. net框架系统。IO对于无效的文件系统字符提供如下功能:

路径。GetInvalidFileNameChars 路径。GetInvalidPathChars

这些函数应该根据. net运行时所在的平台返回适当的结果。也就是说，这些函数的文档页中的备注说:

方法返回的数组不保证包含 文件和目录中无效的完整字符集 的名字。完整的无效字符集可能因文件系统而异。

您可以使用白名单，而不是创建字符黑名单。考虑到所有因素，在文件或目录名称上下文中有意义的字符范围非常短，除非您有一些非常特定的命名要求，否则如果用户不能使用整个ASCII表，他们不会反对您的应用程序。

它不能解决目标文件系统中保留名称的问题，但是使用白名单可以更容易地降低源上的风险。

本着这种精神，这是一个可以被认为是安全的字符范围:

字母(a-z a-z) -如果需要，也可以使用Unicode字符 数字(0 - 9) 下划线(_) 连字符(-) 空间 点号(.)

以及您希望允许的任何其他安全字符。除此之外，您还必须执行一些关于空格和点的附加规则。这通常就足够了:

名称必须包含至少一个字母或数字(以避免只有点/空格) 名称必须以字母或数字开头(以避免前导点/空格) 名称不能以点或空格结尾(如果存在，只需修剪它们，就像资源管理器一样)

这已经允许相当复杂和无意义的名称。例如，在这些规则下，这些名称是可能的，并且在Windows/Linux中是有效的文件名:

一个 ........... ext B -。- - - - - - ext

从本质上讲，即使白名单上的角色很少，你仍然应该决定什么是真正有意义的，并相应地验证/调整名称。在我的一个应用程序中，我使用了与上面相同的规则，但去掉了任何重复的点和空格。

对于Windows，您可以使用PowerShell检查它

$PathInvalidChars = [System.IO.Path]::GetInvalidPathChars() #36 chars

要显示您可以转换的UTF-8代码

$enc = [system.Text.Encoding]::UTF8
$PathInvalidChars | foreach { $enc.GetBytes($_) }

$FileNameInvalidChars = [System.IO.Path]::GetInvalidFileNameChars() #41 chars

$FileOnlyInvalidChars = @(':', '*', '?', '\', '/') #5 chars - as a difference

好吧，如果只是为了研究目的，那么你最好的选择是看看维基百科上关于文件名的条目。

如果您想编写一个可移植的函数来验证用户输入并基于此创建文件名，简单的回答是不要。看一看像Perl的File::Spec这样的可移植模块，了解一下完成这样一个“简单”任务所需的所有跳转。

要让Windows告诉你答案，最简单的方法是尝试通过资源管理器重命名文件，并输入反斜杠/作为新名称。Windows会弹出一个消息框，告诉你非法字符的列表。

A filename cannot contain any of the following characters:
    \ / : * ? " < > | 

微软文档-命名文件，路径和命名空间-命名约定

讨论不同的可能方法

在定义，什么是合法的，什么是不合法的方面的困难已经得到解决，并提出了白名单。但不仅是Windows，很多unix操作系统也支持超过8位的字符，比如Unicode。您还可以在这里讨论编码，例如UTF-8。你可以考虑Jonathan Leffler的评论，他给出了现代Linux的信息，并描述了MacOS的细节。维基百科指出，(例如)

修饰语字母冒号[(见7。有时在Windows文件名中使用，因为它与用于文件名的Segoe UI字体中的冒号相同。[继承的ASCII]冒号本身是不允许的。

因此，我想提出一种更自由的方法，使用Unicode Homoglyph字符替换“非法”字符。我发现在我可比的用例中，结果可读性要强得多，而且它只受所使用字体的限制，它非常广泛，Windows默认为3903个字符。此外，您甚至可以从替换恢复原始内容。

可能的选择和研究笔记

为了保持内容的组织性，我将始终给出字符，它的名称和十六进制数表示。后者不区分大小写，前导零可以自由添加或省略，因此，例如U+002A和U+ 2a是等效的。如果可用，我会尽量指出更多的信息或替代品-请随时向我展示更多或更好的。

Instead of * (U+2A * ASTERISK), you can use one of the many listed, for example U+2217 ∗ (ASTERISK OPERATOR) or the Full Width Asterisk U+FF0A ＊. u+20f0 ⃰ combining asterisk above from combining diacritical marks for symbols might also be a valid choice. You can read 4. for more info about the combining characters. Instead of . (U+2E . full stop), one of these could be a good option, for example ⋅ U+22C5 dot operator. Instead of " (U+22 " quotation mark), you can use “ U+201C english leftdoublequotemark, more alternatives see here. I also included some of the good suggestions of Wally Brockway's answer, in this case u+2036 ‶ reversed double prime and u+2033 ″ double prime - I will from now on denote ideas from that source by ¹³. Instead of / (U+2F / SOLIDUS), you can use ∕ DIVISION SLASH U+2215 (others here), ̸ U+0338 COMBINING LONG SOLIDUS OVERLAY, ̷ COMBINING SHORT SOLIDUS OVERLAY U+0337 or u+2044 ⁄ fraction slash¹³. Be aware about spacing for some characters, including the combining or overlay ones, as they have no width and can produce something like -> ̸th̷is which is ̸th̷is. With added spaces you get -> ̸ th ̷ is, which is ̸ th ̷ is. The second one (COMBINING SHORT SOLIDUS OVERLAY) looks bad in the stackoverflow-font. Instead of \ (U+5C Reverse solidus), you can use ⧵ U+29F5 Reverse solidus operator (more) or u+20E5 ⃥ combining reverse solidus overlay¹³. To replace [ (U+5B [ Left square bracket) and ] (U+005D ] Right square bracket), you can use for example U+FF3B［ FULLWIDTH LEFT SQUARE BRACKET and U+FF3D ］FULLWIDTH RIGHT SQUARE BRACKET (from here, more possibilities here). Instead of : (u+3a : colon), you can use U+2236 ∶ RATIO (for mathematical usage) or U+A789 ꞉ MODIFIER LETTER COLON, (see colon (letter), sometimes used in Windows filenames as it is identical to the colon in the Segoe UI font used for filenames. The colon itself is not permitted ... source and more replacements see here). Another alternative is this one: u+1361 ፡ ethiopic wordspace¹³. Instead of ; (u+3b ; semicolon), you can use U+037E ; GREEK QUESTION MARK (see here). For | (u+7c | vertical line), there are some good substitutes such as: U+2223 ∣ DIVIDES, U+0964 । DEVANAGARI DANDA, U+01C0 ǀ LATIN LETTER DENTAL CLICK (the last ones from Wikipedia) or U+2D4F ⵏ Tifinagh Letter Yan. Also the box drawing characters contain various other options. Instead of , (, U+002C COMMA), you can use for example ‚ U+201A SINGLE LOW-9 QUOTATION MARK (see here). For ? (U+003F ? QUESTION MARK), these are good candidates: U+FF1F ？ FULLWIDTH QUESTION MARK or U+FE56 ﹖ SMALL QUESTION MARK (from here and here). There are also two more from the Dingbats Block (search for "question") and the u+203d ‽ interrobang¹³. While my machine seems to accept it unchanged, I still want to include > (u+3e greater-than sign) and < (u+3c less-than sign) for the sake of completeness. The best replacement here is probably also from the quotation block, such as u+203a › single right-pointing angle quotation mark and u+2039 ‹ single left-pointing angle quotation mark respectively. The tifinagh block only contains ⵦ (u+2D66)¹³ to replace <. The last notion is ⋖ less-than with dot u+22D6 and ⋗ greater-than with dot u+22D7.

对于更多的想法，你也可以在这个块中寻找例子。你还想要更多的想法吗?你可以试着画出你想要的角色，看看这里的建议。

你怎么打这些字符

Say you want to type ⵏ (Tifinagh Letter Yan). To get all of its information, you can always search for this character (ⵏ) on a suited platform such as this Unicode Lookup (please add 0x when you search for hex) or that Unicode Table (that only allows to search for the name, in this case "Tifinagh Letter Yan"). You should obtain its Unicode number U+2D4F and the HTML-code &#11599; (note that 2D4F is hexadecimal for 11599). With this knowledge, you have several options to produce these special characters including the use of

code points to unicode converter or again the Unicode Lookup to reversely convert the numerical representation into the unicode character (remember to set the code point base below to decimal or hexadecimal respectively) a one-liner makro in Autohotkey: :?*:altpipe::{U+2D4F} to type ⵏ instead of the string altpipe - this is the way I input those special characters, my Autohotkey script can be shared if there is common interest Alt Characters or alt-codes by pressing and holding alt, followed by the decimal number for the desired character (more info for example here, look at a table here or there). For the example, that would be Alt+11599. Be aware, that many programs do not fully support this windows feature for all of unicode (as of time writing). Microsoft Office is an exception where it usually works, some other OSes provide similar functionality. Typing these chars with Alt-combinations into MS Word is also the way Wally Brockway suggests in his answer¹³ that was already mentionted - if you don't want to transfer all the hexadecimal values to the decimal asc, you can find some of them there¹³. in MS Office, you can also use ALT + X as described in this MS article to produce the chars if you rarely need it, you can of course still just copy-paste the special character of your choice instead of typing it