You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@thrift.apache.org by "Ben Taitelbaum (JIRA)" <ji...@apache.org> on 2009/11/13 05:01:39 UTC

[jira] Created: (THRIFT-627) should c++ have setters for optional fields?

should c++ have setters for optional fields?
--------------------------------------------

                 Key: THRIFT-627
                 URL: https://issues.apache.org/jira/browse/THRIFT-627
             Project: Thrift
          Issue Type: Improvement
          Components: Compiler (C++)
         Environment: c++
            Reporter: Ben Taitelbaum


It seems non-intuitive to me to have to set __isset.someField = true after setting an optional field someField on a struct. Would it make sense to have a set_someField method that would both set the field and modify __isset?

One of the cases for this is for when a field goes from being required to being optional, and it's easy to forget to set __isset in the code.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Commented: (THRIFT-627) should c++ have setters for optional fields?

Posted by "David Reiss (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/THRIFT-627?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12777374#action_12777374 ] 

David Reiss commented on THRIFT-627:
------------------------------------

I see no downside to generating these methods.

> should c++ have setters for optional fields?
> --------------------------------------------
>
>                 Key: THRIFT-627
>                 URL: https://issues.apache.org/jira/browse/THRIFT-627
>             Project: Thrift
>          Issue Type: Improvement
>          Components: Compiler (C++)
>         Environment: c++
>            Reporter: Ben Taitelbaum
>
> It seems non-intuitive to me to have to set __isset.someField = true after setting an optional field someField on a struct. Would it make sense to have a set_someField method that would both set the field and modify __isset?
> One of the cases for this is for when a field goes from being required to being optional, and it's easy to forget to set __isset in the code.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] Commented: (THRIFT-627) should c++ have setters for optional fields?

Posted by "Rush Manbert (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/THRIFT-627?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12777554#action_12777554 ] 

Rush Manbert commented on THRIFT-627:
-------------------------------------

Would you generate some equivalent for fields that are STL objects? In that case you can't write set_MyMap (value), but you could provide something like wrote_MyMap(void) that would take care of the __isset.

In general, setters would be useful, but they only solve the __isset problem halfway.

I have recently written some classes that I use to write wrappers for Thrift-generated C++ classes. The main point is to handle the __isset setting, especially in the case where I write a data member in a class that is contained within a parent class. If you write a data member of the contained class you need to write its __isset member, but you also need to write the __isset member in the parent class of the contained class. (parent in the ownership-sense, not inheritance) This needs to propagate up the ownership chain.

For STL objects, all I offer is a wasWritten(void) method that handles the __isset chain.

I believe that I could write a script that would read through my Thrift configuration files and auto-generate the wrappers, but it would be a lot better if the compiler would optionally do it for me. I know this is a lot more ambitious than adding setters to the currently generated classes, but it really helped me get out of __isset hell with the latest service that I wrote. While I'm sure that they can be improved, I would be happy to contribute the wrapper base classes. Is there any interest, or should I just shut up and keep this to myself? :-)

> should c++ have setters for optional fields?
> --------------------------------------------
>
>                 Key: THRIFT-627
>                 URL: https://issues.apache.org/jira/browse/THRIFT-627
>             Project: Thrift
>          Issue Type: Improvement
>          Components: Compiler (C++)
>         Environment: c++
>            Reporter: Ben Taitelbaum
>
> It seems non-intuitive to me to have to set __isset.someField = true after setting an optional field someField on a struct. Would it make sense to have a set_someField method that would both set the field and modify __isset?
> One of the cases for this is for when a field goes from being required to being optional, and it's easy to forget to set __isset in the code.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.