[
https://issues.apache.org/jira/browse/THRIFT-4914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16955096#comment-16955096
]
Duru Can Celasun commented on THRIFT-4914:
------------------------------------------
I'm strongly against exposing the transport in handler functions. After all,
you should be able to use the same handler for different transports.
> new type that gives the user more restricted access, which allows them to set
> headers to return
I think we can do this with the return struct. We can make the map unexported
and provide "SetHeader(key, value string)" method.
> Go: Link between THeader and context object
> -------------------------------------------
>
> Key: THRIFT-4914
> URL: https://issues.apache.org/jira/browse/THRIFT-4914
> Project: Thrift
> Issue Type: New Feature
> Components: Go - Library
> Reporter: Yuxuan Wang
> Priority: Major
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> We have "raw" THeader support for Go in THRIFT-4612 now, the next step would
> be to make them more easily accessible.
> The are 2 directions, 4 parts of this ticket:
> * client -> server (requests)
> ** Read headers on server
> ** Write headers on client
> * server -> client (responses)
> ** Write headers on server
> ** Read headers on client
> Take the reading on server as an example. Currently we can read the headers
> from either the transport or the protocol, but neither the transport nor the
> protocol objects are "easily accessible" when you are writing the business
> logic code (writing the request handler in the server code). It would be much
> better if we inject the headers into the context object passed into the
> request handlers.
> We already have code injecting the headers to the context object that lives
> outside of the thrift library working. I'll send out a PR in the coming days
> to add that to the thrift library, so the performance would be better (we no
> longer need to do the extra injecting work), and it would be a lot easier to
> use.
> We'll think about how to best do the client writing part (probably auto read
> the headers from the context object that passed into the client request code,
> and write to THeaderProtocol automatically), and send out a PR soon-ish.
> The other direction, server -> client, is used much less often with headers,
> so we'll probably punt on it for now, and come back revisit them in the
> future.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
