FWD » Core Development » Database

Constantin Asofiei wrote:

Ovidiu: please review 9763a

9763a/15767 only adds conversion support for buffer's xml-node-name attribute. But the implementation will defer to setting the attribute on parent table:

   public void BufferImpl.setXmlNodeName(String name)
   {
      buffer().getParentTable().setXmlNodeName(name);
   }

Also, the other attributes set to the buffer:

[ LABEL label-name ]
[ NAMESPACE-URI namespace ][ NAMESPACE-PREFIX prefix ]
[ XML-NODE-NAME node-name ][ SERIALIZE-NAME serialize-name ]

Constantin Asofiei wrote:

Eduard: please do some small tests and see if conversion of these works properly.

Conversion break for the following statment:

def temp-table tt1 field f1 as int.
def buffer btt1 for tt1 label "mybuffer" 
                        NAMESPACE-URI "buffer_namespace" 
                        NAMESPACE-PREFIX "my_prefix" 
                        xml-node-name "xbtt1" 
                        SERIALIZE-NAME "buf".

The converted output looks like this:

btt1.setXmlNodeName("xbtt1");
"mybuffer";
"buffer_namespace";
"my_prefix";
 "buf";

This will happen with label, NAMESPACE-URI, NAMESPACE-PREFIX and SERIALIZE-NAME when taken separately.

I don't see any label support in BufferImpl, so we may want to defer that.

Otherwise, please create a more complex test for the other options, and see if they are used in JSON or XML write/read. If so, please work on the conversion changes.

Constantin Asofiei wrote:

9763a/15767 only adds conversion support for buffer's xml-node-name attribute. But the implementation will defer to setting the attribute on parent table:

Yes, runtime needs to be fixed. Any concerns about the approach with the conversion?

Taking into consideration that the conversion is broken for these options of define buffer, I am puzzled how this issue surfaced only now. Apparently none of the existing customer projects uses this syntax. Only the latest customer does.

I think the prototype for conversion is correct.

From the conversion point of view, the buffers will have to keep a local copy of these attributes. When they are null / unknown (?), the will delegate the call to parent table, as we do now. This is the simple part. There are some other issues, related to serialisation. Most likely parts of TempTableSchema, XmlExport.java and JsonExport.java (not sure whether the import utilities, too) will have to adjust the accesses to these attributes via the provided BUFFER handle, not its TEMP-TABLE.

Eduard, so I found SERIALIZE-NAME also at the DEF BUFFER statement. Please work on adding all these. I don't know what LABEL is used for - can you find a test which shows how is used?

Constantin Asofiei wrote:

Eduard, so I found SERIALIZE-NAME also at the DEF BUFFER statement. Please work on adding all these. I don't know what LABEL is used for - can you find a test which shows how is used?

According to the docs it is used in error messages instead of the actual buffer name;

LABEL label-name
    Specifies a label for the buffer. This label is used in error messages in place of the buffer name.

def temp-table tt1 field f1 as int index idx1 is unique f1.
def buffer btt1 for tt1 label "my_buffer".

create btt1.
btt1.f1 = 10.
release btt1.

create btt1.
btt1.f1 = 10.
release btt1.

The following program will display the the error message ** my_buffer already exists with 10. (132).

Is this LABEL only for buffers, or temp-table, too? Is it accessible via BUFFER btt1:LABEL = "something"?

Constantin Asofiei wrote:

Is this LABEL only for buffers, or temp-table, too? Is it accessible via BUFFER btt1:LABEL = "something"?

Well in the docs for temp-table statement definition there is no such thing as LABEL.

This syntax BUFFER btt1:LABEL = "something" or message BUFFER btt1:LABEL will give the following warning ** LABEL is not queryable/ setable attribute for BUFFER widget. (4052)

Eduard Soltan wrote:

Constantin Asofiei wrote:

Is this LABEL only for buffers, or temp-table, too? Is it accessible via BUFFER btt1:LABEL = "something"?

Well in the docs for temp-table statement definition there is no such thing as LABEL.

However the following syntax in OE will not raise any errors def temp-table tt1 field f1 as int label "my_table".

Please post the test and how it converts now.

As LABEL is not an attribute for BUFFER or TEMP-TABLE (you can't set or get it via h:LABEL), then please do not use LabelAttribute or mark the methods with LegacyAttribute. Leave them only in Buffer.java, and also use UnimplementedFeature and for the getter just return unknown.

Constantin Asofiei wrote:

Please post the test and how it converts now.

The 4gl test:

def temp-table tt1 field f1 as int.
def buffer btt1 for tt1 label "mybuffer" 
                        NAMESPACE-URI "buffer_namespace" 
                        NAMESPACE-PREFIX "my_prefix" 
                        xml-node-name "xbtt1" 
                        SERIALIZE-NAME "buf".

Converted output is:

 btt1.setXmlNodeName("xbtt1");
 btt1.namespaceURI("buffer_namespace");
 btt1.namespacePrefix("my_prefix");
 btt1.setSerializeName("buf");
 btt1.setLabel("mybuffer");

Constantin Asofiei wrote:

As LABEL is not an attribute for BUFFER or TEMP-TABLE (you can't set or get it via h:LABEL), then please do not use LabelAttribute or mark the methods with LegacyAttribute. Leave them only in Buffer.java, and also use UnimplementedFeature and for the getter just return unknown.

Committed on 9763a, rev. 15770.

9763a rev 15771 changes return new character("label") to return character.UNKNOWN.

I'll move ahead with the conversion testing for this.

Otherwise, for the runtime parts, Eduard, please work in a second branch.

Added runtime support for buffer attributes, on 9763b, rev. 15771.

buffer btt1:write-xml("longchar", lc).

buffer will use the attributes in buffer definition.

temp-table btt1:write-xml("longchar", lc).

temp-table will use the attributes in temp-table definition.

Eduard Soltan wrote:

Added runtime support for buffer attributes, on 9763b, rev. 15771.

buffer btt1:write-xml("longchar", lc).

buffer will use the attributes in buffer definition.

temp-table btt1:write-xml("longchar", lc).

temp-table will use the attributes in temp-table definition.

Test also with buffer tt1:write-xml and buffer tt1:xml-node-name = "something".. I.e. use the default buffer instead of the temp-table definition. After that, check if this affects temp-table tt1:write-xml.

Constantin Asofiei wrote:

Test also with buffer tt1:write-xml and buffer tt1:xml-node-name = "something".. I.e. use the default buffer instead of the temp-table definition. After that, check if this affects temp-table tt1:write-xml.

temp-table definition sets the attributes of the default buffer.

temp-table tt1:write-xml and buffer tt1:write-xml will give the same result.

If I will try to modify some attributes of the default buffer with statement buffer tt1:xml-node-name = "something". Execution of temp-table btt1:write-xml("longchar", lc)., will use the use the updated version of xml-node-name.

Committed on 9763b, rev. 15772 this aspect.

I looked at 9763a and I do NOT agree with the implementation.

IMHO, the values of these new attributes should be stored at BufferImpl level, not in RecordBuffer. The set to null should only be ignored for the default buffer, not the others. The XmlExport.java should simply take the attribute from the buffer, not compute it locally.

At any rate, the changes in (only a new import for IOException) AbstractTempTable.java should be reverted.

Please use the following procedure for tests:

output to /tmp/9763.out.

def temp-table tt1 field f1 as int.
def buffer btt1 for tt1 xml-node-name "xbtt1".

message temp-table tt1:xml-node-name  temp-table btt1:xml-node-name. // tt1 tt1

create tt1.
tt1.f1 = 10.
release tt1.
def var lc as longchar.

buffer btt1:write-xml("longchar", lc).
message "~n" buffer btt1:xml-node-name string(lc). // xbtt1
buffer tt1:write-xml("longchar", lc).
message buffer tt1:xml-node-name string(lc). // tt1

buffer btt1:xml-node-name = ?.  // remove local copy
buffer btt1:write-xml("longchar", lc).
message "~n" buffer btt1:xml-node-name string(lc). // tt1 (from default)
buffer tt1:write-xml("longchar", lc).
message buffer tt1:xml-node-name string(lc). // tt1

buffer btt1:xml-node-name = "bla-bla".  // overwrite local copy
buffer btt1:write-xml("longchar", lc).
message "~n" buffer btt1:xml-node-name string(lc).   // bla-bla
buffer tt1:write-xml("longchar", lc).
message buffer tt1:xml-node-name string(lc).    // tt1

buffer tt1:xml-node-name = ?.       // no effect
temp-table tt1:xml-node-name = ?.   // no effect
buffer btt1:write-xml("longchar", lc).
message "~n" buffer btt1:xml-node-name string(lc). // bla-bla
buffer tt1:write-xml("longchar", lc).
message buffer tt1:xml-node-name string(lc). // tt1

Ovidiu Maxiniuc wrote:

I looked at 9763a and I do NOT agree with the implementation.

I assume you mean 9763b.

Constantin Asofiei wrote:

I assume you mean 9763b.

Yes, you are right. Sorry, I did not realise that there are 2 branches. I have not (yet) looked at 9763a.

I looked at 9763a.

• buffer_definitions.rules:
  • contains some hard tabs (#09, \t)
  • rules should use token names in lower-case: this.downPath(prog.KW_NAMESP_U), etc
  • I think we may have a problem with setXmlNodeName and setSerializeName. They should be replaced with xmlNodeName and serializeName to avoid collisions with property setters of the records.
    However, this will probably be deferred to another task.

Ovidiu, you are right. Eduard: for now, please rename setLabel and getLabel to label; this way we don't introduce new issues. setXmlNodeName and setSerializeName already are defined in Buffer.

I think we need to add some 'safety code' in FWD, so that if it finds a collision between a DMO getter/setter and Buffer methods, then FWD will abend. We need to 'fail early' if there are such cases.

Constantin Asofiei wrote:

Ovidiu, you are right. Eduard: for now, please rename setLabel and getLabel to label; this way we don't introduce new issues. setXmlNodeName and setSerializeName already are defined in Buffer.

I think a field cannot be name label in 4GL.

I think we need to add some 'safety code' in FWD, so that if it finds a collision between a DMO getter/setter and Buffer methods, then FWD will abend. We need to 'fail early' if there are such cases.

My suggestion - I think of some time - is to have a better distinction between the record and the buffer. Assuming the following procedure:

define temp-table tt001 serialize-name "x-data" label "xlabel" 
   field serialize-name as character.
define variable btt0 as handle.
btt0 = buffer tt001:handle.

create tt001.
buffer tt001:serialize-name = "d-data".
tt001.serialize-name = buffer tt001:serialize-name.
message tt001.serialize-name.

         tt001.create();
         buffer(tt001).unwrapNamedSerializable().setSerializeName(new character("d-data"));
         tt001.setSerializeName(buffer(tt001).unwrapNamedSerializable().getSerializeName());
         message((character) new FieldReference(tt001, "serializeName").getValue());

         tt001.create();
         buf(tt001).setSerializeName(new character("d-data"));
         tt001.setSerializeName(buf(tt001).getSerializeName());
         message((character) new FieldReference(tt001, "serializeName").getValue());

This new buf should do the unwrapping automatically for tables and buffers it would return the cast to a BufferImpl:

   public static BufferImpl buf(Buffer b) {
      return (BufferImpl) b;
   }

Is my thinking too simple? Please let me know if I am in error.

Constantin Asofiei wrote:

Ovidiu, you are right. Eduard: for now, please rename setLabel and getLabel to label; this way we don't introduce new issues. setXmlNodeName and setSerializeName already are defined in Buffer.

I think we need to add some 'safety code' in FWD, so that if it finds a collision between a DMO getter/setter and Buffer methods, then FWD will abend. We need to 'fail early' if there are such cases.

I removed the hard tabs, and renamed the token names to lowercase. Since the label is not a valid name for 4gl field, I assume I could leave setLabel and getLabel. Also in this case logic for temp-table tt1:label = "labe2" should also be modified, which could cause problems with other 4gl structures.

Did a conversion of the following programe, and it went propely:

DEFINE VARIABLE bh AS HANDLE  NO-UNDO.

def temp-table tt1 LABEL "label" 
                   NAMESPACE-URI "namespace_url" 
                   NAMESPACE-PREFIX "amespace-prefix" 
                   XML-NODE-NAME "xml_node_name" 
                   SERIALIZE-NAME "seraization_name" 
                   field f1 as int.

def buffer btt1 for tt1 LABEL "label1" 
                        NAMESPACE-URI "namespace_url1" 
                        NAMESPACE-PREFIX "namespace_prefix1" 
                        XML-NODE-NAME "xml_node_name1" 
                        SERIALIZE-NAME "seraization_name1".

buffer tt1:label = "1abel2".
temp-table tt1:label = "labe2".

message buffer tt1:label.

buffer tt1:xml-node-name = "xml-node-name2".
temp-table tt1:xml-node-name = "xml-node-name2".

message buffer tt1:xml-node-name.

buffer tt1:namespace-uri = "namespace_uri2".
temp-table tt1:namespace-uri = "namespace_uri2".

message buffer tt1:namespace-uri.

buffer tt1:namespace-prefix = "namespace-prefix2".
temp-table tt1:namespace-prefix = "namespace-prefix2".

message buffer tt1:namespace-prefix.

buffer tt1:serialize-name = "serialize-name2".
temp-table tt1:serialize-name = "serialize-name2".

message buffer tt1:serialize-name.

I think I will have converted Regression test project just to make sure it does not break.

Eduard, the fact that 4GL doesn't allow label as a field name doesn't mean is not possible for the DMO property name to get converted to getLabel and setLabel.

Constantin Asofiei wrote:

Eduard, the fact that 4GL doesn't allow label as a field name doesn't mean is not possible for the DMO property name to get converted to getLabel and setLabel.

I understand that. I put the change of label methods in rev. 15774.

I performed a conversion of Regression test project with 9763a, rev. 15774. And it went without any problems.

You are missing history entry in a file.

Constantin Asofiei wrote:

You are missing history entry in a file.

Added history entry.

9763a was merged into trunk revision 15785 and archived.

Main faindings regarding buffer attributes:

NAMESPACE-URI

1) for statement buffer btt1:NAMESPACE-URI, it will be used the namepace-uri of the buffer or empty character.

2) for statement temp-table btt1:NAMESPACE-URI, it will be used the namepace-uri of the default buffer or empty character.

NAMESPACE-PREFIX

1) for statement buffer btt1:NAMESPACE-PREFIX, it will be used the namepace-prefix of the buffer or empty character.

2) for statement temp-table btt1:NAMESPACE-PREFIX, it will be used the namepace-prefix of the default buffer or empty character.

XML-NODE-NAME

1) for statement buffer btt1:XML-NODE-NAME, it will be used the xml-node-name of the buffer. If xml-node-name is unknown, it will fallback to SERIALIZE-NAME of the buffer. And if serialization-name is null, it will fallback to the table name.

1) for statement temp-table btt1:XML-NODE-NAME, it will be used the xml-node-name of the default buffer. If xml-node-name is unknown, it will fallback to SERIALIZE-NAME of the default buffer. And if serialization-name is null, it will fallback to the table name.

SERIALIZE-NAME

1) for statement buffer btt1:SERIALIZE-NAME, it will be used the SERIALIZE-NAME of the buffer. If serialization-name is null, it will fallback to the table name.

2) for statement temp-table btt1:SERIALIZE-NAME, it will be used the SERIALIZE-NAME of the default buffer. If SERIALIZE-NAME is null, it will fallback to the table name.

Same applies to the write-xml operation:

1) buffer btt1:write-xml will use only the attributes of the buffer.
2) temp-table btt1:write-xml will only use the attributes of the default table.

Read-Xml operation¶

The only attributes of interest in read operation are xml-node-name and SERIALIZE-NAME.

- for statement buffer btt1:read-xml there are 3 cases, note the attributes are for buffer definitions:
1) xml-node-name is set, in this case it will search only for XML tags that math the value in xml-node-name.
2) xml-node-name not set and SERIALIZE-NAME set, in this case it will search only for XML tags that match the value in SERIALIZE-NAME.
3) xml-node-name not set and SERIALIZE-NAME not set, in this case it will search only for XML tags that match the value in SERIALIZE-NAME.
- same goes for temp-table btt1:read-xml, only the attributes are for temp-table definitions or default buffer.

For NAMESPACE-PREFIX and NAMESPACE-URI are in correct form in XML file, this does not have any effect on attributes of the buffer.

JSON¶

For JSON the only attribute of interests in SERIALIZE-NAME.

1) For buffer btt1:write-json it will be used the SERIALIZE-NAME of the buffer. If SERIALIZE-NAME is null, it will fallback to the table name.

2) For temp-table btt1:write-json it will be used the SERIALIZE-NAME of the default buffer. If SERIALIZE-NAME is null, it will fallback to the table name.

The same goes with read-xml.

I committed on 9763b, rev. 15792, changes to address the behaviour.

Eduard, isn't SERIALIZE-NAME used for JSON? So please try READ-JSON and WRITE-JSON.

Constantin Asofiei wrote:

Eduard, isn't SERIALIZE-NAME used for JSON? So please try READ-JSON and WRITE-JSON.

Yes, SERIALIZE-NAME is used for JSON. I updated my previous post with info of that.

Created task branch 9763c from trunk rev 15790. 15791 fixes a problem with an edge case (regression in trunk 15785), where a DEFINE BUFFER b FOR b. needs to be dropped completely, if 'b' default buffer is used before this statement, all while in an internal procedure.

Branch 9763c was merged to trunk rev 15800 and archived.

Eduard, is 9763b ready for review?

Review of 9763b/15792

• AbstractTempTable.java:
  • line 105: the H entry is a bit misleading/reversed. The implementation was actually delegated from temp-table to the default buffer;
  • line 2535: isn't the correct code is defaultBuffer.namespaceURI(uri);
  • lines 2660-2692: these method seem to be duplicates of 2695-2725? The file most likely fill not compile. At the same time I think we can implement the label accessors methods by raising error 4052:

          ErrorManager.recordOrShowError(4052, "LABEL", "setable" /*sic*/, LegacyResource.TEMP_TABLE + " widget");

    for the setter and

          ErrorManager.recordOrShowError(4052, "LABEL", "queryable", LegacyResource.TEMP_TABLE + " widget");
          return character.UNKNOWN;

    for the getter, instead of the UnimplementedFeature.
• BufferImpl.java:
  • line 4961: the javadoc mentions NAMESPACE-PREFIX but the method's name is namespaceURI();
  • all affected methods: why dropping of @Override annotation?
  • line 4968: comparing to the original code from temp-table, in case the namespaceURI is null / not initialised, an empty string should be returned. Have you tested and found otherwise?
  • line 4979: as above, assigning unknown/null value to NAMESPACE-URI should raise 4083, not only simply return:

          ErrorManager.recordOrShowError(4083, " NAMESPACE-URI", LegacyResource.TEMP_TABLE + " widget");

    Note that, at least in this case, the error is aware of the originating handle. When the buffer handle is used, the LegacyResource.BUFFER is used instead. That means that the test for null/unknown argument should be performed in AbstractTempTable before delegating the call to its default buffer. Please check the other attributes, as well.
  • lines 5036-5067: please see the notes on LABEL attribute from the TEMP-TABLE implementation above;
  • lines 5137-5162: the javadocs seem mangled/mixed;
  • line 5637: a redundant line between the javadoc and the method;
• RecordBuffer.java: the H entry can be dropped, there are no other changes in this file, anymore.

Ovidiu Maxiniuc wrote:

• lines 2660-2692: these method seem to be duplicates of 2695-2725? The file most likely fill not compile. At the same time I think we can implement the label accessors methods by raising error 4052:[...]for the setter and[...]for the getter, instead of the UnimplementedFeature.

The file will compile but an warning will be displayed.

• line 4968: comparing to the original code from temp-table, in case the namespaceURI is null / not initialised, an empty string should be returned. Have you tested and found otherwise?

Yes, for both temp-table and buffer (buffer btt1:NAMESPACE-URI. and btt1:NAMESPACE-URI.) will return ?.

• line 4979: as above, assigning unknown/null value to NAMESPACE-URI should raise 4083, not only simply return:[...]Note that, at least in this case, the error is aware of the originating handle. When the buffer handle is used, the LegacyResource.BUFFER is used instead. That means that the test for null/unknown argument should be performed in AbstractTempTable before delegating the call to its default buffer. Please check the other attributes, as well.

Modified just for NAMESPACE-PREFIX.

Fixed review issues in rev. 15793.

Also initialize to defaultBuffer in StaticTempTable before setting up its attributes.

Eduard, for some reasons I missed some files in #9763-47 (I think I pressed Ctrl+C in wrong console):

• JsonExport.java, JsonImport.java, XmlExport.java, and XmlImport.java:
  • missing the H entry (and copyright year to be updated);
  • line 884, 635, 718, 873, respectively: I do not think there is computation in these lines. The serializeName always uses the currentBuffer, in case isDefaultBuffer is true the tempTable will use its default buffer which is actually currentBuffer. So the original piece of code

          TempTable tempTable = buffer.getParentTable();
          String serializeName = tempTable.getSerializeName().toJavaType();

    should be simply replaced with

    String serializeName = ((Buffer) buffer.getDMOProxy()).getSerializeName().toJavaType();

    if that is the correct behaviour. And similar for rest of the files.
• Do not forget to add the #refid to the comment when you do the commit.

Finally, please check these test casestest cases

output to /tmp/9763.out.

def temp-table tt1 NAMESPACE-URI "namespace_url0" 
                   NAMESPACE-PREFIX "amespace-prefix0" 
                   XML-NODE-NAME "xml_node_name0" 
                   SERIALIZE-NAME "seraization_name0" 
                   LABEL "label" 
                   field f1 as int.   
define buffer btt1 for tt1.
create tt1. f1 = 1.
create tt1.

define variable xml as longchar.
define variable json as longchar.

message "Label:" temp-table tt1:Label  buffer tt1:Label  buffer btt1:Label.

message "NAMESPACE-URI:"    temp-table tt1:NAMESPACE-URI    buffer tt1:NAMESPACE-URI    buffer btt1:NAMESPACE-URI.
message "NAMESPACE-PREFIX:" temp-table tt1:NAMESPACE-PREFIX buffer tt1:NAMESPACE-PREFIX buffer btt1:NAMESPACE-PREFIX.
message "XML-NODE-NAME:"    temp-table tt1:XML-NODE-NAME    buffer tt1:XML-NODE-NAME    buffer btt1:XML-NODE-NAME.
message "SERIALIZE-NAME:"   temp-table tt1:SERIALIZE-NAME   buffer tt1:SERIALIZE-NAME   buffer btt1:SERIALIZE-NAME.
buffer  tt1:write-xml("longchar", xml). message string(xml).
buffer  tt1:write-json("longchar", json). message string(json).
buffer btt1:write-xml("longchar", xml). message string(xml).
buffer btt1:write-json("longchar", json). message string(json).

temp-table tt1:NAMESPACE-URI = "1".
buffer tt1:NAMESPACE-URI = "2".
buffer btt1:NAMESPACE-URI = "3".

temp-table tt1:NAMESPACE-PREFIX = "4".
buffer tt1:NAMESPACE-PREFIX = "5".
buffer btt1:NAMESPACE-PREFIX = "6".

temp-table tt1:XML-NODE-NAME = "7".
buffer tt1:XML-NODE-NAME = "8".
buffer btt1:XML-NODE-NAME = "9".

temp-table tt1:SERIALIZE-NAME = "10".
buffer tt1:SERIALIZE-NAME = "11".
buffer btt1:SERIALIZE-NAME = "12".

message "NAMESPACE-URI:"    temp-table tt1:NAMESPACE-URI    buffer tt1:NAMESPACE-URI    buffer btt1:NAMESPACE-URI.
message "NAMESPACE-PREFIX:" temp-table tt1:NAMESPACE-PREFIX buffer tt1:NAMESPACE-PREFIX buffer btt1:NAMESPACE-PREFIX.
message "XML-NODE-NAME:"    temp-table tt1:XML-NODE-NAME    buffer tt1:XML-NODE-NAME    buffer btt1:XML-NODE-NAME.
message "SERIALIZE-NAME:"   temp-table tt1:SERIALIZE-NAME   buffer tt1:SERIALIZE-NAME   buffer btt1:SERIALIZE-NAME.
buffer  tt1:write-xml("longchar", xml). message string(xml).
buffer  tt1:write-json("longchar", json). message string(json).
buffer btt1:write-xml("longchar", xml). message string(xml).
buffer btt1:write-json("longchar", json). message string(json).

temp-table tt1:NAMESPACE-URI = ?.
buffer tt1:NAMESPACE-URI = ?.
buffer btt1:NAMESPACE-URI = ?.

temp-table tt1:NAMESPACE-PREFIX = ?.
buffer tt1:NAMESPACE-PREFIX = ?.
buffer btt1:NAMESPACE-PREFIX = ?.

temp-table tt1:XML-NODE-NAME = ?.
buffer tt1:XML-NODE-NAME = ?.
buffer btt1:XML-NODE-NAME = ?.

temp-table tt1:SERIALIZE-NAME = ?.
buffer tt1:SERIALIZE-NAME = ?.
buffer btt1:SERIALIZE-NAME = ?.

message "NAMESPACE-URI:"    temp-table tt1:NAMESPACE-URI    buffer tt1:NAMESPACE-URI    buffer btt1:NAMESPACE-URI.
message "NAMESPACE-PREFIX:" temp-table tt1:NAMESPACE-PREFIX buffer tt1:NAMESPACE-PREFIX buffer btt1:NAMESPACE-PREFIX.
message "XML-NODE-NAME:"    temp-table tt1:XML-NODE-NAME    buffer tt1:XML-NODE-NAME    buffer btt1:XML-NODE-NAME.
message "SERIALIZE-NAME:"   temp-table tt1:SERIALIZE-NAME   buffer tt1:SERIALIZE-NAME   buffer btt1:SERIALIZE-NAME.
buffer  tt1:write-xml("longchar", xml). message string(xml).
buffer  tt1:write-json("longchar", json). message string(json).
buffer btt1:write-xml("longchar", xml). message string(xml).
buffer btt1:write-json("longchar", json). message string(json).

temp-table tt1:NAMESPACE-URI = "a".
temp-table tt1:NAMESPACE-PREFIX = "b".
temp-table tt1:XML-NODE-NAME = "c".
temp-table tt1:SERIALIZE-NAME = "d".

message "NAMESPACE-URI:"    temp-table tt1:NAMESPACE-URI    buffer tt1:NAMESPACE-URI    buffer btt1:NAMESPACE-URI.
message "NAMESPACE-PREFIX:" temp-table tt1:NAMESPACE-PREFIX buffer tt1:NAMESPACE-PREFIX buffer btt1:NAMESPACE-PREFIX.
message "XML-NODE-NAME:"    temp-table tt1:XML-NODE-NAME    buffer tt1:XML-NODE-NAME    buffer btt1:XML-NODE-NAME.
message "SERIALIZE-NAME:"   temp-table tt1:SERIALIZE-NAME   buffer tt1:SERIALIZE-NAME   buffer btt1:SERIALIZE-NAME.
buffer  tt1:write-xml("longchar", xml). message string(xml).
buffer  tt1:write-json("longchar", json). message string(json).
buffer btt1:write-xml("longchar", xml). message string(xml).
buffer btt1:write-json("longchar", json). message string(json).

buffer btt1:NAMESPACE-URI = "X".
buffer btt1:NAMESPACE-PREFIX = "Y".
buffer btt1:XML-NODE-NAME = "Z".
buffer btt1:SERIALIZE-NAME = "W".

message "NAMESPACE-URI:"    temp-table tt1:NAMESPACE-URI    buffer tt1:NAMESPACE-URI    buffer btt1:NAMESPACE-URI.
message "NAMESPACE-PREFIX:" temp-table tt1:NAMESPACE-PREFIX buffer tt1:NAMESPACE-PREFIX buffer btt1:NAMESPACE-PREFIX.
message "XML-NODE-NAME:"    temp-table tt1:XML-NODE-NAME    buffer tt1:XML-NODE-NAME    buffer btt1:XML-NODE-NAME.
message "SERIALIZE-NAME:"   temp-table tt1:SERIALIZE-NAME   buffer tt1:SERIALIZE-NAME   buffer btt1:SERIALIZE-NAME.
buffer  tt1:write-xml("longchar", xml). message string(xml).
buffer  tt1:write-json("longchar", json). message string(json).
buffer btt1:write-xml("longchar", xml). message string(xml).
buffer btt1:write-json("longchar", json). message string(json).

define buffer btt2 for tt1.
message buffer btt1:NAMESPACE-URI buffer btt2:NAMESPACE-PREFIX buffer btt2:XML-NODE-NAME buffer btt2:SERIALIZE-NAME.
buffer btt2:write-xml("longchar", xml). message string(xml).
buffer btt2:write-json("longchar", json). message string(json).

Compare the output from FWD and 4GL. I used output to file to be easier to compare (meld) their results.

Committed on 9763a, rev. 15796.
I modified the JsonImport, JsonExport, XmlExport and XmlImport, to use just the currentBuffer in getting the attributes.

I had to modify the namespaceUri methods not to generate a empty Java string when trying to set a character.UNKNOWN.
And the fwd output of the previous test matches completely the 4GL output.

I tested read-json and read-xml, and they are fine.

Also made some tests with datasets (define dataset dtt1 for btt1.), and noticed a few things:

- write-json and read-json for dataset are good, since for buffer write it will use JsonExport.serializeTempTable method that it also used in buffer:writeJson. And for

- read-json is good, since the JsonImport.readDataSetContent method will use SerializeName of the buffers.

- write-xml for dataset is not ok, some attributes of the temp-table are used (namespaceURI, namespacePrefix).

Ovidiu, why do we use if (buffer.getSerializeName().toStringMessage().equalsIgnoreCase(name) in XmlImport.importDataset, to match the XML token to a buffer? From what I tested xml-node-name always takes precedence.

Testing on another customer application was successful. No more errors encountered nor strange behaviors. ✅

The unit tests and smoke tests of a large GUI application went fine.✅

Also got some news that unit test of a very large application have a 5% increase rate when run 9763b. But it has to be checked with a new baseline.

Eduard Soltan wrote:

Ovidiu, why do we use if (buffer.getSerializeName().toStringMessage().equalsIgnoreCase(name) in XmlImport.importDataset, to match the XML token to a buffer? From what I tested xml-node-name always takes precedence.

I did a bit of digging and I think the reason is because the original implementation of getSerializeName() was returning the SERIALIZE-NAME attribute, only if it was expressly set. For that, the name of the buffer was checked in XmlImport. More recent code added this check right into getSerializeName():

return new character(serializeName == null || serializeName.trim().isEmpty() ? name : serializeName);

There is a regression in at dataset ddt1:read-xml, when has no buffer set.

In XmlImport.importDataset it go on if (numBuf == 0) branch and will create a TempTableBuilder.

ctx.tt = new TempTableBuilder();
String ttNsUri = reader.getNamespaceURI();
String ttNsPrefix = reader.getPrefix();
if (ttNsUri != null)
{
   ctx.tt.namespaceURI(ttNsUri);
}
if (ttNsPrefix != null)
{
   ctx.tt.namespacePrefix(ttNsPrefix);
}

The problem is that TempTableBuilder does not initialize the defaultBuffer. Hence setting the namespaceURI or namespacePrefix will result in NPE.

Eduard Soltan wrote:

The problem is that TempTableBuilder does not initialize the defaultBuffer. Hence setting the namespaceURI or namespacePrefix will result in NPE.

I am thinking to initialize the ctx.tt.tempTablePrepare and create the defaultBuffer, but this will result in a error because I don't have any fields added yet.

The AbstractTempTable has two specialised subclasses: StaticTempTable and TempTableBuilder (for dynamic temp-tables). In the former case case, the default buffer is created automatically (statically) even before the code of procedure that declares it starts to be executed. For the latter, the buffer is created only AFTER the temp-table is prepared. This is the case you are encountering. Only in this particular case (AbstractTempTable._dynamic() returns true), the attributes should be temporarily stored in TempTableBuilder until it is prepared and the default buffer created (see createDefaultBuffer()). That why it is called a 'builder', after all.

Eduard Soltan wrote:

I am thinking to initialize the ctx.tt.tempTablePrepare and create the defaultBuffer, but this will result in a error because I don't have any fields added yet.

No, the defaultBuffer should be initialised at that specific time. Therefore, the attributes should be accumulated in the builder and set all at once, when the buffer is created.

Ovidiu Maxiniuc wrote:

Only in this particular case (AbstractTempTable._dynamic() returns true), the attributes should be temporarily stored in TempTableBuilder until it is prepared and the default buffer created (see createDefaultBuffer()). That why it is called a 'builder', after all.

I was think about that, working now to add back the attributes to AbstractTempTable and will set them in AbstractTempTable only when defaultBuffer is null. And when the defualtBuffer is indeed created, will set the attributes.

Not in AbstractTempTable, just in TempTableBuilder.

LE: and the attributes will be set in TempTableBuilder.createDefaultBuffer(), as noted above, after initialisation of defaultBuffer.

Ovidiu Maxiniuc wrote:

Not in AbstractTempTable, just in TempTableBuilder.

But this implies to override all the setter and getter for the attributes in TempTableBuilder.

Indeed, it does. This is specific to dynamic temp-tables, therefore, this is the proper place to do it.

LE: the implementation of a setter would look like:

setAttribute(value) :
   if (!prepared)
      store-locally(value)
   else
      super.setAttribute(value)

Committed on 9763b, rev. 15832.
Sure thing, I added the method in TempTableBuilder, and the attributes will be set in tempTablePrepareImpl, the only place where createDefaultBuffer is called.

Good, that's the idea.

• methods like namespacePrefix(character prefix) do not need to be overridden. The code is identical as in super class;
• lines 2781-2804: why don't set the attributes directly on the default buffer?
• lines 1241, 1286: the code is difficult to read because the ternary operator is used with different layouts. Also the length of the line is too long. Try a solution would be to replace the first with an if (which is also consequent with the similar/neighbour methods);
• note: the way 4GL checks if a dynamic temp-table is 'ready' for use is the PREPARED attribute. Along the code we used _prepared() method, which is the actual unwrapped implementation. Testing with defaultBuffer == null will do, though.

I have made the modification in TempTableBuilder.

But also made some changes in XmlExport and XmlExport to match the write-xml and read-xml methods.

Some test wrote for testing of write-xml:

def temp-table tt1 NAMESPACE-URI "namespace_url0" 
                   NAMESPACE-PREFIX "amespace-prefix0" 
                   XML-NODE-NAME "xml_node_name0" 
                   SERIALIZE-NAME "seraization_name0" 
                   LABEL "label" 
                   BEFORE-TABLE before-table-name
                   field f1 as int.   
define buffer btt1 for tt1  NAMESPACE-URI "namespace_url01" 
                    NAMESPACE-PREFIX "amespace-prefix01" 
                   XML-NODE-NAME "xml_node_name01" 
                   SERIALIZE-NAME "seraization_name01" 

temp-table tt1:TRACKING-CHANGES = true.
create tt1. f1 = 1.
create tt1.

define variable xml as longchar.

define dataset dtt1 for tt1.

dataset dtt1:write-xml("longchar", xml).

message string(xml).

For write-xml:

def temp-table tt1 NAMESPACE-URI "namespace_url0" 
                   NAMESPACE-PREFIX "amespace-prefix0" 
                   XML-NODE-NAME "xml_node_name0" 
                   SERIALIZE-NAME "seraization_name0" 
                   LABEL "label" 
                   BEFORE-TABLE before-table-name
                   field f1 as int.   
define buffer btt1 for tt1  NAMESPACE-URI "namespace_url01" 
                    NAMESPACE-PREFIX "amespace-prefix01" 
                   XML-NODE-NAME "xml_node_name01" 
                   SERIALIZE-NAME "seraization_name01" 

temp-table tt1:TRACKING-CHANGES = true.
create tt1. f1 = 1.
create tt1.

DEFINE VARIABLE cSourceType             AS CHARACTER NO-UNDO.
DEFINE VARIABLE cReadMode               AS CHARACTER NO-UNDO.
DEFINE VARIABLE lOverrideDefaultMapping AS LOGICAL   NO-UNDO.
DEFINE VARIABLE cFile                   AS CHARACTER NO-UNDO.
DEFINE VARIABLE cSchemaLocation         AS CHARACTER NO-UNDO.
DEFINE VARIABLE cFieldTypeMapping       AS CHARACTER NO-UNDO.
DEFINE VARIABLE cVerifySchemaMode       AS CHARACTER NO-UNDO.
DEFINE VARIABLE lRetOK                  AS LOGICAL   NO-UNDO.

ASSIGN
  cSourceType             = "file" 
  cFile                   = "file.json" 
  cReadMode               = "empty" 
  cSchemaLocation         = "" 
  lOverrideDefaultMapping = ?
  cFieldTypeMapping       = ?
  cVerifySchemaMode       = ?.

lRetOK = dataset dtt1:READ-json(cSourceType, cFile, cReadMode).

for each btt1:
   message btt1.f1 .
end.

Eduard Soltan wrote:

I have made the modification in TempTableBuilder.

The setters use _prepared() and the getters defaultBuffer != null for same condition.

But also made some changes in XmlExport and XmlExport to match the write-xml and read-xml methods.

No objections except for copyright year.

Some test wrote for testing of write-xml:
[...]
For write-xml:
[...]

I see READ-json, I was expecting a read-xml based on the changes in XmlImport.java, right?

Ovidiu Maxiniuc wrote:

[...]

I see READ-json, I was expecting a read-xml based on the changes in XmlImport.java, right?

Sorry, meant to post the following test:

def temp-table tt1 NAMESPACE-URI "namespace_url0" 
                   NAMESPACE-PREFIX "amespace-prefix0" 
                   XML-NODE-NAME "xml_node_name0" 
                   SERIALIZE-NAME "seraization_name0" 
                   LABEL "label" 
                   BEFORE-TABLE before-table-name
                   field f1 as int.   
define buffer btt1 for tt1  NAMESPACE-URI "namespace_url01" 
                    NAMESPACE-PREFIX "amespace-prefix01" 
                   XML-NODE-NAME "xml_node_name01" 
                   SERIALIZE-NAME "seraization_name01" 

temp-table tt1:TRACKING-CHANGES = true.
create tt1. f1 = 1.
create tt1.

DEFINE VARIABLE cSourceType             AS CHARACTER NO-UNDO.
DEFINE VARIABLE cReadMode               AS CHARACTER NO-UNDO.
DEFINE VARIABLE lOverrideDefaultMapping AS LOGICAL   NO-UNDO.
DEFINE VARIABLE cFile                   AS CHARACTER NO-UNDO.
DEFINE VARIABLE cSchemaLocation         AS CHARACTER NO-UNDO.
DEFINE VARIABLE cFieldTypeMapping       AS CHARACTER NO-UNDO.
DEFINE VARIABLE cVerifySchemaMode       AS CHARACTER NO-UNDO.
DEFINE VARIABLE lRetOK                  AS LOGICAL   NO-UNDO.

ASSIGN
  cSourceType             = "file" 
  cFile                   = "file.xml" 
  cReadMode               = "empty" 
  cSchemaLocation         = "" 
  lOverrideDefaultMapping = ?
  cFieldTypeMapping       = ?
  cVerifySchemaMode       = ?.

lRetOK = dataset dtt1:read-xml(cSourceType, cFile, cReadMode, cSchemaLocation, lOverrideDefaultMapping, cFieldTypeMapping, cVerifySchemaMode).

for each btt1:
   message btt1.f1 .
end.

Test from the previous post I used to read-json and it works as expected.

Testing performed with rev. 15835:

- unit tests of a large customer application ✅
- harness ✅
- Chui regression testing ✅
- unit test and smoke tests of a large GUI application ✅
- smoke of a GUI application ✅

Rebased to latest trunk, rev. 15937. Did some quick smoke testing and unit tests of a large GUI application and it passed.

Do you mean 9763b? The current revision is 15835. Is it possible you pushed to different branch?

Ovidiu Maxiniuc wrote:

Do you mean 9763b? The current revision is 15835. Is it possible you pushed to different branch?

Yes it the 9763b branch, I run 9763b bzr push --overwrite ~/secure/code/p2j_repo/p2j/active/9763b, but it is taking a very long time.

Eduard Soltan wrote:

Yes it the 9763b branch, I run 9763b bzr push --overwrite ~/secure/code/p2j_repo/p2j/active/9763b, but it is taking a very long time.

Now the 9763b, is at revision 15949.

I had one more quick look over 9763b, r15949.

In BufferImpl.java, I think there is a rebase error, ** 139 ES 20250508 was overwritten by ** 139 ES 20250313. And there are a few usages of the character constructor which should have been replaced by character.of() static method. But since you have already tested the branch and it was a success, let's defer this for another time, do avoid the unnecessary risk.

In RecordBuffer.java, all code changes were removed and only H entry remained. Please revert this as well.

Side note:
Instead of ~/secure/code/p2j_repo/p2j/active/9763b, try using bzr+ssh://localhost:2224/opt/secure/code/p2j_repo/p2j/active/9763b. It is much faster. You can do this when checking out, as well. You only need to have your signature installed on your /.ssh folder on the dev server.

Fixed rebase issues in rev. 15950. Is additional testing necessary?

• BufferImpl - missing history number
• RecordBuffer - no actual change, use 'bzr revert' to get it to branch's trunk level

Ovidiu: any other concerns?

No, everything seems correct now.

Exporting of child buffer (in a dataset relation) is using the attributes from TempTableSchema, when it should switch to buffer attributes.

Committed on 9763c, rev. 16177.

Eduard, please merge after 10586a.