Regular Expressions faster in Java ?..

[pawel]

I have made some comparision C# to Java RegularExpressi on. The problem was
to find out if the rule match some text.
Matching were done for precompiled regular expressions, in 100000 iterations
loop. Those loops were executed 11 times and average value of consumend time
was calculated. Below are codes for both classes.
And I found, that Java implementation is 2 to 5 times faster than C# (it
depends on complexity of expression).
Maybe my test were to simple? And Java made some optimisations, that the
code doesnt run (couse it really does nothing usefull)?

--
Pawe³

<<RegMatchTest. java>>
public class RegMatchTest
{

public static void main(String[] args) throws Exception
{
String pat[] = {"a*c?(d|f+) ", "g.*c?(r|d+ )"};
String word ;
long num = 100000;

File f = new File ("text.txt") ;

char buff[] = new char[(int)f.length()];
FileReader fr = new FileReader (f);
fr.read(buff);
word = new String (buff);
System.out.prin tln("Testing for "+num+" loops.");
long avgSum = 0 ;
for (int n = 0 ; n <= 10 ; ++n)
{
long t1 = System.currentT imeMillis();
new RegMatchTest(). Test1 (pat, word, num);
long t2 = System.currentT imeMillis();
System.out.prin tln("Elapsed time : " + (t2-t1) + " ms");
if (n > 0)
avgSum += t2-t1;
}
System.out.prin tln("\nAverage time : "+ (avgSum/10) +" ms");
}

boolean Test1 (String[] pat, String word, long len)
{
Pattern p[] = {Pattern.compil e(pat[0]),
Pattern.compile (pat[1])};

boolean b = false ;
for (int n = 0 ; n < len ; ++n)
{
Matcher m = p[n%2].matcher(word);
b = m.matches();
}
return b ;
}
<<class1.cs>>
class Class1
{
[STAThread]
static void Main(string[] args)
{
string[] pat = {@"a*c?(d|f+) ", @"g.*c?(r|d+)"} ;
string word ;
long num = 100000;

System.IO.Strea mReader tr = new System.IO.Strea mReader ("text.txt") ;
word = tr.ReadToEnd () ;

Console.WriteLi ne("Testing for "+num+" loops.");
long avgSum = 0 ;
for (int n = 0 ; n <= 10 ; ++n)
{
DateTime t1 = DateTime.Now;
new Class1().Test1 (pat, word, num);
DateTime t2 = DateTime.Now;
TimeSpan ts = t2 - t1 ;
Console.WriteLi ne("Elapsed time : " + (ts.TotalMillis econds) + " ms");
if (n > 0)
avgSum += (long)ts.TotalM illiseconds;
}
Console.WriteLi ne("\nAverage time : "+ (avgSum/10) +" ms");
}

bool Test1 (string[] pat, String word, long len)
{

Regex[] p = {new Regex (pat[0], RegexOptions.Co mpiled),
new Regex (pat[1], RegexOptions.Co mpiled)};
bool b = false ;
for (int n = 0 ; n < len ; ++n)
{
Match m = p[n%2].Match(word);
}
return b ;
}
}

[phoenix]

I noticed the same in the past. Regex seems to be poorly supported by C#.
They are real slow even when compiled. I heard there are people porting the
boost package to C# but haven't found it yet.

Yves

"pawel" <pa************ @interia.pl> schreef in bericht
news:ez******** ******@TK2MSFTN GP12.phx.gbl...

I have made some comparision C# to Java RegularExpressi on. The problem was
to find out if the rule match some text.
Matching were done for precompiled regular expressions, in 100000 iterations loop. Those loops were executed 11 times and average value of consumend time was calculated. Below are codes for both classes.
And I found, that Java implementation is 2 to 5 times faster than C# (it
depends on complexity of expression).
Maybe my test were to simple? And Java made some optimisations, that the
code doesnt run (couse it really does nothing usefull)?

--
Pawe³

<<RegMatchTest. java>>
public class RegMatchTest
{

public static void main(String[] args) throws Exception
{
String pat[] = {"a*c?(d|f+) ", "g.*c?(r|d+ )"};
String word ;
long num = 100000;

File f = new File ("text.txt") ;

char buff[] = new char[(int)f.length()];
FileReader fr = new FileReader (f);
fr.read(buff);
word = new String (buff);
System.out.prin tln("Testing for "+num+" loops.");
long avgSum = 0 ;
for (int n = 0 ; n <= 10 ; ++n)
{
long t1 = System.currentT imeMillis();
new RegMatchTest(). Test1 (pat, word, num);
long t2 = System.currentT imeMillis();
System.out.prin tln("Elapsed time : " + (t2-t1) + " ms");
if (n > 0)
avgSum += t2-t1;
}
System.out.prin tln("\nAverage time : "+ (avgSum/10) +" ms");
}

boolean Test1 (String[] pat, String word, long len)
{
Pattern p[] = {Pattern.compil e(pat[0]),
Pattern.compile (pat[1])};

boolean b = false ;
for (int n = 0 ; n < len ; ++n)
{
Matcher m = p[n%2].matcher(word);
b = m.matches();
}
return b ;
}
<<class1.cs>>
class Class1
{
[STAThread]
static void Main(string[] args)
{
string[] pat = {@"a*c?(d|f+) ", @"g.*c?(r|d+)"} ;
string word ;
long num = 100000;

System.IO.Strea mReader tr = new System.IO.Strea mReader ("text.txt") ;
word = tr.ReadToEnd () ;

Console.WriteLi ne("Testing for "+num+" loops.");
long avgSum = 0 ;
for (int n = 0 ; n <= 10 ; ++n)
{
DateTime t1 = DateTime.Now;
new Class1().Test1 (pat, word, num);
DateTime t2 = DateTime.Now;
TimeSpan ts = t2 - t1 ;
Console.WriteLi ne("Elapsed time : " + (ts.TotalMillis econds) + " ms");
if (n > 0)
avgSum += (long)ts.TotalM illiseconds;
}
Console.WriteLi ne("\nAverage time : "+ (avgSum/10) +" ms");
}

bool Test1 (string[] pat, String word, long len)
{

Regex[] p = {new Regex (pat[0], RegexOptions.Co mpiled),
new Regex (pat[1], RegexOptions.Co mpiled)};
bool b = false ;
for (int n = 0 ; n < len ; ++n)
{
Match m = p[n%2].Match(word);
}
return b ;
}
}