Is there a standard (preferably Apache Commons or similarly non-viral) library for doing "glob" type matches in Java? When I had to do similar in Perl once, I just changed all the "." to "\.", the "*" to ".*" and the "?" to "." and that sort of thing, but I'm wondering if somebody has done the work for me.
Similar question: Create regex from glob expression
Globbing is also planned for implemented in Java 7.
See FileSystem.getPathMatcher(String) and the "Finding Files" tutorial.
There's nothing built-in, but it's pretty simple to convert something glob-like to a regex:
public static String createRegexFromGlob(String glob)
{
String out = "^";
for(int i = 0; i < glob.length(); ++i)
{
final char c = glob.charAt(i);
switch(c)
{
case '*': out += ".*"; break;
case '?': out += '.'; break;
case '.': out += "\\."; break;
case '\\': out += "\\\\"; break;
default: out += c;
}
}
out += '$';
return out;
}
this works for me, but I'm not sure if it covers the glob "standard", if there is one :)
Update by Paul Tomblin: I found a perl program that does glob conversion, and adapting it to Java I end up with:
private String convertGlobToRegEx(String line)
{
LOG.info("got line [" + line + "]");
line = line.trim();
int strLen = line.length();
StringBuilder sb = new StringBuilder(strLen);
// Remove beginning and ending * globs because they're useless
if (line.startsWith("*"))
{
line = line.substring(1);
strLen--;
}
if (line.endsWith("*"))
{
line = line.substring(0, strLen-1);
strLen--;
}
boolean escaping = false;
int inCurlies = 0;
for (char currentChar : line.toCharArray())
{
switch (currentChar)
{
case '*':
if (escaping)
sb.append("\\*");
else
sb.append(".*");
escaping = false;
break;
case '?':
if (escaping)
sb.append("\\?");
else
sb.append('.');
escaping = false;
break;
case '.':
case '(':
case ')':
case '+':
case '|':
case '^':
case '$':
case '@':
case '%':
sb.append('\\');
sb.append(currentChar);
escaping = false;
break;
case '\\':
if (escaping)
{
sb.append("\\\\");
escaping = false;
}
else
escaping = true;
break;
case '{':
if (escaping)
{
sb.append("\\{");
}
else
{
sb.append('(');
inCurlies++;
}
escaping = false;
break;
case '}':
if (inCurlies > 0 && !escaping)
{
sb.append(')');
inCurlies--;
}
else if (escaping)
sb.append("\\}");
else
sb.append("}");
escaping = false;
break;
case ',':
if (inCurlies > 0 && !escaping)
{
sb.append('|');
}
else if (escaping)
sb.append("\\,");
else
sb.append(",");
break;
default:
escaping = false;
sb.append(currentChar);
}
}
return sb.toString();
}
I'm editing into this answer rather than making my own because this answer put me on the right track.
Thanks to everyone here for their contributions. I wrote a more comprehensive conversion than any of the previous answers:
/**
* Converts a standard POSIX Shell globbing pattern into a regular expression
* pattern. The result can be used with the standard {@link java.util.regex} API to
* recognize strings which match the glob pattern.
* <p/>
* See also, the POSIX Shell language:
* http://pubs.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#tag_02_13_01
*
* @param pattern A glob pattern.
* @return A regex pattern to recognize the given glob pattern.
*/
public static final String convertGlobToRegex(String pattern) {
StringBuilder sb = new StringBuilder(pattern.length());
int inGroup = 0;
int inClass = 0;
int firstIndexInClass = -1;
char[] arr = pattern.toCharArray();
for (int i = 0; i < arr.length; i++) {
char ch = arr[i];
switch (ch) {
case '\\':
if (++i >= arr.length) {
sb.append('\\');
} else {
char next = arr[i];
switch (next) {
case ',':
// escape not needed
break;
case 'Q':
case 'E':
// extra escape needed
sb.append('\\');
default:
sb.append('\\');
}
sb.append(next);
}
break;
case '*':
if (inClass == 0)
sb.append(".*");
else
sb.append('*');
break;
case '?':
if (inClass == 0)
sb.append('.');
else
sb.append('?');
break;
case '[':
inClass++;
firstIndexInClass = i+1;
sb.append('[');
break;
case ']':
inClass--;
sb.append(']');
break;
case '.':
case '(':
case ')':
case '+':
case '|':
case '^':
case '$':
case '@':
case '%':
if (inClass == 0 || (firstIndexInClass == i && ch == '^'))
sb.append('\\');
sb.append(ch);
break;
case '!':
if (firstIndexInClass == i)
sb.append('^');
else
sb.append('!');
break;
case '{':
inGroup++;
sb.append('(');
break;
case '}':
inGroup--;
sb.append(')');
break;
case ',':
if (inGroup > 0)
sb.append('|');
else
sb.append(',');
break;
default:
sb.append(ch);
}
}
return sb.toString();
}
And the unit tests to prove it works:
/**
* @author Neil Traft
*/
public class StringUtils_ConvertGlobToRegex_Test {
@Test
public void star_becomes_dot_star() throws Exception {
assertEquals("gl.*b", StringUtils.convertGlobToRegex("gl*b"));
}
@Test
public void escaped_star_is_unchanged() throws Exception {
assertEquals("gl\\*b", StringUtils.convertGlobToRegex("gl\\*b"));
}
@Test
public void question_mark_becomes_dot() throws Exception {
assertEquals("gl.b", StringUtils.convertGlobToRegex("gl?b"));
}
@Test
public void escaped_question_mark_is_unchanged() throws Exception {
assertEquals("gl\\?b", StringUtils.convertGlobToRegex("gl\\?b"));
}
@Test
public void character_classes_dont_need_conversion() throws Exception {
assertEquals("gl[-o]b", StringUtils.convertGlobToRegex("gl[-o]b"));
}
@Test
public void escaped_classes_are_unchanged() throws Exception {
assertEquals("gl\\[-o\\]b", StringUtils.convertGlobToRegex("gl\\[-o\\]b"));
}
@Test
public void negation_in_character_classes() throws Exception {
assertEquals("gl[^a-n!p-z]b", StringUtils.convertGlobToRegex("gl[!a-n!p-z]b"));
}
@Test
public void nested_negation_in_character_classes() throws Exception {
assertEquals("gl[[^a-n]!p-z]b", StringUtils.convertGlobToRegex("gl[[!a-n]!p-z]b"));
}
@Test
public void escape_carat_if_it_is_the_first_char_in_a_character_class() throws Exception {
assertEquals("gl[\\^o]b", StringUtils.convertGlobToRegex("gl[^o]b"));
}
@Test
public void metachars_are_escaped() throws Exception {
assertEquals("gl..*\\.\\(\\)\\+\\|\\^\\$\\@\\%b", StringUtils.convertGlobToRegex("gl?*.()+|^$@%b"));
}
@Test
public void metachars_in_character_classes_dont_need_escaping() throws Exception {
assertEquals("gl[?*.()+|^$@%]b", StringUtils.convertGlobToRegex("gl[?*.()+|^$@%]b"));
}
@Test
public void escaped_backslash_is_unchanged() throws Exception {
assertEquals("gl\\\\b", StringUtils.convertGlobToRegex("gl\\\\b"));
}
@Test
public void slashQ_and_slashE_are_escaped() throws Exception {
assertEquals("\\\\Qglob\\\\E", StringUtils.convertGlobToRegex("\\Qglob\\E"));
}
@Test
public void braces_are_turned_into_groups() throws Exception {
assertEquals("(glob|regex)", StringUtils.convertGlobToRegex("{glob,regex}"));
}
@Test
public void escaped_braces_are_unchanged() throws Exception {
assertEquals("\\{glob\\}", StringUtils.convertGlobToRegex("\\{glob\\}"));
}
@Test
public void commas_dont_need_escaping() throws Exception {
assertEquals("(glob,regex),", StringUtils.convertGlobToRegex("{glob\\,regex},"));
}
}
There are couple of libraries that do Glob-like pattern matching that are more modern than the ones listed:
Theres Ants Directory Scanner And Springs AntPathMatcher
I recommend both over the other solutions since Ant Style Globbing has pretty much become the standard glob syntax in the Java world (Hudson, Spring, Ant and I think Maven).
I recently had to do it and used \Q and \E to escape the glob pattern:
private static Pattern getPatternFromGlob(String glob) {
return Pattern.compile(
"^" + Pattern.quote(glob)
.replace("*", "\\E.*\\Q")
.replace("?", "\\E.\\Q")
+ "$");
}
This is a simple Glob implementation which handles * and ? in the pattern
public class GlobMatch {
private String text;
private String pattern;
public boolean match(String text, String pattern) {
this.text = text;
this.pattern = pattern;
return matchCharacter(0, 0);
}
private boolean matchCharacter(int patternIndex, int textIndex) {
if (patternIndex >= pattern.length()) {
return false;
}
switch(pattern.charAt(patternIndex)) {
case '?':
// Match any character
if (textIndex >= text.length()) {
return false;
}
break;
case '*':
// * at the end of the pattern will match anything
if (patternIndex + 1 >= pattern.length() || textIndex >= text.length()) {
return true;
}
// Probe forward to see if we can get a match
while (textIndex < text.length()) {
if (matchCharacter(patternIndex + 1, textIndex)) {
return true;
}
textIndex++;
}
return false;
default:
if (textIndex >= text.length()) {
return false;
}
String textChar = text.substring(textIndex, textIndex + 1);
String patternChar = pattern.substring(patternIndex, patternIndex + 1);
// Note the match is case insensitive
if (textChar.compareToIgnoreCase(patternChar) != 0) {
return false;
}
}
// End of pattern and text?
if (patternIndex + 1 >= pattern.length() && textIndex + 1 >= text.length()) {
return true;
}
// Go on to match the next character in the pattern
return matchCharacter(patternIndex + 1, textIndex + 1);
}
}
Similar to Tony Edgecombe's answer, here is a short and simple globber that supports * and ? without using regex, if anybody needs one.
public static boolean matches(String text, String glob) {
String rest = null;
int pos = glob.indexOf('*');
if (pos != -1) {
rest = glob.substring(pos + 1);
glob = glob.substring(0, pos);
}
if (glob.length() > text.length())
return false;
// handle the part up to the first *
for (int i = 0; i < glob.length(); i++)
if (glob.charAt(i) != '?'
&& !glob.substring(i, i + 1).equalsIgnoreCase(text.substring(i, i + 1)))
return false;
// recurse for the part after the first *, if any
if (rest == null) {
return glob.length() == text.length();
} else {
for (int i = glob.length(); i <= text.length(); i++) {
if (matches(text.substring(i), rest))
return true;
}
return false;
}
}
It may be a slightly hacky approach. I've figured it out from NIO2's Files.newDirectoryStream(Path dir, String glob) code. Pay attention that every match new Path object is created. So far I was able to test this only on Windows FS, however, I believe it should work on Unix as well.
// a file system hack to get a glob matching
PathMatcher matcher = ("*".equals(glob)) ? null
: FileSystems.getDefault().getPathMatcher("glob:" + glob);
if ("*".equals(glob) || matcher.matches(Paths.get(someName))) {
// do you stuff here
}
UPDATE Works on both - Mac and Linux.
The previous solution by Vincent Robert/dimo414 relies on Pattern.quote() being implemented in terms of \Q...\E, which is not documented in the API and therefore may not be the case for other/future Java implementations. The following solution removes that implementation dependency by escaping all occurrences of \E instead of using quote(). It also activates DOTALL mode ((?s)) in case the string to be matched contains newlines.
public static Pattern globToRegex(String glob)
{
return Pattern.compile(
"(?s)^\\Q" +
glob.replace("\\E", "\\E\\\\E\\Q")
.replace("*", "\\E.*\\Q")
.replace("?", "\\E.\\Q") +
"\\E$"
);
}
I don't know about a "standard" implementation, but I know of a sourceforge project released under the BSD license that implemented glob matching for files. It's implemented in one file, maybe you can adapt it for your requirements.
There is sun.nio.fs.Globs but it is not part of the public API.
You can use it indirectly via:
FileSystems.getDefault().getPathMatcher("glob:<myPattern>")
But it returns PathMatcher, which is inconvenient to work with. Since it can accept only Path as parameter (not File).
One possible option is to convert the PathMatcher to regex pattern (just call its 'toString()' method).
Another option is to use dedicated Glob library like glob-library-java.
Long ago I was doing a massive glob-driven text filtering so I've written a small piece of code (15 lines of code, no dependencies beyond JDK). It handles only '*' (was sufficient for me), but can be easily extended for '?'. It is several times faster than pre-compiled regexp, does not require any pre-compilation (essentially it is a string-vs-string comparison every time the pattern is matched).
Code:
public static boolean miniglob(String[] pattern, String line) {
if (pattern.length == 0) return line.isEmpty();
else if (pattern.length == 1) return line.equals(pattern[0]);
else {
if (!line.startsWith(pattern[0])) return false;
int idx = pattern[0].length();
for (int i = 1; i < pattern.length - 1; ++i) {
String patternTok = pattern[i];
int nextIdx = line.indexOf(patternTok, idx);
if (nextIdx < 0) return false;
else idx = nextIdx + patternTok.length();
}
if (!line.endsWith(pattern[pattern.length - 1])) return false;
return true;
}
}
Usage:
public static void main(String[] args) {
BufferedReader in = new BufferedReader(new InputStreamReader(System.in));
try {
// read from stdin space separated text and pattern
for (String input = in.readLine(); input != null; input = in.readLine()) {
String[] tokens = input.split(" ");
String line = tokens[0];
String[] pattern = tokens[1].split("\\*+", -1 /* want empty trailing token if any */);
// check matcher performance
long tm0 = System.currentTimeMillis();
for (int i = 0; i < 1000000; ++i) {
miniglob(pattern, line);
}
long tm1 = System.currentTimeMillis();
System.out.println("miniglob took " + (tm1-tm0) + " ms");
// check regexp performance
Pattern reptn = Pattern.compile(tokens[1].replace("*", ".*"));
Matcher mtchr = reptn.matcher(line);
tm0 = System.currentTimeMillis();
for (int i = 0; i < 1000000; ++i) {
mtchr.matches();
}
tm1 = System.currentTimeMillis();
System.out.println("regexp took " + (tm1-tm0) + " ms");
// check if miniglob worked correctly
if (miniglob(pattern, line)) {
System.out.println("+ >" + line);
}
else {
System.out.println("- >" + line);
}
}
} catch (IOException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
Copy/paste from here
By the way, it seems as if you did it the hard way in Perl
This does the trick in Perl:
my @files = glob("*.html")
# Or, if you prefer:
my @files = <*.html>