[
https://issues.apache.org/jira/browse/THRIFT-2939?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14272899#comment-14272899
]
Radoslaw Gruchalski commented on THRIFT-2939:
---------------------------------------------
This does not break the exposed interface. It only changes generated code. This
means that code generated prior to 1.0.0 would not work with 1.0.0. Seems
acceptable to me. I can provide additional validation in 1.0.0 and human
readable errors. Does this help? Or try/catch in generated code but this one
seems nasty.
> JavaScript generated code for Node and browser is different
> -----------------------------------------------------------
>
> Key: THRIFT-2939
> URL: https://issues.apache.org/jira/browse/THRIFT-2939
> Project: Thrift
> Issue Type: Improvement
> Components: JavaScript - Compiler, JavaScript - Library
> Affects Versions: 0.9.2
> Reporter: Radoslaw Gruchalski
> Labels: javascript, patch
>
> The generated code for regular JS differs from NodeJS code. NodeJS uses
> direct return values from {{read<Type>()}} while regular JS code expects a
> value from {{read<Type>()}} to be an object with {{.value}} property.
> This makes it impossible to share generated code between browser JS and
> NodeJS. The fix is rather simple:
> - do not append {{.value}} in generated code while accessing values read from
> protocol / transport
> - do not return an object with the value property from {{read}}, return
> result read
> I am not sure why such difference. Both environment implementations take an
> input, which is a protocol, and execute the same read methods. There is no
> need to have an additional complexity in regular JS.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
