Inefficiency of http_request::extract_vector() method · microsoft/cpprestsdk#103

(2 留言) (0 反應) (0 負責人)C++ (7,596 star) (1,670 fork)batch import

enhancementhelp wanted

描述

I'm looking at using the C++ REST SDK to implement a relatively simple but performance-critical REST server in our project. I've put together a PoC and it all looks good. But I've just hit upon what I think is a major performance issue.

When using the http_request::extract_vector() method, the Task completion routine is passed the retrieved vector by value. Our server will sometimes be posted a large content body (1MB or more). The pass-by-value means that this vector will be copied. This completely removes one of the major advantage of using C/C++ for this kind of work, i.e. the fact that if you know what you're doing, you can avoid the huge amount of copying that can be a bottleneck on platforms like NodeJS.

Am I missing something obvious here or will I have to implement a custom extract_vector_ptr() method myself?

貢獻者指南

技術棧: cpp
領域: backendperformance
議題類型: performance
難度: 3
預計時間: half day
活動狀態: stale
清晰度: clear
前置要求: Basic knowledge of C++Understanding of C++ REST SDK async patterns
新手友善度: 45
研究方向: Investigate the implementation of http request::extract vector() in the cpprestsdk source code. Look for the Task completion routine that passes the vector by value. Consider modifying the method to return a shared pointer or implement a move semantics variant. Review any existing discussions in the issue comments for suggested approaches.