Rumah > Artikel > pembangunan bahagian belakang > 深入理解PHP内核(六)函数的定义、传参及返回值,深入理解内核_PHP教程
一、函数的定义
用户函数的定义从function 关键字开始,如下
<span>function</span> foo(<span>$var</span><span>) { </span><span>echo</span> <span>$var</span><span>; }</span>
1、词法分析
在Zend/zend_language_scanner.l中我们找到如下所示的代码:
<ST_IN_SCRIPTING><span>"</span><span>function</span><span>"</span><span> { </span><span>return</span><span> T_FUNCTION; }</span>
它所表示的含义是function将会生成T_FUNCTION标记。在获取这个标记后,我们开始语法分析。
2、语法分析
在Zend/zend_language_parser.y文件中找到函数的声明过程标记如下:
<span>function: T_FUNCTION { $$.u.opline_num </span>=<span> CG(zend_lineno); } ; is_reference: </span><span>/*</span><span> empty </span><span>*/</span> { $$.op_type =<span> ZEND_RETURN_VAL; } </span>| <span>'</span><span>&</span><span>'</span> { $$.op_type =<span> ZEND_RETURN_REF; } ; unticked_function_declaration_statement: function is_reference T_STRING { zend_do_begin_function_declaration(</span>&$<span>1</span>, &$<span>3</span>, <span>0</span>, $<span>2</span><span>.op_type, NULL TSRMLS_CC); } </span><span>'</span><span>(</span><span>'</span> parameter_list <span>'</span><span>)</span><span>'</span> <span>'</span><span>{</span><span>'</span> inner_statement_list <span>'</span><span>}</span><span>'</span><span> { zend_do_end_function_declaration(</span>&$<span>1</span><span> TSRMLS_CC); } ;</span>
关注点在function is_reference T_STRING,表示function关键字,是否引用,函数名
T_FUNCTION标记只是用来定位函数的声明,表示这是一个函数,而更多的工作是与这个函数相关的东西,包括参数,返回值。
3、生成中间代码
语法解析后,我们看到所执行编译函数为zend_do_begin_function_declaration。在Zend/zend_complie.c文件找到其实现如下:
<span>void</span> zend_do_begin_function_declaration(znode *<span>function_token, znode </span>*<span>function_name, </span><span>int</span> is_method, <span>int</span> return_reference, znode *fn_flags_znode TSRMLS_DC) <span>/*</span><span> {{{ </span><span>*/</span><span> { ...</span><span>//</span><span>省略</span> function_token->u.op_array =<span> CG(active_op_array); lcname </span>=<span> zend_str_tolower_dup(name, name_len); orig_interactive </span>=<span> CG(interactive); CG(interactive) </span>= <span>0</span><span>; init_op_array(</span>&<span>op_array, ZEND_USER_FUNCTION, INITIAL_OP_ARRAY_SIZE TSRMLS_CC); CG(interactive) </span>=<span> orig_interactive; ...</span><span>//</span><span>省略</span> <span>if</span><span> (is_method) { ...</span><span>//</span><span>省略,类方法 在后面的章节介绍</span> !<span>ǶGH } </span><span>else</span><span> { zend_op </span>*opline =<span> get_next_op(CG(active_op_array) TSRMLS_CC); opline</span>->opcode =<span> ZEND_DECLARE_FUNCTION; opline</span>->op1.op_type =<span> IS_CONST; build_runtime_defined_function_key(</span>&opline-><span>op1.u.constant, lcname, name_len TSRMLS_CC); opline</span>->op2.op_type =<span> IS_CONST; opline</span>->op2.u.constant.type =<span> IS_STRING; opline</span>->op2.u.constant.value.str.val =<span> lcname; opline</span>->op2.u.constant.value.str.len =<span> name_len; Z_SET_REFCOUNT(opline</span>->op2.u.constant, <span>1</span><span>); opline</span>->extended_value =<span> ZEND_DECLARE_FUNCTION; zend_hash_update(CG(function_table), opline</span>- ><span>op1.u.constant.value.str.val, opline</span>->op1.u.constant.value.str.len, &<span>op_array, </span><span>sizeof</span><span>(zend_op_array), (</span><span>void</span> **) &<span>CG(active_op_array)); } } </span><span>/*</span><span> }}} </span><span>*/</span><span><br /></span>
生成的代码为ZEND_DECLARE_FUNCTION,根据这个中间的代码及操作数对应的op_type。我们可以找到中间代码的执行函数为ZEND_DECLARE_FUNCTION_SPEC_HANDLER。
在生成中间代码的时候,可以看到已经统一了函数名全部为小写,表示函数的名称不是区 分大小写的。
为验证这个实现,我们看一段代码
<span>function T() { echo </span><span>1</span><span>; } function t() { echo </span><span>2</span><span>; }</span>
执行代码会报错Fatal error: Cannot redeclare t() (previously declared in ...)
表示对于PHP来说T和t是同一个函数名,校验函数名是否重复,这个过程是在哪进行的呢?
4、执行中间代码
在Zend/zend_vm_execute.h文件中找到ZEND_DECLARE_FUNCTION中间代码对应的执行函数:ZEND_DECLARE_FUNCTION_SPEC_HANDLER。此函数只调用了函数do_bind_function。其调用代码为:
do_bind_function(EX(opline), EG(function_table), <span>0</span>);
在这个函数中将EX(opline)所指向的函数添加到EG(function_table)中,并判断是否已经存在相同名字的函数,如果存在则报错,EG(function_table)用来存放执行过程中全部的函数信息,相当于函数的注册表。它的结构是一个HashTable,所以在do_bind_function函数中添加新的函数使用的是HashTable的操作函数zend_hash_add
二、函数的参数
函数的定义只是一个将函数名注册到函数列表的过程。
1、用户自定义函数的参数
我们知道对于函数的参数检查是通过zend_do_receive_arg函数来实现的,在此函数中对于参数的关键代码如下:
CG(active_op_array)->arg_info = erealloc(CG(active_op_array)-><span>arg_info, </span><span>sizeof</span>(zend_arg_info)*(CG(active_op_array)-><span>num_args)); cur_arg_info </span>= &CG(active_op_array)->arg_info[CG(active_op_array)->num_args-<span>1</span><span>]; cur_arg_info</span>->name = estrndup(varname-><span>u.constant.value.str.val, varname</span>-><span>u.constant.value.str.len); cur_arg_info</span>->name_len = varname-><span>u.constant.value.str.len; cur_arg_info</span>->array_type_hint = <span>0</span><span>; cur_arg_info</span>->allow_null = <span>1</span><span>; cur_arg_info</span>->pass_by_reference =<span> pass_by_reference; cur_arg_info</span>->class_name =<span> NULL; cur_arg_info</span>->class_name_len = <span>0</span>;
整个参数的传递是通过给中间代码的arg_info字段执行赋值操作完成。关键点是在arg_info字段,arg_info字段的结构如下:
typedef <span>struct</span><span> _zend_arg_info { </span><span>const</span> <span>char</span> *name; <span>/*</span><span>参数的名称</span><span>*/</span><span> zend_uint name_len; </span><span>/*</span><span>参数名称的长度</span><span>*/</span> <span>const</span> <span>char</span> *class_name; <span>/*</span><span> 类名</span><span>*/</span><span> zend_uint class_name_len; </span><span>/*</span><span>类名长度</span><span>*/</span><span> zend_bool array_type_hint; </span><span>/*</span><span>数组类型提示</span><span>*/</span><span> zend_bool allow_null; </span><span>/*</span><span>是否允许为NULLͺ</span><span>*/</span><span> zend_bool pass_by_reference; </span><span>/*</span><span>是否引用传递</span><span>*/</span><span> zend_bool return_reference; </span><span>int</span><span> required_num_args; } zend_arg_info;</span>
参数的值传递和参数传递的区别是通过pass_by_reference参数在生成中间代码时实现的。
对于参数的个数,中间代码中包含的arg_nums字段在每次执行**zend_do_receive_argxx时都会加1.如下代码:
CG(active_op_array)->num_args++;
并且当前参数的索引为CG(active_op_array)->num_args-1.如下代码:
cur_arg_info = &CG(active_op_array)->arg_info[CG(active_op_array)->num_args-<span>1</span>];
以上的分析是针对函数定义时的参数设置,这些参数是固定的。而在实际编写程序时可能我们会用到可变参数。此时我们会用到函数func_num_args和func_get_args。它们是以内部函数存在。于是在Zend\zend_builtin_functions.c文件中找到这两个函数的实现。我们首先来看func_num_args函数的实现,其代码如下:
<span>/*</span><span> {{{ proto int func_num_args(void) Get the number of arguments that were passed to the function </span><span>*/</span><span> ZEND_FUNCTION(func_num_args) { zend_execute_data </span>*ex = EG(current_execute_data)-><span>prev_execute_data; </span><span>if</span> (ex && ex-><span>function_state.arguments) { RETURN_LONG((</span><span>long</span>)(zend_uintptr_t)*(ex-><span>function_state.arguments)); } </span><span>else</span><span> { zend_error(E_WARNING, </span><span>"</span><span>func_num_args(): Called from the global scope - no function context</span><span>"</span><span>); RETURN_LONG(</span>-<span>1</span><span>); } } </span><span>/*</span><span> }}} </span><span>*/</span>
在存在ex->function_state.arguments的情况下,及函数调用时,返回ex->function_state.arguments转化后的值,否则显示错误并返回-1。这里最关键的一点是EG(current_execute_data)。这个变量存放的是当前执行程序或函数的数据,此时我们需要取前一个执行程序的数据,为什么呢?因为这个函数的调用是在进入函数后执行的。函数的相关数据等都在之前执行过程中,于是调用的是:
zend_execute_data *ex = EG(current_execute_data)->prev_execute_data;
2、内部函数的参数
以常见的count函数为例,其参数处理部分的代码如下:
<span>/*</span><span> {{{ proto int count(mixed var [, int mode]) Count the number of elements in a variable (usually an array) </span><span>*/</span><span> PHP_FUNCTION(count) { zval </span>*<span>array; </span><span>long</span> mode =<span> COUNT_NORMAL; </span><span>if</span> (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, <span>"</span><span>z|l</span><span>"</span><span>, </span>&array, &mode) ==<span> FAILURE) { </span><span>return</span><span>; } ... </span><span>//</span><span>省略</span> }
这里包括了两个操作:一个是取参数的个数,一个是解析参数列表。
(1)取参数的个数
取参数的个数是通过ZEND_NUM_ARGS()宏来实现的,其定义如下:
<span>#define</span> ZEND_NUM_ARGS() (ht)
ht是在Zend/zend.h文件中定义的宏INTERNAL_FUNCTION_PARAMETERS中的ht,如下
<span>#define</span> INTERNAL_FUNCTION_PARAMETERS int ht, zval *return_value,<span> zval </span>**return_value_ptr, zval *this_ptr, <span>int</span> return_value_used TSRMLS_DC
(2)解析参数列表
PHP内部函数在解析参数时使用的是zend_parse_parameters。它可以大大简化参数的接收处理工作,虽然它在处理可变参数时还有点弱。
其声明如下:
ZEND_API <span>int</span> zend_parse_parameters(<span>int</span> num_args TSRMLS_DC, <span>char</span> *<span>type_spec, ...)</span>
zend_parse_parameters()在解析参数的同时户尽可能的转换参数类型,这样就可以确保我们总是能得到所期望的类型的变量
3、函数的返回值
PHP中函数都有返回值,没return返回null
(1)return语句
从Zend/zend_language_parser.y文件中可以确认其生成中间代码调用的是zend_do_return函数。
<span>void</span> zend_do_return(znode *expr, <span>int</span> do_end_vparse TSRMLS_DC) <span>/*</span><span> {{{ </span><span>*/</span><span> { zend_op </span>*<span>opline; </span><span>int</span><span> start_op_number, end_op_number; </span><span>if</span><span> (do_end_vparse) { </span><span>if</span> (CG(active_op_array)-><span>return_reference </span>&& !<span>zend_is_function_or_method_call(expr)) { zend_do_end_variable_parse(expr, BP_VAR_W, </span><span>0</span> TSRMLS_CC);<span>/*</span><span> 处理返回引用 </span><span>*/</span><span> } </span><span>else</span><span> { zend_do_end_variable_parse(expr, BP_VAR_R, </span><span>0</span> TSRMLS_CC);<span>/*</span><span> 处理常规变量返回 </span><span>*/</span><span> } } ...</span><span>//</span><span> 省略,取其他中间代码操作</span> <span> opline</span>->opcode =<span> ZEND_RETURN; </span><span>if</span><span> (expr) { opline</span>->op1 = *<span>expr; </span><span>if</span> (do_end_vparse &&<span> zend_is_function_or_method_call(expr)) { opline</span>->extended_value =<span> ZEND_RETURNS_FUNCTION; } } </span><span>else</span><span> { opline</span>->op1.op_type =<span> IS_CONST; INIT_ZVAL(opline</span>-><span>op1.u.constant); } SET_UNUSED(opline</span>-><span>op2); } </span><span>/*</span><span> }}} </span><span>*/</span>
生成中间代码为ZEND_RETURN。第一个操作数的类型在返回值为可用的表达式时,其类型为表达式的操作类型,否则类型为IS_CONST。这在后续计算执行中间代码函数时有用到。根据操作数的不同,ZEND_RETURN中间代码会执行ZEND_RETURN_SPEC_CONST_HANDLER,ZEND_RETURN_SPEC_TMP_HANDLER或ZEND_RETURN_SPEC_TMP_HANDLER。这三个函数的执行流程基本类似,包括对一些错误的处理。这里我们以ZEND_RETURN_SPEC_CONST_HANDLER为例说明函数返回值的执行过程:
<span>static</span> <span>int</span><span> ZEND_FASTCALL ZEND_RETURN_SPEC_CONST_HANDLER(ZEND_OPCODE_HANDLER_ARGS) { zend_op </span>*opline =<span> EX(opline); zval </span>*<span>retval_ptr; zval </span>**<span>retval_ptr_ptr; </span><span>if</span> (EG(active_op_array)->return_reference ==<span> ZEND_RETURN_REF) { </span><span>//</span><span> ǓǔŷsÁ\ɁƶMļ@ɗÁĻļ</span> <span>if</span> (IS_CONST == IS_CONST || IS_CONST ==<span> IS_TMP_VAR) { </span><span>/*</span><span> Not supposed to happen, but we'll allow it </span><span>*/</span><span> zend_error(E_NOTICE, </span><span>"</span><span>Only variable references \</span> should be returned by reference<span>"</span><span>);</span> <span>goto</span><span> return_by_value; } retval_ptr_ptr </span>= NULL; <span>//</span><span> ǓǔŔ</span> <span>if</span> (IS_CONST == IS_VAR && !<span>retval_ptr_ptr) { zend_error_noreturn(E_ERROR, </span><span>"</span><span>Cannot return string offsets by </span> reference<span>"</span><span>);</span> <span> } </span><span>if</span> (IS_CONST == IS_VAR && !<span>Z_ISREF_PP(retval_ptr_ptr)) { </span><span>if</span> (opline->extended_value == ZEND_RETURNS_FUNCTION &&<span> EX_T(opline</span>->op1.u.<span>var</span>).<span>var</span><span>.fcall_returned_reference) { } </span><span>else</span> <span>if</span> (EX_T(opline->op1.u.<span>var</span>).<span>var</span>.ptr_ptr == &EX_T(opline->op1.u.<span>var</span>).<span>var</span><span>.ptr) { </span><span>if</span> (IS_CONST == IS_VAR && !<span>0</span><span>) { </span><span>/*</span><span> undo the effect of get_zval_ptr_ptr() </span><span>*/</span><span> PZVAL_LOCK(</span>*<span>retval_ptr_ptr); } zend_error(E_NOTICE, </span><span>"</span><span>Only variable references \</span> should be returned by reference<span>"</span><span>);</span> <span>goto</span><span> return_by_value; } } </span><span>if</span> (EG(return_value_ptr_ptr)) { <span>//</span><span> Ǔǔŷs</span> SEPARATE_ZVAL_TO_MAKE_IS_REF(retval_ptr_ptr); <span>//</span><span> is_ref__gcőę</span> <span>1</span><span> Z_ADDREF_PP(retval_ptr_ptr); </span><span>//</span><span> refcount__gcŒď×1</span> <span> (</span>*EG(return_value_ptr_ptr)) = (*<span>retval_ptr_ptr); } } </span><span>else</span><span> { return_by_value: retval_ptr </span>= &opline-><span>op1.u.constant; </span><span>if</span> (!<span>EG(return_value_ptr_ptr)) { </span><span>if</span> (IS_CONST ==<span> IS_TMP_VAR) { } } </span><span>else</span> <span>if</span> (!<span>0</span>) { <span>/*</span><span> Not a temp var </span><span>*/</span> <span>if</span> (IS_CONST == IS_CONST ||<span> EG(active_op_array)</span>->return_reference == ZEND_RETURN_REF ||<span> (PZVAL_IS_REF(retval_ptr) </span>&& Z_REFCOUNT_P(retval_ptr) > <span>0</span><span>)) { zval </span>*<span>ret; ALLOC_ZVAL(ret); INIT_PZVAL_COPY(ret, retval_ptr); </span><span>//</span><span> ŁͿʍǓǔŔ </span> <span> zval_copy_ctor(ret); </span>*EG(return_value_ptr_ptr) =<span> ret; } </span><span>else</span><span> { </span>*EG(return_value_ptr_ptr) = retval_ptr; <span>//</span><span> ħ6ɶŔ</span> <span> Z_ADDREF_P(retval_ptr); } } </span><span>else</span><span> { zval </span>*<span>ret; ALLOC_ZVAL(ret); INIT_PZVAL_COPY(ret, retval_ptr); </span><span>//</span><span> ŁͿʍǓǔŔ </span> *EG(return_value_ptr_ptr) =<span> ret; } } </span><span>return</span> zend_leave_helper_SPEC(ZEND_OPCODE_HANDLER_ARGS_PASSTHRU); <span>//</span><span> Ǔ</span> <span>ǔĉșʒ }</span>
函数的返回值在程序执行时存储在*EG(return_value_ptr_ptr)。ZEND内核对值返回和引用返回作了区别,并且在此基础上对常量,临时变量和其他类型的变量在返回时作了不同的处理。在return执行完之后,ZEND内核通过调用zend_leave_helper_SPEC函数,清除函数内部使用的变量等。这也是ZEND内核自动给函数加上NULL返回的原因之一。