php的true和TRUE引发的一些思考

最近在写php代码的时候突然很纠结一个问题，就是php里面的那些预定义常量也预定义变量到底有什么区别。比如说true和TRUE, false和FALSE，null和NULL... 那么这些变量对到底有什么区别呢，作为一名热衷技术热衷死磕的码农，果断要测试一下。

var_dump(true === TRUE);

结果为bool(true)，那意味着他们是没有区别的。那么问题来了，php为什么要放2个一毛一样的预定义变量呢？是有什么用意呢，还是本身语言设计的历史遗留问题？百度google一大圈 没有找到合适的答案，于是果断去看php的编译源代码，php的整个编译程序只有一个文件，都在zend/zend_compile.c里面，大概7000多行，看着眼花，没办法，也要慢慢看，结果还是找到了想要的答案，这里记录过程，以便查阅。

这里看看php字面量常量的编译

static int zend_add_const_name_literal(zend_op_array *op_array, zend_string *name, zend_bool unqualified) 
{
	zend_string *tmp_name;

	int ret = zend_add_literal_string(op_array, &name);

	size_t ns_len = 0, after_ns_len = ZSTR_LEN(name);
	const char *after_ns = zend_memrchr(ZSTR_VAL(name), '\\', ZSTR_LEN(name));
	if (after_ns) { 
		after_ns += 1;
		ns_len = after_ns - ZSTR_VAL(name) - 1;
		after_ns_len = ZSTR_LEN(name) - ns_len - 1;

		/* lowercased namespace name & original constant name */
		tmp_name = zend_string_init(ZSTR_VAL(name), ZSTR_LEN(name), 0);
		zend_str_tolower(ZSTR_VAL(tmp_name), ns_len);
		zend_add_literal_string(op_array, &tmp_name);

		/* lowercased namespace name & lowercased constant name */
		tmp_name = zend_string_tolower(name);
		zend_add_literal_string(op_array, &tmp_name);

		if (!unqualified) {
			return ret;
		}
	} else {
		after_ns = ZSTR_VAL(name);
	}

	/* original unqualified constant name */
	tmp_name = zend_string_init(after_ns, after_ns_len, 0);
	zend_add_literal_string(op_array, &tmp_name);

	/* lowercased unqualified constant name */ 
	tmp_name = zend_string_alloc(after_ns_len, 0);
	zend_str_tolower_copy(ZSTR_VAL(tmp_name), after_ns, after_ns_len);
	zend_add_literal_string(op_array, &tmp_name);

	return ret;
}

看到其中一个片段

/* lowercased namespace name & lowercased constant name */
tmp_name = zend_string_tolower(name);
zend_add_literal_string(op_array, &tmp_name);

tmp_name = zend_string_tolower(name)，这里全部转换成小写了

接着往下看，发现php的好多地方都是不区分大小写的，包括IF和if， 函数名，方法名...都是不区分大小写的

void zend_begin_method_decl(zend_op_array *op_array, zend_string *name, zend_bool has_body) {
	op_array->scope = ce;
	op_array->function_name = zend_string_copy(name);

	lcname = zend_string_tolower(name);
	lcname = zend_new_interned_string(lcname);
	....

方法名称也全部在编译的时候转换成小写了！

static void zend_begin_func_decl(znode *result, zend_op_array *op_array, zend_ast_decl *decl) {
	unqualified_name = decl->name;
	op_array->function_name = name = zend_prefix_with_ns(unqualified_name);
	lcname = zend_string_tolower(name);
	...

函数名称在编译的时候也全部转换成小写了！

key = zend_build_runtime_definition_key(lcname, decl->lex_pos);
zend_hash_update_ptr(CG(function_table), key, op_array);

这里表示把编译好的函数通过key,op_array代码送入进CG(function_table)的全局函数符号表！

void zend_compile_class_decl(zend_ast *ast) {
	name = zend_new_interned_string(name);
	lcname = zend_string_tolower(name);

	name = zend_generate_anon_class_name(decl->lex_pos);
	lcname = zend_string_tolower(name);
	...

编译类的定义，也全部把类名称转换成小写了，所以类的名字大小写也没关系！

继续往下看...

void zend_compile_use(zend_ast *ast) 
{
	zend_ast_list *list = zend_ast_get_list(ast);
	uint32_t i;
	zend_string *current_ns = FC(current_namespace);
	uint32_t type = ast->attr;
	HashTable *current_import = zend_get_import_ht(type);
	zend_bool case_sensitive = type == T_CONST;
	...
	
	if (case_sensitive) {
		lookup_name = zend_string_copy(new_name);
	} else {
		lookup_name = zend_string_tolower(new_name);

	}
	...

这段代码可以得出2个结论：

1. use 语法是不区分大小写的
2. define常量是严格区分大小写的

继续看...

if (type == T_CLASS && zend_is_reserved_class_name(new_name)) {
	zend_error_noreturn(E_COMPILE_ERROR, "Cannot use %s as %s because '%s' "
			"is a special class name", ZSTR_VAL(old_name), ZSTR_VAL(new_name), ZSTR_VAL(new_name));
}

类名字定义，不能是预定义类名字的限制逻辑！

reserved_class_name的定义如下：

static const struct reserved_class_name reserved_class_names[] = {
	{ZEND_STRL("bool")},
	{ZEND_STRL("false")},
	{ZEND_STRL("float")},
	{ZEND_STRL("int")},
	{ZEND_STRL("null")},
	{ZEND_STRL("parent")},
	{ZEND_STRL("self")},
	{ZEND_STRL("static")},
	{ZEND_STRL("string")},
	{ZEND_STRL("true")},
	{ZEND_STRL("void")},
	{ZEND_STRL("iterable")},
	{NULL, 0}
};

所以，类名都不能是，上述预定义字符串！

接下来看看类的编译

static int zend_add_class_name_literal(zend_op_array *op_array, zend_string *name) {
{
	/* Original name */
	int ret = zend_add_literal_string(op_array, &name);

	/* Lowercased name */
	zend_string *lc_name = zend_string_tolower(name);
	zend_add_literal_string(op_array, &lc_name);

	zend_alloc_cache_slot(ret);

	return ret;
}

void zend_begin_method_decl(zend_op_array *op_array, zend_string *name, zend_bool has_body) 
{
	zend_class_entry *ce = CG(active_class_entry);
	zend_bool in_interface = (ce->ce_flags & ZEND_ACC_INTERFACE) != 0;
	zend_bool in_trait = (ce->ce_flags & ZEND_ACC_TRAIT) != 0;
	zend_bool is_public = (op_array->fn_flags & ZEND_ACC_PUBLIC) != 0;
	zend_bool is_static = (op_array->fn_flags & ZEND_ACC_STATIC) != 0;

	zend_string *lcname;

	if (in_interface) {
		if (!is_public || (op_array->fn_flags & (ZEND_ACC_FINAL|ZEND_ACC_ABSTRACT))) {
			zend_error_noreturn(E_COMPILE_ERROR, "Access type for interface method "
					"%s::%s() must be omitted", ZSTR_VAL(ce->name), ZSTR_VAL(name));
		}
		op_array->fn_flags |= ZEND_ACC_ABSTRACT;
	}

	if (op_array->fn_flags & ZEND_ACC_ABSTRACT) {
		if (op_array->fn_flags & ZEND_ACC_PRIVATE) {
			zend_error_noreturn(E_COMPILE_ERROR, "%s function %s::%s() cannot be declared private",
					in_interface ? "Interface" : "Abstract", ZSTR_VAL(ce->name), ZSTR_VAL(name));
		}

		if (has_body) {
			zend_error_noreturn(E_COMPILE_ERROR, "%s function %s::%s() cannot contain body",
					in_interface ? "Interface" : "Abstract", ZSTR_VAL(ce->name), ZSTR_VAL(name));
		}

		ce->ce_flags |= ZEND_ACC_IMPLICIT_ABSTRACT_CLASS;
	} else if (!has_body) {
		zend_error_noreturn(E_COMPILE_ERROR, "Non-abstract method %s::%s() must contain body",
				ZSTR_VAL(ce->name), ZSTR_VAL(name));
	}

	op_array->scope = ce;
	op_array->function_name = zend_string_copy(name);

	lcname = zend_string_tolower(name);
	lcname = zend_new_interned_string(lcname);

这段代码有意思，php的什么和抽象类的修饰权限，全是在编译的时候检测的，而且，检测与否其实就是一个if

if (in_interface) {		
	if (!is_public || (op_array->fn_flags & (ZEND_ACC_FINAL|ZEND_ACC_ABSTRACT))) {			
		zend_error_noreturn(E_COMPILE_ERROR, "Access type for interface method "				
				"%s::%s() must be omitted", ZSTR_VAL(ce->name), ZSTR_VAL(name));		
	}		
	op_array->fn_flags |= ZEND_ACC_ABSTRACT;	
}

如果是接口，但是修饰flag不是public就报错了！

这也可以说明在php里面，写个interface，然后写很多实现，这种模式不是很有必要，也就是欺骗一下编译器，运行的时候还是到CG(function_table)或者CG(active_class_entry)中取查找，会让编译过程变慢，所以特么opcache这么有用的很大原因之一！

最后献上PHP之父Rasmus在php巴黎会议的一段话

Rasmus in a PHP conference in Paris saying more or less: "I'm definitely not a good programmer, in terms of following strict coding rules or standards, 
	   but I can say that if you rely on case sensitivity to recognize one function name from another, you're in kind of serious trouble

这里简单翻译一下：“我绝对不是一个好的程序员，至少在遵循严格的编码规则和标准方面， 但是我不得不说，如果你要依靠大小写来区分自己写的另一个函数，那你将会非常麻烦” 到此为止，这估计是php不区分大小写的最好的理由了。