[cp-patches] Enabling iconv charset provider

[Mark Wielaard]

Hi,

This patch makes it easy to enable and experiment with the new iconv
charset provider.

2005-04-29  Sven de Marothy  <address@hidden>
            Mark Wielaard  <address@hidden>

       * java/nio/charset/Charset.java (defaultCharset): Use
       SystemProperties.
       (provider): Check gnu.classpath.nio.charset.provider.iconv system
       property and return the IconvProvider when set.

       * NEWS: Document new character encoder framework.

Committed.

Enabling the IconvProvider does show a few regressions with mauve so I
have not enabled it by default yet. These are probably all simple
non-standard charset alias names missing from iconv.

-PASS: gnu.testlet.java.lang.String.getBytes13: String.getBytes("UTF-16") 
(number 1)
-PASS: gnu.testlet.java.lang.String.getBytes13: String.getBytes("UnicodeBig") 
(number 1)
+FAIL: gnu.testlet.java.lang.String.getBytes13: String.getBytes("UTF-16") 
(number 1)
+FAIL: gnu.testlet.java.lang.String.getBytes13: String.getBytes("UnicodeBig") 
(number 1)
 PASS: gnu.testlet.java.lang.String.getBytes13: 
String.getBytes("UnicodeBigUnmarked") (number 1)
-PASS: gnu.testlet.java.lang.String.getBytes13: 
String.getBytes("UnicodeLittle") (number 1)
+FAIL: gnu.testlet.java.lang.String.getBytes13: 
String.getBytes("UnicodeLittle") (number 1)
-PASS: gnu.testlet.java.nio.charset.Charset.forName2: Charset 
'IsO_646.IRV:1991' supported (number 1)
+FAIL: gnu.testlet.java.nio.charset.Charset.forName2: Charset 
'IsO_646.IRV:1991' supported (number 1)

About the release. I have dragged into a fight with cvs checkouts
through external cvs with eclipse. That was a mistake. Eclipse won :{
And as it turned out this wasn't a regression since 0.14 also had an
issue here. Sorry for fighting windmills.

There is one issue I want to investigate and fix before the release with
FileInputStream not always throwing FileNotFoundException when
appropriate. After that it is just updating docs and release!

Cheers,

Mark

Index: NEWS
===================================================================
RCS file: /cvsroot/classpath/classpath/NEWS,v
retrieving revision 1.73
diff -u -r1.73 NEWS
--- NEWS        28 Apr 2005 21:36:09 -0000      1.73
+++ NEWS        29 Apr 2005 14:32:34 -0000
@@ -1,7 +1,37 @@
-New in release 0.15 ()
+New in release 0.15 (Apr 29, 2005)
+
+* The old character encoding framework (gnu.java.io.EncodingManager)
+has been replaced by a system based completely on nio.charset
+providers. Many converters have been added, both the io, lang and nio
+frameworks now use the same set of converters and the whole character
+stream framework (Readers and Writers) have been optimized. For some
+workloads this leads to 2x till 20x speedups.
+
+The default charsets supported are:
+
+  Cp424, Cp437, Cp737, Cp775, Cp850, Cp852, Cp855, Cp857, Cp860, Cp861,
+  Cp862, Cp863, Cp864, Cp865, Cp866, Cp869, Cp874, ISO_8859_1, ISO_8859_13,
+  ISO_8859_15, ISO_8859_2, ISO_8859_3, ISO_8859_4, ISO_8859_5, ISO_8859_6,
+  ISO_8859_7, ISO_8859_8, ISO_8859_9, KOI_8, MS874, MacCentralEurope,
+  MacCroatian, MacCyrillic, MacDingbat, MacGreek, MacIceland, MacRoman,
+  MacRomania, MacSymbol, MacThai, MacTurkish, US_ASCII, UTF_16, UTF_16BE,
+  UTF_16Decoder, UTF_16Encoder, UTF_16LE, UTF_8, UnicodeLittle, Windows1250,
+  Windows1251, Windows1252, Windows1253, Windows1254, Windows1255,
+  Windows1256, Windows1257, Windows1258.
+
+Many more encoding are supported through the new IconvProvider
+depending on the platform iconv support. GNU libiconv is recommended.
+The IconvProvider is currently not enabled by default. To enable it
+define the system property gnu.classpath.nio.charset.provider.iconv=true.
+Some runtimes might choose to enable this by default by setting it
+through VMSystemProperties. We would like to get feedback on whether
+enabling or disabling the IconvProvider by default results in the
+highest speedups.
 
 Runtime interface changes:
 
+* jni.h changed to better support more VMs; see VM integration guide
+  for details.
 * New --enable-default-toolkit option to configure can be used to set
   the fully qualified class name of the default AWT toolkit to use.
   If not given, the old default of gnu.java.awt.peerk.gtk.GtkToolkit
@@ -10,11 +40,6 @@
   "core" JNI libraries.  This is primarily useful if your VM can use the
   Gtk peers but not the core JNI libraries.
 
-VM Interface changes:
-
-* jni.h changed to better support more VMs; see VM integration guide
-  for details.
-
 Other changes:
 
 * Fixed compatibility problems in the Java Beans API which affected Eclipse's
Index: java/nio/charset/Charset.java
===================================================================
RCS file: /cvsroot/classpath/classpath/java/nio/charset/Charset.java,v
retrieving revision 1.20
diff -u -r1.20 Charset.java
--- java/nio/charset/Charset.java       29 Apr 2005 06:49:57 -0000      1.20
+++ java/nio/charset/Charset.java       29 Apr 2005 14:32:34 -0000
@@ -38,7 +38,10 @@
 
 package java.nio.charset;
 
+import gnu.classpath.SystemProperties;
+
 import gnu.java.nio.charset.Provider;
+import gnu.java.nio.charset.iconv.IconvProvider;
 
 import java.io.BufferedReader;
 import java.io.InputStreamReader;
@@ -129,7 +132,7 @@
     
     try 
       {
-       encoding = System.getProperty("file.encoding");
+       encoding = SystemProperties.getProperty("file.encoding");
       }
     catch(SecurityException e)
       {
@@ -231,6 +234,12 @@
 
   private static CharsetProvider provider()
   {
+    String useIconv = SystemProperties.getProperty
+      ("gnu.classpath.nio.charset.provider.iconv");
+
+    if (useIconv != null)
+      return IconvProvider.provider();
+
     return Provider.provider();
   }

signature.asc
Description: This is a digitally signed message part