Write up a response flow handler that gets the response message from the MessageContext object and removes the xsi:nil = true elements.
-----Original Message----- From: Tim R J Langford [mailto:[EMAIL PROTECTED] Sent: Wednesday, February 15, 2006 9:42 AM To: [email protected] Subject: Re: How to not send nillable="true" for null elements in a SOAP message Hello All, I am still trying to remove 'nil="true"' elements from my messages sent out by axis. I have modified the wsdl in elements where I want axis not to send these strings (i.e: have a "blank" element) as follows : <element name="sendNothingIfNull" nillable="true" type="xsd:string"/> to : <element name="sendNothingIfNull" nillable="false" minOccurs="0" type="xsd:string"/> so that instead of sending : <someXML> <sendNothingIfNull xsi:type="xsd:string" xsi:nil="true"/> </someXML> I send something like : <someXML></someXML> I have tried a few other methods (such as the one Rod suggested not specifying "nillable" at all in the schema and setting minOccurs="0") but axis keeps sending these elements as nillable. Should this work or am I making a mistake somwhere? It may be PEBCAK but I just cant get axis1.3 to exhibit the behavior the server expects. Is there any other lower level method in axis to deal with this? (I traced the code calls down to the HTTP sender class but even at this level there is no easy access to the string as it is being passed by chunk encoded HTTP - and of course in some cases the server we are trying to attach too actually wants us to send these nillable="true" elements so this approach is not suitable). I know I must have exhausted all you guys by now! Sorry, and thanks. Tim PS: I am generating the axis code form the wsdl using the following ant task configuration : <axis-wsdl2java output="${src.dir}" deployScope="Application" verbose="true" serverSide="true" wrapArrays="true" url="eurostar_hack2.wsdl"> </axis-wsdl2java> Rodrigo Ruiz wrote: > By what you describe, I think you should add minOccurs='0' and remove > nillable='true' > > Regards, > Rodrigo Ruiz > > Tim R J Langford wrote: > >> Thanks Anne, >> >> Thats what I thought. Unfortunatly our provider does not seem too >> clued up on their tech, and the wsdl does not allow minOccurs="0". I >> guess I will have to update our automated build process to fix their >> schema before generating my soap beans and classes. The best way to >> do this would be by adding minOccurs="0" to the faulty elements I >> presume? >> >> Thanks to everybody for their time, >> >> Tim >> >> >> Anne Thomas Manes wrote: >> >>> If the service cannot accept xsi:nil="true" then the service >>> provider should adjust the schema accordingly. Does the schema allow >>> minOccurs="0"? If not, then there's no valid way to send no element >>> instead of xsi:nil="true". >>> >>> Anne >>> >>> On 2/14/06, *Tim R J Langford* <[EMAIL PROTECTED] >>> <mailto:[EMAIL PROTECTED]>> wrote: >>> >>> >>> Hi Jeff, >>> >>> Thanks for your response. >>> >>> Sorry for being unclear. Yes you are correct. I am sending >>> 'xsi:nil="true"' and I want to configure axis to send nothing >>> instead >>> of this string. e.g : * >>> >>> <XXX* xsi:type="xsd:string" xsi:nil="true"/>* >>> <YYY* xsi:type="zzz:YYY" xsi:nil="true" xmlns:zzz="KKK"*/> >>> >>> *I think Axis 1.2 did it this way? Is there anyway of configuring >>> Axis >>> 1.3 to not send these 'xsi:nil="true"' elements? >>> >>> Thanks for you time, >>> >>> Tim >>> >>> >>> PS: The service wsdl does have 'xsd:nillable="true"' elements >>> in the >>> message schema, but their system cannot actually handle the >>> situation >>> where it is null (even if they return it in a response), and >>> they have >>> asked us to remove the 'xsi:nil="true"' elements from our >>> requests. We >>> could fix the wsdl schema, but this would impede our codegen >>> system as >>> we are the client, so were wondering if there was a way to do it >>> from >>> within axis?* >>> >>> >>> >>> * >>> >>> Jeff Greif wrote: >>> >>> >Just to be sure, you're sending xsi:nil="true", not >>> xsd:nillable="true", right? >>> > >>> >The latter is used only in the schema, and means that the >>> element is >>> >allowed to have no content. The former means that this >>> particular use >>> >of the element has no content. >>> > >>> >Jeff >>> > >>> >On 2/14/06, Tim R J Langford <[EMAIL PROTECTED] >>> <mailto:[EMAIL PROTECTED]>> wrote: >>> > >>> > >>> >>Hello All, >>> >> >>> >>I am writing a client interface into a provider SOAP web >>> service, and >>> >>their system fails and returns a null pointer exception when I >>> send them >>> >>a 'nillable="true"' element in my request. I think the reason >>> for this >>> >>is that they are using an older version of axis than we are >>> using ( 1.3). >>> >> >>> >> >>> > >>> > >>> > >>> >>> >> >> >> >
