"Performance vs stability". Net's try-catch efficiency

Reprint please mark: Shu Xiaolong http://www.cnblogs.com/shuxiaolong/archive/2013/04/10/3012439.html

 

Cause:

In most projects, try-catch is not added. Ask why and answer the question about performance;

However, no matter whether the details are correct or not, the errors are caused by exceptions (program or website Yellow Pages). Developers are tired of following up on these bugs;

So I want to know how much efficiency is wasted by try-catch;

 

Code:

1 datetime time1 = datetime. now; 2 for (INT I = 0; I <10000000; I ++) 3 {4 try {int temp = convert. toint32 ("1234567890");} 5 catch (exception exp) {} 6} 7 datetime time2 = datetime. now; 8 for (INT I = 0; I <10000000; I ++) 9 {10 int temp = convert. toint32 ("1234567890"); 11} 12 datetime time3 = datetime. now; 13 14 console. writeline ("Time Difference (MS): \ t has try \ t {0}", (time2-time1 ). totalmilliseconds); 15 console. writeline ("Time Difference (MS): \ t no try \ t {0}", (time3-time2 ). totalmilliseconds); 16 console. writeline ();

 

Result: (run 10 times)

Note: 7th, 10th times: the number of try-catch instances exceeds that of try-catch instances.

10 million try-catch operations only take 1.6 seconds.

 

If the error occurs:

1 datetime time1 = datetime. now; 2 for (INT I = 0; I <10000000; I ++) 3 {4 try {int temp = convert. toint32 ("ABC");} 5 catch (exception exp) {} 6} 7 datetime time2 = datetime. now; 8 console. writeline ("Time Difference (MS): \ t has try \ t {0}", (time2-time1 ). totalmilliseconds ));

Catch performance instantly loses 200 times;

 

Conclusion:

Try is not a waste of efficiency. Catch is a waste of efficiency;

Catch will waste your efficiency, and will crash without catch;

Change the question to "How do you choose to waste efficiency during errors and program crashes during errors ?"

 

 

Case:

In the previous company, there were 6 servers on one website, which were visited by 40 million people every day;

Development Department Code requirements: Your modules do not affect others' modules, so in order to avoid accidents, the code is common: a function is try from the beginning, catch at the end of the function to write exception logs (try-catch is like a part of the function body );

Result: The website is stable and has high performance.

 

 

My thoughts:

The real performance is not that you don't need try-catch; you don't need LINQ; you don't need EF;

Projects that have really participated in, especially projects that deliver others (excluding products and main websites) are not efficient. They are often:

Incorrect algorithms (such as memory paging );

The dog skin plaster that will not cure the symptoms in the later stage (as long as the bug keeps him from reporting an error );

Code that can not be executed is executed in the architecture;

 

Real stability is not necessarily a try-catch anytime, anywhere;

Carefully write the code of each function (try-catch can be added when exceptions may occur) or (even if exceptions occur) or (certainly no exception Code );

Do not trust the returned values of calls from other methods-do not think that other methods will certainly return the value you need;

For all network communications, try-catch must be added to distributed systems;