FWD » Core Development » Base Language

There are 2 patterns used that should be supported.

Both start with an assignment of RCODE-INFORMATION:FILENAME to some program name or class name.

Then they use:

• RCODE-INFORMATION:DB-REFERENCES which we could easily add as an annotation during conversion. It appears to be a list of databases in use for that program. The code that uses this checks that the database is connected (at runtime).
• RCODE-INFORMATION:MD5-VALUE which is used as described in #9439 (see that task for implementation ideas)

Greg Shah wrote:

Both start with an assignment of RCODE-INFORMATION:FILENAME to some program name or class name.

This is already supported.

• RCODE-INFORMATION:DB-REFERENCES which we could easily add as an annotation during conversion. It appears to be a list of databases in use for that program. The code that uses this checks that the database is connected (at runtime).

This is already supported.

• RCODE-INFORMATION:MD5-VALUE which is used as described in #9439 (see that task for implementation ideas)

This needs to be added.

The class needs to be read from the .jar. Look into ThinClient.loadWindowIcon, how that works, I assume something similar can be used to get a stream for the .class file.

See RcodeInfoOps.initFileInfo, this resolves the qualified Java class name, and the Class file.

Keep in mind that we don't want this attribute to be calculated always - only first time the MD5-VALUE is read.

Constantin Asofiei wrote:

The class needs to be read from the .jar. Look into ThinClient.loadWindowIcon, how that works, I assume something similar can be used to get a stream for the .class file.

I don't understand how this method loadWindowIcon from ThinClient applies to my problem. Is there a specific method I should use? I only need the absolute path to the root of the project. From there, I'm able to build the absolute path to the .class file which is in the build/lib or deploy/lib. And I also need the name of the .jar file.

Please try this example:

      String file = "com/goldencode/p2j/testcases/Test1.class";
      URL url = MultiClassLoader.getClassLoader().getResource(file);
      InputStream s = url.openStream();

I got it and it works. Thank you!

Committed my solution to 9442a revision 15653 and 15654.

• please update the copyright year in the changed files, i.e. 2017-2025
• do not set the history number if is not set, that was from a branch which was promoted to trunk at some time. Otherwise, we do fix history numbers if they are out-of-order
• why is getMD5Val returning BaseDataType? This is The return value is a 32 character hexadecimal number., to quote from the docs. If you want to return an unknown value, use new character().
• the @return needs to have the first javadoc text on the same line
• hexazecimal needs to be hexadecimal
• intermediate comments need to use // and not the /**/
• when reading from jar files, you need to use UNIX-style '/' separator - using File.separator it may not work on Windows
• there is a TODO in the catch (Exception e) of resolveMd5Value - at the least, please add WARNING logging of the exception
• getMD5Val() needs to compute md5Value on the first read:
  • initialize it to null in initFileInfo
  • if is null, call resolveMd5Value
• look for kw_md5_val in gaps/expressions.rules and change rw.rt_lvl_stub to rw.rt_lvl_full_restr - add a comment that the MD5 value is computed from the .class file, and not the .r file
• I don't recall exactly the cases where FWD emits (anon) inner classes in the converted code, but these need to be included in the MD5 value, too:
  • find the folder where the main .class file is located
  • look in the folder for files which have the main class package + simple name + dollar sign as prefix, like com/goldencode/testcases/Test1$1.class - i.e. the <simpleName>$ is a prefix of that file name.

Addendum: kw_md5_val needs to be registered in SignatureHelper, too.

Constantin Asofiei wrote:

Addendum: kw_md5_val needs to be registered in SignatureHelper, too.

It is necessary to add it if it represents an attribute?

Delia Mitric wrote:

Constantin Asofiei wrote:

Addendum: kw_md5_val needs to be registered in SignatureHelper, too.

It is necessary to add it if it represents an attribute?

You are right, that's an attribute, not a method. So is not needed to be added.

• expressions.rules needs a history entry and the copyright year change
• instead of LOG.warning(e.getMessage());, log both a message, the file and a full stacktrace, like this (untested):
  

  if (LOG.isLoggable(Level.WARNING)
  {
     LOG.warning("Error while trying to calculate the MD5-VALUE for file " + url, e);
  }
  

• in wa.info.fileName != null && wa.info.md5Value == null test, instead of wa.info.fileName != null you need to have wa.info != null && wa.info.initialized.

Otherwise, did you test resolveMd5Value with a .class having an inner class?

Constantin Asofiei wrote:

Otherwise, did you test resolveMd5Value with a .class having an inner class?

About this, I found at https://docs.progress.com/bundle/openedge-oo-abl-develop-applications-117/page/Class-definition-files-and-object-type-names.html that in 4GL we cannot have a class defined into another class. The problem is I didn't find any specific example for it, but I've tested it with a bit of an unorthodox method:
I added a procedure with a specific name (say test.p).
I manually added another procedure with the name prefixed by the name of the main procedure. I've done this to "mimic" an inner class and to see if the program takes it into account for calculating the md5 value.

Delia Mitric wrote:

Constantin Asofiei wrote:

Otherwise, did you test resolveMd5Value with a .class having an inner class?

About this, I found at https://docs.progress.com/bundle/openedge-oo-abl-develop-applications-117/page/Class-definition-files-and-object-type-names.html that in 4GL we cannot have a class defined into another class.

Yes, but FWD emits I think only a kind of triggers as inner classes, at this time.

The problem is I didn't find any specific example for it, but I've tested it with a bit of an unorthodox method:
I added a procedure with a specific name (say test.p).
I manually added another procedure with the name prefixed by the name of the main procedure. I've done this to "mimic" an inner class and to see if the program takes it into account for calculating the md5 value.

You meant you changed the Test.java file to have an inner class? That is a good way to check the code if it picks up the inner class or not.

Constantin Asofiei wrote:

You meant you changed the Test.java file to have an inner class? That is a good way to check the code if it picks up the inner class or not.

Using this test I found that compiled inner classes don't have a different path from that of main .class file. The "inner" compiled classes contents are just included in the main compiled file. This seems to be a pattern in FWD.
I don't think that it makes sense to treat them separately unless you found an example where inner classes are compiled in a separate file.

For this example:

package testing;

public class InnerTest
{
   public static void main(String[] args)
   {
      Runnable anon = new Runnable()
      {
         @Override
         public void run()
         {
         }
      };

      anon.run();
   }

   public static class OtherClass
   {

   }
}

-rw-rw---- 1 ca ca 543 Jan 30 09:35 'InnerTest$1.class'
-rw-rw---- 1 ca ca 357 Jan 30 09:35 'InnerTest$OtherClass.class'
-rw-rw---- 1 ca ca 667 Jan 30 09:35  InnerTest.class

Constantin Asofiei wrote:

For this example:
[...]
Java 8 compiles these files:
[...]

Yes, I know this is possible in Java, but in FWD it seems inner classes are not compiled this way.

Delia Mitric wrote:

Constantin Asofiei wrote:

For this example:
[...]
Java 8 compiles these files:
[...]

Yes, I know this is possible in Java, but in FWD it seems inner classes are not compiled this way.

I'm talking about what is possible in Java, to cover any case where FWD will emit in the converted Java file an inner Java class. See this example:

def var p1 as char.

on write of book do:
  message p1.
end.

Committed to 9442a revision 15656.

Tested with inner classes.

Delia Mitric wrote:

Committed to 9442a revision 15656.

The changes are OK, thank you.

Preparing to merge now.