[
https://issues.apache.org/jira/browse/HDFS-7011?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14153114#comment-14153114
]
Zhanwei Wang commented on HDFS-7011:
------------------------------------
Blocking SIGPIPE in backend worker thread is not important since it does
nothing, I agree to discuss it later.
But I have to say use "std::tr1" namespace directly is not inappropriate for
the following reasons.
1) The basic rule is that we should only use features in C++ standard directly.
Wrap the nonstandard things if we have to use.
At this time, shared_ptr is already in C++ standard and "std::tr1" namespace is
NOT in C++ standard. I understand that using "std::tr1::shared_ptr" can make
old compiler work. But I cannot agree to use nonstandard things everywhere in
the project.
I'm ok to use "std::tr1::shared_ptr", but introduce it into hdfs::internal
namespace and do not use nonstandard namespace "std::tr1" directly. Or use
"std::shared_ptr" directly.
2) Another reason is that I know in some companies, a very old version GCC is
still used and it does not support C++ 11 features, they use boost to get new
features of C++ . I do not mean to introduce the debate if we introduce boost
into the project. My suggestion is that if we can introduce all C++ 11
features into namespace "hdfs::internal", it give people opportunity to change
only few places to switch the implementation of C++ 11 features from compiler
provided runtime library to boost instead of changing the code everywhere in
the project. This provide the opportunity that people can create a private
branch and switch the implementation of C++ 11 features and continually merge
code from official branch. That is the reason I introduce many C++ 11 features
such as unordered_map, thread, shared_ptr into namespace "hdfs::internal".
You also agree that we should benefit as many people as possible, right? Do
such wrap work will not introduce the confusion since the name "unordered_map"
or "shared_ptr" is known by every c++ programmer and people does not pay much
attention on which implementation is used. Even if people is curious about the
implementation, it is easy to find out.
> Implement basic utilities for libhdfs3
> --------------------------------------
>
> Key: HDFS-7011
> URL: https://issues.apache.org/jira/browse/HDFS-7011
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: hdfs-client
> Reporter: Zhanwei Wang
> Assignee: Colin Patrick McCabe
> Attachments: HDFS-7011-pnative.001.patch,
> HDFS-7011-pnative.002.patch, HDFS-7011.patch
>
>
> Implement basic utilities such as hash, exception handling, logger, configure
> parser, checksum calculate and so on.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
