Some principles of php

http://www.thinkphp.cn/topic/10421.html

php compilation and execution are separated, that is: the compilation is completed first, and then executed. Many people will say: The same is true for c++, indeed. However, this separation of PHP can provide us with a lot of convenience, but of course it also has many disadvantages.

　Let’s talk about the whole process first: 　①php will call the compilation function zend_compile_file() to compile. The specific implementation of this function actually includes two main processes: lexical analysis (Lex implementation) and syntax analysis (Yacc implementation). After executing this function: the compilation of the php script is complete. The input of this function is: php script file, and the output is op_array. To put it simply, the Suncity compilation process is to parse the script into instructions that the php virtual machine can process, and op_array is just an array made of these instructions (this is very similar to the assembly code generated by some compiled languages, which are also commands one by one). 　②: Then the php virtual machine will call the zend_execute() function to execute. The input of this function is the op_array generated in the compilation stage above, where it will parse each command and process it. Since there are about 150 op commands in total, it needs to process these 150 commands. A very interesting question arises here: How does it handle these 150 commands? First of all, each command has its corresponding processor for processing. Therefore: the virtual machine will be distributed to the corresponding processor for processing based on the type of each command in op_array. There are two small questions here: 1: What is the processor here? 2: How is it distributed? To answer these two questions, we need to explain it from the distribution mechanism: There are three mechanisms for PHP virtual machine to distribute commands: CALL, SWITCH, and GOTO. PHP uses the CALL method by default, which is all opcode processors. are defined as functions, The system is then called by the virtual machine. This method is the traditional method and is generally considered the most stable method. The SWITCH method and the GOTO method use switch and goto to distribute the opcode to the corresponding processing logic (segment) for execution. . Now let’s answer the two questions above: 1: The processor actually processes the logic of the op command. It can exist in the form of a function or a logical segment, depending on how the command is distributed. 　2: There are three distribution methods: call, switch and goto. Which one is more efficient? In fact, you can already have a preliminary understanding from the above explanation. Both switch and goto have corresponding logical segments in the zend_execute() function, which can be executed directly. The call is to execute the function call in the zend_execute() function. obvious Note: The efficiency of function calling is the lowest. If you call it once, you have to push it on the stack! So in terms of efficiency: call is the lowest. For switch and goto: For example, if you want to execute the third command: switch must first judge whether it is the first two, but goto does not need to judge at all, and jumps directly to the logical code segment of the third command for execution. This is better than Switch reduces the loss of sequential judgment from top to bottom, so: goto efficiency is higher than switch. So these three distribution methods are generally off topic: since PHP defaults to call, if you want to further squeeze out the performance of PHP, you can change its command distribution method to goto. However, although the goto method improves the execution speed, the compilation speed is actually the slowest. Let’s talk about the weakness of PHP’s separation of compilation and execution: In fact, it cannot be regarded as a weakness, although zend The engine (php's virtual machine) strictly separates compilation and execution, but for Suncity users: it is as if they are not separated, because every time I execute a PHP script request, I have to execute: compile -> execute these two stage. No stage is missing.So we can compare this with a compiled language like c++: Run the same request 100 times　　①For c++, since it only needs to be compiled once in the early stage, it will not be compiled again after compilation. The execution is ok, so the loss is: 　 1 compilation + 100 executions 　 ② For php, it needs to be compiled + executed every time, so the loss is: 100 compilations + 100 executions Obviously: from a quantitative point of view, interpreted languages ​​consume much more than compiled languages. To put it bluntly: the separation of compilation and execution of PHP is not a real separation. The c++ kind is the real separation. PHP has been aware of this problem for a long time, so it thought of a way to solve this problem: this solution is eAccelerator. The main idea is as follows: when the script is run for the first time, the compiled script is saved in a certain way. Within the cache validity time specified by us, when the script is run for the second time, it will no longer perform repeated compilation work, but directly Calling and executing the compiled file saved previously greatly improves program performance. This method improves the efficiency of PHP to a certain extent, but it is not the ultimate method. The ultimate method is to change it to a compiled language. Finally, let me talk about the advantages of separating PHP compilation and execution; this advantage is actually It's for programmers, not for users. Because of the separation of these two stages, we can do something we http://tyangcdh.com want to do here. For example, if you want to do file encryption and decryption, you want to encrypt some php script source code files so that users cannot see the source code. At the same time, this encrypted source code file can be parsed and processed by the PHP virtual machine. Of course: to achieve this, you must first think about the encryption and decryption algorithm and ensure that this is a reversible process. Now that you have encrypted the php source code file, you need to define the suffix 6 of this encrypted file, assuming it is. So the question is: How can we enable the PHP virtual machine to process files with this suffix? This requires the separation of compilation and execution mentioned above. Recall: the input to the compilation phase is the php source file, and the output is op_array. OK, let’s make a fuss at this stage. The main idea is: first in the compilation function zend_compile_file(): look at the suffix of the input file: if it is normal .php, then follow the normal logic; if it is *.buaa, then decrypt it first and then follow the normal logic. Ha~ It’s that simple. Of course: this process is not as simple as said, and you cannot directly modify the zend_compile_file() function. In the end, you need to extend and implement a module yourself to handle this process.

The above introduces some principles of PHP, including aspects of content. I hope it will be helpful to friends who are interested in PHP tutorials.