[
https://issues.apache.org/jira/browse/THRIFT-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Anthony Molinaro resolved THRIFT-1126.
--------------------------------------
Resolution: Fixed
Fix Version/s: 0.7
Assignee: Bryan Duxbury (was: Anthony Molinaro)
I applied the patch then was looking at the generated code for the
ConstantsDemo.thrift in the test directory and noticed that the constants for
maps and sets seemed weird. See if you can spot the problem with these
{noformat}
-define(constantsDemo_GEN_MAP, dict:from_list([(35532,233),(43523,853)])).
-define(constantsDemo_GEN_SET,
sets:from_list([(235,true),(235,true),(53235,true)])).
{noformat}
Yeah, turns out if you actually try to use those you get errors, as 1, parens
are not how you create tuples, and 2, then you use sets:from_list its also not
what you expect (which as far as I am concerned this this case is a set with 2
elements as sets dedupe).
As far as I can tell complex constants probably never worked, guess that shows
how much they are used :).
Anyway, I went ahead and fixed that bug and commited the patch.
Thanks again Anatoly!
> Extending struct_info for erlang bindings
> -----------------------------------------
>
> Key: THRIFT-1126
> URL: https://issues.apache.org/jira/browse/THRIFT-1126
> Project: Thrift
> Issue Type: Improvement
> Components: Erlang - Compiler
> Reporter: Anatoly Kanivetsky
> Assignee: Bryan Duxbury
> Fix For: 0.7
>
> Attachments: thrift-1086852.diff
>
>
> In addition to recently fixed THRIFT-27, I've made some extra changes, if no
> one objects.
> In this patch i've added more type specifiers to generated records.
> Now it covers all possible types (I've forgot about set(), map() and
> #another_record{}).
> It now initializes all record fields without default value with 'undefined'.
> This is a bit redundant, but I like it. (maybe you're not? let me know if you
> have some objections!).
> Also in this patch adds generation of additional structure info, that might
> be useful from runtime.
> It contains:
> * FieldID,
> * FieldReq,
> * FieldType
> * Identifier
> * ConstValue
> ConstValue is constructed everytime when struct_info_ext(...) is called.
> This isn't good, but perhaps, it is ok if erlang optimizes such things. Does
> it?
> I planned to use struct_info_ext(...) in further modifications of erlang
> thrift library, to implement runtime validation of presence of required
> fields during reading/writing, and to avoid encoding of optional fields with
> default values. But it is pretty usable of itself.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
