On 13/12/06, Jean-Sebastien Delfino <[EMAIL PROTECTED]> wrote:
Pete Robbins wrote: > On 13/12/06, Pete Robbins <[EMAIL PROTECTED]> wrote: > >> OK... I've checked in a change for this. The result is that xsi:type is >> written correctly. Unfortunately the element prefix is not always >> written >> correctly. The problem is that we do not record the namespace uri of a >> property so when we come to write the graph to xml we are, at the >> moment, >> assuming that the property element namespace uri is the same as the >> uri of >> the property type. This is not always the case as in Sebastien's example >> above. >> >> I think what is there now will work for Caroline and also for Sebastien >> (though the xml is actually incorrect). >> >> A perfect fix will need more thought and changes in schema parsing, >> document loading and document writing. It is possible to have an >> improved >> fix without too much rework where we only write out the element >> prefix after >> validating that a global element of the property name was defined >> (Property >> on the "RootType" of the namespace) exists. So for the examples in this >> thread: >> >> >> - return would be written as <return .. > because there is no >> property named "return" on the Type >> http://www.bigbank.com/AccountService#RootType. >> >> - address would be written as <tns2:address ... > because "address" >> is a property (of the correct Type) on >> >> http://schemas.xmlsoap.org/wsdl/soap/< http://schemas.xmlsoap.org/wsdl/soap/#RootType>< http://schemas.xmlsoap.org/wsdl/soap/> >> >> #RootType >> >> >> I'll look at this again tomorrow. >> >> Cheers, >> >> > > > > I've checked in another update which should write the element prefix > correctly. I think there are still cases where we should write > prefixes on > attributes as well. > > I also fixed the fact that Sequence elements were written as the property > name and not the element name. There is a similar problem when writing > references which I'll fix in the next update. > > Cheers, > > > >> >> On 12/12/06, Pete Robbins <[EMAIL PROTECTED]> wrote: >> > >> > >> > >> > On 12/12/06, Jean-Sebastien Delfino < [EMAIL PROTECTED]> wrote: >> > > >> > > Pete Robbins wrote: >> > > > On 11/12/06, Jean-Sebastien Delfino < [EMAIL PROTECTED]> wrote: >> > > >> >> > > >> With the SDO head (r485883) the XML response from the >> > > AccountDataService >> > > >> in the HttpdBigBank scenario is incorrect (and breaks the >> > > scenario): >> > > >> >> > > >> <getStockAccountResponse xmlns="http://tempuri.org"; >> > > >> xmlns:xsi=" http://www.w3.org/2001/XMLSchema-instance " >> > > >> xmlns:tns="http://tempuri.org"; >> > > >> xmlns:tns2=" http://www.bigbank.com/AccountService";><tns2:return >> > > >> >> > > >> >> > > >> xsi:type="StockAccount"><accountNumber>1234_STA12345</accountNumber><symbol>IBM</symbol><quantity>100</quantity><balance></balance></tns2:return></getStockAccountResponse> >> >> > > >> > > >> >> > > >> >> > > >> - xsi:type="StockAccount" is missing the tns2 namespace prefix. >> > > >> - return is not from the tns2 namespace and should not be >> prefixed >> > > by >> > > >> tns2. >> > > >> >> > > >> >> > > >> Going back to an older revision of SDOXMLWriter.cpp (r480964) >> I am >> > > >> getting the correct XML: >> > > >> >> > > >> <getStockAccountResponse xmlns=" http://tempuri.org"; >> > > >> xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance " >> > > >> xmlns:tns="http://tempuri.org " >> > > >> xmlns:tns2=" http://www.bigbank.com/AccountService";><return >> > > >> >> > > >> >> > > >> xsi:type="tns2:StockAccount"><accountNumber>1234_STA12345</accountNumber><symbol>IBM</symbol><quantity>100</quantity><balance></balance></return></getStockAccountResponse> >> >> > > >> > > >> >> > > >> >> > > >> I have stepped through the SDO runtime code and I'm not sure yet >> > > why >> > > >> this is happening, but this is caused by some of the recent >> changes >> > > to >> > > >> SDOXMLWriter.cpp. Any idea? >> > > >> >> > > >> -- >> > > >> Jean-Sebastien >> > > > >> > > > >> > > > What namespace is the element <return> supposed to be in? In the >> > > first >> > > > it is >> > > > in "tns2" and in the second the default namespace >> http://tempuri.org >> > > . >> > > > What >> > > > Types/Properties are defined in the DataFactory? >> > > > >> > > > I'll take a look at this. There have been a number of "fixes" in >> > > this >> > > > area >> > > > and each change breaks something else! >> > > > >> > > > Cheers, >> > > > >> > > Pete, >> > > >> > > In my example, checkingAccountResponse is open content so the >> <return> >> > > element is not defined by a schema, but the instance of StockAccount >> > > is >> > > from tns2. >> > > >> > > -- >> > > Jean-Sebastien >> > >> > >> > >> > OK. I know how to fix this. Sebastien, you put some code in to >> check for >> > an open content property and not write the element uri. This caused a >> > failure that Caroline saw so we backed out that change. >> > >> > The problem is actually that your check at the top of the writeDO >> method >> > was using the *element* name and not the *property* name. The name of >> > the element to be written is passed into writeDO, not the name of the >> > property that is being processed. These are often the same, but are >> not the >> > same in cases where substitutions, aliases or sdo:name annotations >> are used. >> > >> > >> > I'm on the case and hope to have a fix in the next couple of hours. >> > >> > Cheers, >> > >> > -- >> > Pete >> > >> >> >> >> -- >> Pete > > > > Pete, I'm afraid that a number of scenarios are broken with the latest SDO changes (I just tried the SDO head r486738): - PythonWeatherForecast - HttpdBigBank - RestCalculator - RestCustomer Going back to r480964 of SDOXMLWriter.cpp they all work.
That's a long time ago! 485600 is the revision before I made these changes. Could you test against that so we can eliminate the other changes. We need to add some of these more complex tests into the SDO testing. Can you pin down exactly what is failing? Cheers, -- Pete
