FWD » Core Development » Database

Review of 7645a/15017.

So far, so good. I am OK with the revision, I like the splitting of the code in very specific methods.

• there are some unused imports (java.net.*, com.goldencode.util.*, org.w3c.dom.*);
• method saveMetaSetup() can be private (and moved down in the appropriate section). Unless there is there a reason to keep it public? Why rename it to saveMetaRecord, maybe?
• method populateDatabase():
  • please update the javadoc. It still say "information read from XML files stored in the application jar". This is not the case, any more;
  • it uses { at the end of try / for / while (Java standard). Please move them to next line (GCD code style). Same for method savePropertyMetaRecord();
  • the lists is too vague as an identifier. It might have been suggested by the IDE, but the human reader will understand better the code if a more suggestive name is used (for example: dmoInterfacesList);
  • typo in indecies identifier;
• methods saveMetaDbRecord(), saveFileMetaRecord(), savePropertyMetaRecord(), saveIndexMetaRecord(), saveIndexComponentMetaRecord(), and createMetaAux():
  • lines dmo.primaryKey(Long.parseLong(String.valueOf(recId))); look strange. Why not using a cast to long? Maybe use long as the type of recId in populateDatabase(), the only place where these methods are being called;
  • maybe change the names of the methods. they do not save anything, just create and initialize a record to be saved later by saveMetaSetup();
  • since the methods also initialize the record to be returned from the passed arguments (like Property, P2JIndex), this fact can be specified in the javadoc, too;
  • I think it would be more readable if the declaration and initialization of dmo is done by a single statement, on a single line;
• method saveAuxMetaRecords():
  • The multiple lookups in tableSet set can be replaced by the faster and more compact way line this:
    

          for (String tableName : tableSet)
          {
             switch (tableName)
             {
                case "_actindex": createMetaAux(p, ++recId, "Actindex"); break;
                case "_actlock":  createMetaAux(p, ++recId, "Actlock");  break;
                [...]
             }
          }
    

  • I guess each of these tables will have to be populated with specific values in the future, as the customers will need access to these tables, so this method will be updated at respective moments;
  • method prepareRecord(), fieldClassMap, PK_TYPE fields are not needed any more so they can be dropped;
• other things I identified while doing the review, not introduced by this update:
  • the SystemTable.forIface() has a String parameter. I wonder whether we can boost the performance by comparing the interfaces using = instead of String.equalsIgnoreCase(). In both cases where this method is called, the interface Class is available.
  • there is a typo with isManadatory() method. Please correct it.
• question: when the SystemTable._File is created, it uses MetadataManager::populateFileTable() as initializer. This will take each table from the database (method populateFileTable()) and use FileTableRecordBuilder.build() to build a record which will save (setMetaTableRecord()) similar records in _File, _Field and the other meta tables. It looks to me that these pieces of information are at lease overwritten if not duplicated.

There is only one affected file MetadataManager.java. We also need changes in conversion, so that the <schema>.meta.xml resource is not generated any more. Since the meta information is picked from the annotation, this resource is no longer necessary. Therefore I expect there will be a new revision which removes rules/metaschema.xml from FWD project and the code from ConversionDriver.middle() which generates metaschema XML. Please feel free to update the status of the task when these are ready.

The fixes are done. I also did a bit of cleanup and code maintenance.
Here is a bit of code which dumps the metadata content:

output to /tmp/7645.out.

for each _Db:
   message "D: " recid(_Db) _Db-name _Db-Type _Db-Comm _Db-Slave.
   for each _File of _Db:
      message "~tT: " recid(_File) _Db-recid _File-Name _File-Number _Template _Owner _User-Misc _File-Label _Tbl-Type _numfld _Prime-Index _Valmsg _Valexp _Dump-name.

      for each _Field of _File:
         message "~t~tF: " recid(_Field) _File-recid _Field-Name _Data-Type _Mandatory _Fld-case _Label _Col-label _Format _Order _Field-Physpos _Width.
      end.

      for each _Index of _File:
         message "~t~tI: " recid(_Index) _File-recid _Index-Name _Unique.

         for each _Index-Field of _Index:
            message "~t~t~tC: " recid(_Index-Field) _Index-recid  _Field-recid  _Index-Seq.
         end.
      end.

   end.
end.

It is focused on the items which are now processed from annotations instead of XML. I compared the output with and without the 7645a and they are identical up to the RECIDs of the records (note that they are used in PARENT-ID relations).

Committed revision 15054 (after rebase).

Although I am pretty confident of the code, it should be reviewed. Thank You!

Branch 7645a was merged into trunk as rev. 15070 and archived.

Reconversion is not mandatory. But if done the <db-name>.meta.xml files will not be created any more.
You need to update the build script to skip packaging the deprecated artefact (old locations: <pkgroot>/dmo/<db-name>/<db-name>.meta.xml). Otherwise the build script may end with file not found error.