FWD » Core Development » Database

There is another case where there is a field f1 serialize-name "oa:f1", and the XML has the field like f1 - in this and only this case, it will be picked up even if the XML file doesn't have the oa namespace.

I have a fix for this and the issue reported by Eduard, but needs more testing. I'll create and commit it to 10560a.

Eduard, see 10560a rev 16164 for my current changes. Please try to break them (see the TODO in XmlImport) by creating misc relations.

I have created the following dataset structure, one thing to note that Second table has a relationship with with ID and oaID table (serialization-name ID) :

DEFINE TEMP-TABLE Main NO-UNDO
    FIELD first_id  AS RECID XML-NODE-TYPE "HIDDEN":U.

DEFINE TEMP-TABLE Second NO-UNDO NAMESPACE-URI "http://www.openapplications.org/oagis/10":U
    FIELD Main_id AS RECID XML-NODE-TYPE "HIDDEN":U.

DEFINE TEMP-TABLE Third NO-UNDO NAMESPACE-URI "http://www.openapplications.org/oagis/10":U
    FIELD Main_id AS RECID XML-NODE-TYPE "HIDDEN":U.

DEFINE TEMP-TABLE ID NO-UNDO
    FIELD ID_Text1        AS CHARACTER  XML-NODE-TYPE "TEXT":U
    FIELD Second_id       AS RECID      XML-NODE-TYPE "HIDDEN":U
    FIELD Third_id        AS RECID      XML-NODE-TYPE "HIDDEN":U.

DEFINE TEMP-TABLE oaID NO-UNDO NAMESPACE-URI "http://www.openapplications.org/oagis/10":U SERIALIZE-NAME "ID":U
         FIELD ID_Text1    AS CHARACTER  XML-NODE-TYPE "TEXT":U
         FIELD Second_id   AS RECID      XML-NODE-TYPE "HIDDEN":U
         FIELD Third_id    AS RECID      XML-NODE-TYPE "HIDDEN":U.

DEFINE DATASET dsProcessField  XML-NODE-TYPE "HIDDEN":U
    FOR Main, Second, Third, ID, oaID
    PARENT-ID-RELATION FOR Main, Second PARENT-ID-FIELD Main_id
    PARENT-ID-RELATION FOR Main, Third  PARENT-ID-FIELD Main_id
    PARENT-ID-RELATION FOR Second, ID    PARENT-ID-FIELD Second_id
    PARENT-ID-RELATION FOR Second, oaID PARENT-ID-FIELD Second_id
    PARENT-ID-RELATION FOR Third, ID    PARENT-ID-FIELD Third_id.

Exported the following structure into an XML:

create Main.

create Second.
Second.Main_id = RECID(Main).
create Third.
Third.Main_id = recid(Main).

create oaID.
oaID.ID_Text1 = 'text1'.
oaID.Second_id = recid(Second).
release oaID.

create ID.
ID.ID_Text1 = 'text2'.
ID.Second_id = recid(Second).
release ID.

create ID.
ID.ID_Text1 = 'text3'.
ID.Third_id = recid(Third).
release ID.

release Second.
release Third.
release Main.

And got the following result:

<?xml version="1.0"?>
<Main xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
  <Second xmlns="http://www.openapplications.org/oagis/10">
    <ID>text1</ID>
    <ID>text2</ID>
  </Second>
  <Third xmlns="http://www.openapplications.org/oagis/10">
    <ID>text3</ID>
  </Third>
</Main>

Later when trying to load the same XML into the dataset, both <ID> nodes load into a single table. It chooses the first relationship that matches the serialization-name of the XML tag. And the second table that has the same serialization-name, but it's relationship is just ignored.

I think I found a problem in rev. 16164.

In XmlImport.resolveTableByXmlName the current context should be used as opposed to the context saved on stack ImportContext ctx = peekContext();

Committed on 10560a, rev. 16165.

On XmlImport.resolveTableByXmlName, I don't think is right to use the ctx.ttname, because the value is the local name for the current XML node and the relation.getParentTable() contains the legacy name.
I suggest changing it to ctx.tt._name(). This fixes some unit tests.

Should I commit the change from #10560-8?

Teodor Gorghe wrote:

Should I commit the change from #10560-8?

Go ahead.

Teodor Gorghe wrote:

Committed revision 16166 on task branch 16166:
- Resolving the parent buffer by its legacy name instead of the XML local name.

What task branch is this in?

Oops, this commit is in this task, 10560a.

I've committed my changes to 10560a rev 16167 - use a context in 'resolveTableByXmlName'.

I have a test case which does not work with 10352a, trunk or this branch: ABL CodeABL Code

BLOCK-LEVEL ON ERROR UNDO, THROW.

USING OpenEdge.Core.Assert FROM PROPATH.

CLASS tests.TestReadXmlPID:
    DEFINE TEMP-TABLE tt1
        FIELD f1 AS INTEGER
        FIELD Name AS CHARACTER.

    DEFINE TEMP-TABLE Name
        FIELD f1 AS INTEGER
        FIELD parentId AS RECID.

    DEFINE DATASET ds1 FOR tt1, Name
        PARENT-ID-RELATION REL1 FOR tt1, NAME PARENT-ID-FIELD parentId.

    METHOD PUBLIC VOID PopulateDataset():
        CREATE tt1.
        ASSIGN
            tt1.f1 = 1
            tt1.Name = "Name".

        CREATE Name.
        ASSIGN
            Name.f1 = 2
            Name.parentId = RECID(tt1).
    END METHOD.

    @Test.
    METHOD PUBLIC VOID TestReadXml():
        DEFINE VARIABLE cOutput AS LONGCHAR NO-UNDO.
        PopulateDataset().
        DATASET ds1:WRITE-XML ("longchar", cOutput).

        DATASET ds1:READ-XML ("longchar", cOutput, "EMPTY", ?, ?).

        FIND FIRST tt1 NO-ERROR.
        Assert:IsTrue(AVAILABLE tt1, "AVAILABLE tt1").
        Assert:Equals("Name", tt1.Name).
        Assert:Equals(1, tt1.f1).

        FIND FIRST Name NO-ERROR.
        Assert:IsTrue(AVAILABLE Name, "AVAILABLE Name").
        Assert:Equals(2, Name.f1).
    END METHOD.
END CLASS.

Constantin, should I create a new task or work here/on 10352a?

Work it on this branch.

For the test case from #10560-16, 4GL passes, but FWD fails because the name field is empty.

I am currently exploiting all hidden behavior from 4GL.

1. When the parentId field has XML-NODE-TYPE to HIDDEN, 4GL acts like FWD. The name field is empty and there is no Name record.
2. When the f1 field is a type of TEXT, the serialized XML gets something like that: XMLXML

   <?xml version="1.0"?>
   <ds1 xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
     <tt1>
       <f1>1</f1>
       <Name>Name</Name>
       <Name>2</Name>
     </tt1>
   </ds1>
   

   
   When this is read with READ-XML, the name field from tt1 table is written twice (same as FWD).

Later update:
- And when the Name temp-table has a structure, like this NameName

    DEFINE TEMP-TABLE Name
        FIELD f1 AS INTEGER XML-NODE-TYPE "TEXT" 
        FIELD attrName AS CHARACTER XML-NODE-TYPE "ATTRIBUTE" 
        FIELD parentId AS RECID XML-NODE-TYPE "HIDDEN".

Teodor: please make sure to run your tests in a later version than OE 11.6 - especially when you find 'weird quirks'.

I know that this behavior also happens for OE 11.7. Do we have another OE version available?

I have tested with OE 12.8 and I can confirm that everything that happens on OE 11.6 is the same as OE 12.8.

Committed revision 16168 on task branch 10560a:
- Added support on DATASET:READ-XML for matching XML field and temp-table with same serializable name/XML node name.

Constantin, the work is done, but I am not very confident on my changes, might be some regressions. I am starting to do some testing, but I have committed because I need a review.

Committed revision 16169 on task branch 10560a:
- Added some missing javadoc description. Added one more case which 4GL treats as a temp-table node.

• please make it compile in Java 8 - see XmlImport line 3949, the .toList() call;
• update the copyright year
• LookAheadXmlStreamReader
  • not all methods use the replay stack - see isStandalone(), isWhiteSpace() and others
  • move the defined classes to the end of the file

Committed revision 16170 on task branch 10560a:
- Addressed code review #10560-27.

Committed revision 16171 on task branch 10560a:
- Replaced ctx.ttName with ctx.tt._name().

Teodor: please rebase 10560a.

Rebased task branch 10560a to trunk revision 16250.

Rebased task branch 10560a to trunk revision 16358.