaceToWrapAt >= 0) {
wrappedLine.append(str.substring(offset, spaceToWrapAt));
wrappedLine.append(newLineStr);
offset = spaceToWrapAt + 1;
} else {
wrappedLine.append(str.substring(offset));
offset = inputLineLength;
}
}
}
}
// Whatever is left in line is short enough to just pass through
wrappedLine.append(str.substring(offset));
return wrappedLine.toString();
}
/**
* Extract the 'original' subject value, by ignoring leading
* response/forward marker and '[XX]' formatted tags (as many mailing-list
* softwares do).
*
* <p>
* Result is also trimmed.
* </p>
*
* @param subject
* Never <code>null</code>.
* @return Never <code>null</code>.
*/
public static String stripSubject(final String subject) {
int lastPrefix = 0;
final Matcher tagMatcher = TAG_PATTERN.matcher(subject);
String tag = null;
// whether tag stripping logic should be active
boolean tagPresent = false;
// whether the last action stripped a tag
boolean tagStripped = false;
if (tagMatcher.find(0)) {
tagPresent = true;
if (tagMatcher.start() == 0) {
// found at beginning of subject, considering it an actual tag
tag = tagMatcher.group();
// now need to find response marker after that tag
lastPrefix = tagMatcher.end();
tagStripped = true;
}
}
final Matcher matcher = RESPONSE_PATTERN.matcher(subject);
// while:
// - lastPrefix is within the bounds
// - response marker found at lastPrefix position
// (to make sure we don't catch response markers that are part of
// the actual subject)
while (lastPrefix < subject.length() - 1
&& matcher.find(lastPrefix)
&& matcher.start() == lastPrefix
&& (!tagPresent || tag == null || subject.regionMatches(matcher.end(), tag, 0,
tag.length()))) {
lastPrefix = matcher.end();
if (tagPresent) {
tagStripped = false;
if (tag == null) {
// attempt to find tag
if (tagMatcher.start() == lastPrefix) {
tag = tagMatcher.group();
lastPrefix += tag.length();
tagStripped = true;
}
} else if (lastPrefix < subject.length() - 1 && subject.startsWith(tag, lastPrefix)) {
// Re: [foo] Re: [foo] blah blah blah
// ^ ^
// ^ ^
// ^ new position
// ^
// initial position
lastPrefix += tag.length();
tagStripped = true;
}
}
}
// Null pointer check is to make the static analysis component of Eclipse happy.
if (tagStripped && (tag != null)) {
// restore the last tag
lastPrefix -= tag.length();
}
if (lastPrefix > -1 && lastPrefix < subject.length() - 1) {
return subject.substring(lastPrefix).trim();
} else {
return subject.trim();
}
}
/**
* @param parentDir
* @param name
* Never <code>null</code>.
*/
public static void touchFile(final File parentDir, final String name) {
final File file = new File(parentDir, name);
try {
if (!file.exists()) {
file.createNewFile();
} else {
file.setLastModified(System.currentTimeMillis());
}
} catch (Exception e) {
Log.d(K9.LOG_TAG, "Unable to touch file: " + file.getAbsolutePath(), e);
}
}
/**
* Creates a unique file in the given directory by appending a hyphen
* and a number to the given filename.
*
* @param directory
* @param filename
* @return
*/
public static File createUniqueFile(File directory, String filename) {
File file = new File(directory, filename);
if (!file.exists()) {
return file;
}
// Get the extension of the file, if any.
int index = filename.lastIndexOf('.');
String format;
if (index != -1) {
String name = filename.substring(0, index);
String extension = filename.substring(index);
format = name + "-%d" + extension;
} else {
format = filename + "-%d";
}
for (int i = 2; i < Integer.MAX_VALUE; i++) {
file = new File(directory, String.format(format, i));
if (!file.exists()) {
return file;
}
}
return null;
}
/**
* @param from
* @param to
* @return
*/
public static boolean move(final File from, final File to) {
if (to.exists()) {
to.delete();
}
to.getParentFile().mkdirs();
try {
FileInputStream in = new FileInputStream(from);
FileOutputStream out = new FileOutputStream(to);
byte[] buffer = new byte[1024];
int count = -1;
while ((count = in.read(buffer)) > 0) {
out.write(buffer, 0, count);
}
out.close();
in.close();
from.delete();
return true;
} catch (Exception e) {
Log.w(K9.LOG_TAG, "cannot move " + from.getAbsolutePath() + " to " + to.getAbsolutePath(), e);
return false;
}
}
/**
* @param fromDir
* @param toDir
*/
public static void moveRecursive(final File fromDir, final File toDir) {
if (!fromDir.exists()) {
return;
}
if (!fromDir.isDirectory()) {
if (toDir.exists()) {
if (!toDir.delete()) {
Log.w(K9.LOG_TAG, "cannot delete already existing file/directory " + toDir.getAbsolutePath());
}
}
if (!fromDir.renameTo(toDir)) {
Log.w(K9.LOG_TAG, "cannot rename " + fromDir.getAbsolutePath() + " to " + toDir.getAbsolutePath() + " - moving instead");
move(fromDir, toDir);
}
return;
}
if (!toDir.exists() || !toDir.isDirectory()) {
if (toDir.exists()) {
toDir.delete();
}
if (!toDir.mkdirs()) {
Log.w(K9.LOG_TAG, "cannot create directory " + toDir.getAbsolutePath());
}
}
File[] files = fromDir.listFiles();
for (File file : files) {
if (file.isDirectory()) {
moveRecursive(file, new File(toDir, file.getName()));
file.delete();
} else {
File target = new File(toDir, file.getName());
if (!file.renameTo(target)) {
Log.w(K9.LOG_TAG, "cannot rename " + file.getAbsolutePath() + " to " + target.getAbsolutePath() + " - moving instead");
move(file, target);
}
}
}
if (!fromDir.delete()) {
Log.w(K9.LOG_TAG, "cannot delete " + fromDir.getAbsolutePath());
}
}
private static final String IMG_SRC_REGEX = "(?is:<img[^>]+src\\s*=\\s*['\"]?([a-z]+)\\:)";
private static final Pattern IMG_PATTERN = Pattern.compile(IMG_SRC_REGEX);
/**
* Figure out if this part has images.
* TODO: should only return true if we're an html part
* @param message Content to evaluate
* @return True if it has external images; false otherwise.
*/
public static boolean hasExternalImages(final String message) {
Matcher imgMatches = IMG_PATTERN.matcher(message);
while (imgMatches.find()) {
if (!imgMatches.group(1).equals("content")) {
if (K9.DEBUG) {
Log.d(K9.LOG_TAG, "External images found");
}
return true;
}
}
if (K9.DEBUG) {
Log.d(K9.LOG_TAG, "No external images.");
}
return false;
}
}