你写的字符(串)忽略大小写函数真的严谨吗？
提示
阅读本文需要同时对c++和java有一定了解。
背景
有时我们比较两个字符串时不考虑它们是大写还是小写；举个例子，在这种情况下我们认为“BanAna”和“baNaNA”是等价的。
其中一种思路是：
1. 将两个字符串都转换为小写（或者都转换为大写）；
2.比较转换后的两个字符串是否相同。
这里给出一段C++示例代码：
//C++ example that we offen use bool testIgnoreCase(string str1, string str2){ transform(str1.begin(),str1.end(),str1.begin(),::tolower); transform(str2.begin(),str2.end(),str2.begin(),::tolower); //Or //transform(str1.begin(),str1.end(),str1.begin(),::toupper); //transform(str2.begin(),str2.end(),str2.begin(),::toupper); cout<<str1<<" "<<str2<<endl;//apple apple return str1 == str2; } int main() { string str1 = "ApplE"; string str2 = "apPle"; cout<<testIgnoreCase(str1,str2);//1 return 0; }
上面的代码同一将两个字符串转换为了小写，然后比较。当然你先转换为大写也行。
看起来功能已经实现了。
但这种做法真的严谨吗？
考虑下面的两个例子：
//C++ example1 bool testIgnoreCase(string str1, string str2){ transform(str1.begin(),str1.end(),str1.begin(),::tolower); transform(str2.begin(),str2.end(),str2.begin(),::tolower); //Or //transform(str1.begin(),str1.end(),str1.begin(),::toupper); //transform(str2.begin(),str2.end(),str2.begin(),::toupper); cout<<str1<<" "<<str2<<endl;//ı i return str1 == str2; } int main() { string str1 = "ı";//unicode=305,注意不在ascii范围内 string str2 = "I";//常见的大写字母I cout<<testIgnoreCase(str1,str2);//0 return 0; }
//C++ example2 bool testIgnoreCase(string str1, string str2){ //transform(str1.begin(),str1.end(),str1.begin(),::tolower); //transform(str2.begin(),str2.end(),str2.begin(),::tolower); //Or transform(str1.begin(),str1.end(),str1.begin(),::toupper); transform(str2.begin(),str2.end(),str2.begin(),::toupper); cout<<str1<<" "<<str2<<endl;//İ I return str1 == str2; } int main() { string str1 = "İ";//unicode=304,注意不在ascii范围内 string str2 = "i";//常见的小写字母i cout<<testIgnoreCase(str1,str2);//0 return 0; }
从上面两个例子中，可以看到，不管是全部转换为小写还是全部转换为大写，再比较的方式，都是不严谨的。主要的原因是我们没有考虑超出ascii编码范围的字符。
上面的例子中，总共涉及到四个字符，分别为：
| i | 常见的小写字母i，Ascii=105 |
| I | 常见的大写字母I，Ascii=73 |
ı
|
unicode=305 |
İ
|
unicode=304 |
因此引出一个疑问：这四个字符，是一族的吗？换句话说，它们是否真的被视为等价？如果它们不等价，上面的问题就不算是问题了。
这个问题就涉及到两种语言之间的差异了：
Java中，它们之间大小写转换关系如下：
而C++中，这几个字符不被视为等价，这就意味着，就算你这样写（先转换为小写，如果还不相等，再转换为大写判断；当然先转换为大写后转换为小写是一样的思路）：
bool testIgnoreCase(string str1, string str2){ transform(str1.begin(),str1.end(),str1.begin(),::tolower); transform(str2.begin(),str2.end(),str2.begin(),::tolower); if(str1 == str2) { return true; } transform(str1.begin(),str1.end(),str1.begin(),::toupper); transform(str2.begin(),str2.end(),str2.begin(),::toupper); return str1 == str2; }
也不会起丝毫作用。
那Java中是如何实现IgnoreCace的呢？
看Java中的equalsIgnoreCase()函数源码：
//Java public boolean equalsIgnoreCase(String anotherString) { return (this == anotherString) ? true : (anotherString != null) && (anotherString.value.length == value.length) && regionMatches(true, 0, anotherString, 0, value.length); } public boolean regionMatches(boolean ignoreCase, int toffset, String other, int ooffset, int len) { char ta[] = value; int to = toffset; char pa[] = other.value; int po = ooffset; // Note: toffset, ooffset, or len might be near -1>>>1. if ((ooffset < 0) || (toffset < 0) || (toffset > (long)value.length - len) || (ooffset > (long)other.value.length - len)) { return false; } while (len-- > 0) { char c1 = ta[to++]; char c2 = pa[po++]; if (c1 == c2) { continue; } if (ignoreCase) { // If characters don't match but case may be ignored, // try converting both characters to uppercase. // If the results match, then the comparison scan should // continue. char u1 = Character.toUpperCase(c1); char u2 = Character.toUpperCase(c2); if (u1 == u2) { continue; } // Unfortunately, conversion to uppercase does not work properly // for the Georgian alphabet, which has strange rules about case // conversion. So we need to make one last check before // exiting. if (Character.toLowerCase(u1) == Character.toLowerCase(u2)) { continue; } } return false; } return true; }
可以看到，Java中的忽略大小写比较先将字符转换为大写，对于不相等的字符，又转换为小写比较；这样做相当于多了一层保障。
再细究，我们先看小写转换，观察其更为底层的实现：
1 int toLowerCase(int ch) { 2 int mapChar = ch; 3 int val = getProperties(ch); 4 5 if ((val & 0x00020000) != 0) { 6 if ((val & 0x07FC0000) == 0x07FC0000) { 7 switch(ch) { 8 // map the offset overflow chars 9 case 0x0130 : mapChar = 0x0069; break; 10 case 0x2126 : mapChar = 0x03C9; break; 11 case 0x212A : mapChar = 0x006B; break; 12 case 0x212B : mapChar = 0x00E5; break; 13 // map the titlecase chars with both a 1:M uppercase map 14 // and a lowercase map 15 case 0x1F88 : mapChar = 0x1F80; break; 16 case 0x1F89 : mapChar = 0x1F81; break; 17 case 0x1F8A : mapChar = 0x1F82; break; 18 case 0x1F8B : mapChar = 0x1F83; break; 19 case 0x1F8C : mapChar = 0x1F84; break; 20 case 0x1F8D : mapChar = 0x1F85; break; 21 case 0x1F8E : mapChar = 0x1F86; break; 22 case 0x1F8F : mapChar = 0x1F87; break; 23 case 0x1F98 : mapChar = 0x1F90; break; 24 case 0x1F99 : mapChar = 0x1F91; break; 25 case 0x1F9A : mapChar = 0x1F92; break; 26 case 0x1F9B : mapChar = 0x1F93; break; 27 case 0x1F9C : mapChar = 0x1F94; break; 28 case 0x1F9D : mapChar = 0x1F95; break; 29 case 0x1F9E : mapChar = 0x1F96; break; 30 case 0x1F9F : mapChar = 0x1F97; break; 31 case 0x1FA8 : mapChar = 0x1FA0; break; 32 case 0x1FA9 : mapChar = 0x1FA1; break; 33 case 0x1FAA : mapChar = 0x1FA2; break; 34 case 0x1FAB : mapChar = 0x1FA3; break; 35 case 0x1FAC : mapChar = 0x1FA4; break; 36 case 0x1FAD : mapChar = 0x1FA5; break; 37 case 0x1FAE : mapChar = 0x1FA6; break; 38 case 0x1FAF : mapChar = 0x1FA7; break; 39 case 0x1FBC : mapChar = 0x1FB3; break; 40 case 0x1FCC : mapChar = 0x1FC3; break; 41 case 0x1FFC : mapChar = 0x1FF3; break; 42 43 case 0x023A : mapChar = 0x2C65; break; 44 case 0x023E : mapChar = 0x2C66; break; 45 case 0x10A0 : mapChar = 0x2D00; break; 46 case 0x10A1 : mapChar = 0x2D01; break; 47 case 0x10A2 : mapChar = 0x2D02; break; 48 case 0x10A3 : mapChar = 0x2D03; break; 49 case 0x10A4 : mapChar = 0x2D04; break; 50 case 0x10A5 : mapChar = 0x2D05; break; 51 case 0x10A6 : mapChar = 0x2D06; break; 52 case 0x10A7 : mapChar = 0x2D07; break; 53 case 0x10A8 : mapChar = 0x2D08; break; 54 case 0x10A9 : mapChar = 0x2D09; break; 55 case 0x10AA : mapChar = 0x2D0A; break; 56 case 0x10AB : mapChar = 0x2D0B; break; 57 case 0x10AC : mapChar = 0x2D0C; break; 58 case 0x10AD : mapChar = 0x2D0D; break; 59 case 0x10AE : mapChar = 0x2D0E; break; 60 case 0x10AF : mapChar = 0x2D0F; break; 61 case 0x10B0 : mapChar = 0x2D10; break; 62 case 0x10B1 : mapChar = 0x2D11; break; 63 case 0x10B2 : mapChar = 0x2D12; break; 64 case 0x10B3 : mapChar = 0x2D13; break; 65 case 0x10B4 : mapChar = 0x2D14; break; 66 case 0x10B5 : mapChar = 0x2D15; break; 67 case 0x10B6 : mapChar = 0x2D16; break; 68 case 0x10B7 : mapChar = 0x2D17; break; 69 case 0x10B8 : mapChar = 0x2D18; break; 70 case 0x10B9 : mapChar = 0x2D19; break; 71 case 0x10BA : mapChar = 0x2D1A; break; 72 case 0x10BB : mapChar = 0x2D1B; break; 73 case 0x10BC : mapChar = 0x2D1C; break; 74 case 0x10BD : mapChar = 0x2D1D; break; 75 case 0x10BE : mapChar = 0x2D1E; break; 76 case 0x10BF : mapChar = 0x2D1F; break; 77 case 0x10C0 : mapChar = 0x2D20; break; 78 case 0x10C1 : mapChar = 0x2D21; break; 79 case 0x10C2 : mapChar = 0x2D22; break; 80 case 0x10C3 : mapChar = 0x2D23; break; 81 case 0x10C4 : mapChar = 0x2D24; break; 82 case 0x10C5 : mapChar = 0x2D25; break; 83 case 0x10C7 : mapChar = 0x2D27; break; 84 case 0x10CD : mapChar = 0x2D2D; break; 85 case 0x1E9E : mapChar = 0x00DF; break; 86 case 0x2C62 : mapChar = 0x026B; break; 87 case 0x2C63 : mapChar = 0x1D7D; break; 88 case 0x2C64 : mapChar = 0x027D; break; 89 case 0x2C6D : mapChar = 0x0251; break; 90 case 0x2C6E : mapChar = 0x0271; break; 91 case 0x2C6F : mapChar = 0x0250; break; 92 case 0x2C70 : mapChar = 0x0252; break; 93 case 0x2C7E : mapChar = 0x023F; break; 94 case 0x2C7F : mapChar = 0x0240; break; 95 case 0xA77D : mapChar = 0x1D79; break; 96 case 0xA78D : mapChar = 0x0265; break; 97 case 0xA7AA : mapChar = 0x0266; break; 98 // default mapChar is already set, so no 99 // need to redo it here. 100 // default : mapChar = ch; 101 } 102 } 103 else { 104 int offset = val << 5 >> (5+18); 105 mapChar = ch + offset; 106 } 107 } 108 return mapChar; 109 }
源码中的getProperties，获取到字符所处的属性集，然后根据不同的情况执行对应的操作，对于我们的例子，源码第9行
case 0x0130 : mapChar = 0x0069; break;
将İ（304）转换为i（105）。注意程序中是16进制的。
再看大写转换：
1 int toUpperCase(int ch) { 2 int mapChar = ch; 3 int val = getProperties(ch); 4 5 if ((val & 0x00010000) != 0) { 6 if ((val & 0x07FC0000) == 0x07FC0000) { 7 switch(ch) { 8 // map chars with overflow offsets 9 case 0x00B5 : mapChar = 0x039C; break; 10 case 0x017F : mapChar = 0x0053; break; 11 case 0x1FBE : mapChar = 0x0399; break; 12 // map char that have both a 1:1 and 1:M map 13 case 0x1F80 : mapChar = 0x1F88; break; 14 case 0x1F81 : mapChar = 0x1F89; break; 15 case 0x1F82 : mapChar = 0x1F8A; break; 16 case 0x1F83 : mapChar = 0x1F8B; break; 17 case 0x1F84 : mapChar = 0x1F8C; break; 18 case 0x1F85 : mapChar = 0x1F8D; break; 19 case 0x1F86 : mapChar = 0x1F8E; break; 20 case 0x1F87 : mapChar = 0x1F8F; break; 21 case 0x1F90 : mapChar = 0x1F98; break; 22 case 0x1F91 : mapChar = 0x1F99; break; 23 case 0x1F92 : mapChar = 0x1F9A; break; 24 case 0x1F93 : mapChar = 0x1F9B; break; 25 case 0x1F94 : mapChar = 0x1F9C; break; 26 case 0x1F95 : mapChar = 0x1F9D; break; 27 case 0x1F96 : mapChar = 0x1F9E; break; 28 case 0x1F97 : mapChar = 0x1F9F; break; 29 case 0x1FA0 : mapChar = 0x1FA8; break; 30 case 0x1FA1 : mapChar = 0x1FA9; break; 31 case 0x1FA2 : mapChar = 0x1FAA; break; 32 case 0x1FA3 : mapChar = 0x1FAB; break; 33 case 0x1FA4 : mapChar = 0x1FAC; break; 34 case 0x1FA5 : mapChar = 0x1FAD; break; 35 case 0x1FA6 : mapChar = 0x1FAE; break; 36 case 0x1FA7 : mapChar = 0x1FAF; break; 37 case 0x1FB3 : mapChar = 0x1FBC; break; 38 case 0x1FC3 : mapChar = 0x1FCC; break; 39 case 0x1FF3 : mapChar = 0x1FFC; break; 40 41 case 0x023F : mapChar = 0x2C7E; break; 42 case 0x0240 : mapChar = 0x2C7F; break; 43 case 0x0250 : mapChar = 0x2C6F; break; 44 case 0x0251 : mapChar = 0x2C6D; break; 45 case 0x0252 : mapChar = 0x2C70; break; 46 case 0x0265 : mapChar = 0xA78D; break; 47 case 0x0266 : mapChar = 0xA7AA; break; 48 case 0x026B : mapChar = 0x2C62; break; 49 case 0x0271 : mapChar = 0x2C6E; break; 50 case 0x027D : mapChar = 0x2C64; break; 51 case 0x1D79 : mapChar = 0xA77D; break; 52 case 0x1D7D : mapChar = 0x2C63; break; 53 case 0x2C65 : mapChar = 0x023A; break; 54 case 0x2C66 : mapChar = 0x023E; break; 55 case 0x2D00 : mapChar = 0x10A0; break; 56 case 0x2D01 : mapChar = 0x10A1; break; 57 case 0x2D02 : mapChar = 0x10A2; break; 58 case 0x2D03 : mapChar = 0x10A3; break; 59 case 0x2D04 : mapChar = 0x10A4; break; 60 case 0x2D05 : mapChar = 0x10A5; break; 61 case 0x2D06 : mapChar = 0x10A6; break; 62 case 0x2D07 : mapChar = 0x10A7; break; 63 case 0x2D08 : mapChar = 0x10A8; break; 64 case 0x2D09 : mapChar = 0x10A9; break; 65 case 0x2D0A : mapChar = 0x10AA; break; 66 case 0x2D0B : mapChar = 0x10AB; break; 67 case 0x2D0C : mapChar = 0x10AC; break; 68 case 0x2D0D : mapChar = 0x10AD; break; 69 case 0x2D0E : mapChar = 0x10AE; break; 70 case 0x2D0F : mapChar = 0x10AF; break; 71 case 0x2D10 : mapChar = 0x10B0; break; 72 case 0x2D11 : mapChar = 0x10B1; break; 73 case 0x2D12 : mapChar = 0x10B2; break; 74 case 0x2D13 : mapChar = 0x10B3; break; 75 case 0x2D14 : mapChar = 0x10B4; break; 76 case 0x2D15 : mapChar = 0x10B5; break; 77 case 0x2D16 : mapChar = 0x10B6; break; 78 case 0x2D17 : mapChar = 0x10B7; break; 79 case 0x2D18 : mapChar = 0x10B8; break; 80 case 0x2D19 : mapChar = 0x10B9; break; 81 case 0x2D1A : mapChar = 0x10BA; break; 82 case 0x2D1B : mapChar = 0x10BB; break; 83 case 0x2D1C : mapChar = 0x10BC; break; 84 case 0x2D1D : mapChar = 0x10BD; break; 85 case 0x2D1E : mapChar = 0x10BE; break; 86 case 0x2D1F : mapChar = 0x10BF; break; 87 case 0x2D20 : mapChar = 0x10C0; break; 88 case 0x2D21 : mapChar = 0x10C1; break; 89 case 0x2D22 : mapChar = 0x10C2; break; 90 case 0x2D23 : mapChar = 0x10C3; break; 91 case 0x2D24 : mapChar = 0x10C4; break; 92 case 0x2D25 : mapChar = 0x10C5; break; 93 case 0x2D27 : mapChar = 0x10C7; break; 94 case 0x2D2D : mapChar = 0x10CD; break; 95 // ch must have a 1:M case mapping, but we 96 // can't handle it here. Return ch. 97 // since mapChar is already set, no need 98 // to redo it here. 99 //default : mapChar = ch; 100 } 101 } 102 else { 103 int offset = val << 5 >> (5+18); 104 mapChar = ch - offset; 105 } 106 } 107 return mapChar; 108 }
转换ı（305）时，程序跳到了第103行：
int offset = val << 5 >> (5+18);
将其转换为I（73）。
至此，上面的例子可以正常运行了。
总结
对于Java:
1. 对于Ascii码表中的字符，传统方法（只转换为大写或小写）完全没有问题；
2. 若要考虑更多字符集，需多加考虑，这时要多加一次转换和比较。除了文中列举的字符，还有其他字符存在类似的问题。
对于C++:
1. 对于Ascii码表中的字符，传统方法（只转换为大写或小写）完全没有问题；
2. C++对于超出Ascii码表的字符处理方式和Java不同。由于看不到tolower的源码，这里没有进一步分析，有知晓的读者欢迎留言。
后记
文中涉及到了“等价”和“相等”的概念，这里不做具体区分，可参考《Effective C++》详细了解。
参考话题
https://stackoverflow.com/questions/15518731/understanding-logic-in-caseinsensitivecomparator
版权声明：本文为xiaoxi666原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。