在 C # 中将包含命令行参数的字符串拆分为 string []

我有一个包含要传递给另一个可执行文件的命令行参数的字符串,我需要提取包含单个参数的字符串[] ,方式与在命令行上指定命令时 C # 的方式相同。当通过反射执行另一个程序集入口点时,将使用字符串[]。

这个有标准函数吗?或者有一个首选的方法(regex?)正确地分解参数?它必须处理“”’分隔的字符串,这些字符串可能正确地包含了空格,因此我不能只在“’上分割。”。

示例字符串:

string parameterString = @"/src:""C:\tmp\Some Folder\Sub Folder"" /users:""abcdefg@hijkl.com"" tasks:""SomeTask,Some Other Task"" -someParam foo";

示例结果:

string[] parameterArray = new string[] {
@"/src:C:\tmp\Some Folder\Sub Folder",
@"/users:abcdefg@hijkl.com",
@"tasks:SomeTask,Some Other Task",
@"-someParam",
@"foo"
};

我不需要命令行解析库,只需要一种获取应该生成的 String []的方法。

更新 : 我必须更改预期的结果,以匹配 C # 实际生成的内容(删除了拆分字符串中多余的“ s”)

102581 次浏览

Yes, the string object has a built in function called Split() that takes a single parameter specifying the character to look for as a delimiter, and returns an array of strings (string[]) with the individual values in it.

I am not sure if I understood you, but is the problem that the character used as splitter, is also to be found inside the text? (Except for that it is escaped with double "?)

If so, I would create a for loop, and replace all instances where <"> is present with <|> (or another "safe" character, but make sure that it only replaces <">, and not <"">

After iterating the string, I would do as previously posted, split the string, but now on the character <|>.

This The Code Project article is what I've used in the past. It's a good bit of code, but it might work.

This MSDN article is the only thing I could find that explains how C# parses command line arguments.

The Windows command-line parser behaves just as you say, split on space unless there's a unclosed quote before it. I would recommend writing the parser yourself. Something like this maybe:

    static string[] ParseArguments(string commandLine)
{
char[] parmChars = commandLine.ToCharArray();
bool inQuote = false;
for (int index = 0; index < parmChars.Length; index++)
{
if (parmChars[index] == '"')
inQuote = !inQuote;
if (!inQuote && parmChars[index] == ' ')
parmChars[index] = '\n';
}
return (new string(parmChars)).Split('\n');
}

It annoys me that there's no function to split a string based on a function that examines each character. If there was, you could write it like this:

    public static IEnumerable<string> SplitCommandLine(string commandLine)
{
bool inQuotes = false;


return commandLine.Split(c =>
{
if (c == '\"')
inQuotes = !inQuotes;


return !inQuotes && c == ' ';
})
.Select(arg => arg.Trim().TrimMatchingQuotes('\"'))
.Where(arg => !string.IsNullOrEmpty(arg));
}

Although having written that, why not write the necessary extension methods. Okay, you talked me into it...

Firstly, my own version of Split that takes a function that has to decide whether the specified character should split the string:

    public static IEnumerable<string> Split(this string str,
Func<char, bool> controller)
{
int nextPiece = 0;


for (int c = 0; c < str.Length; c++)
{
if (controller(str[c]))
{
yield return str.Substring(nextPiece, c - nextPiece);
nextPiece = c + 1;
}
}


yield return str.Substring(nextPiece);
}

It may yield some empty strings depending on the situation, but maybe that information will be useful in other cases, so I don't remove the empty entries in this function.

Secondly (and more mundanely) a little helper that will trim a matching pair of quotes from the start and end of a string. It's more fussy than the standard Trim method - it will only trim one character from each end, and it will not trim from just one end:

    public static string TrimMatchingQuotes(this string input, char quote)
{
if ((input.Length >= 2) &&
(input[0] == quote) && (input[input.Length - 1] == quote))
return input.Substring(1, input.Length - 2);


return input;
}

And I suppose you'll want some tests as well. Well, alright then. But this must be absolutely the last thing! First a helper function that compares the result of the split with the expected array contents:

    public static void Test(string cmdLine, params string[] args)
{
string[] split = SplitCommandLine(cmdLine).ToArray();


Debug.Assert(split.Length == args.Length);


for (int n = 0; n < split.Length; n++)
Debug.Assert(split[n] == args[n]);
}

Then I can write tests like this:

        Test("");
Test("a", "a");
Test(" abc ", "abc");
Test("a b ", "a", "b");
Test("a b \"c d\"", "a", "b", "c d");

Here's the test for your requirements:

        Test(@"/src:""C:\tmp\Some Folder\Sub Folder"" /users:""abcdefg@hijkl.com"" tasks:""SomeTask,Some Other Task"" -someParam",
@"/src:""C:\tmp\Some Folder\Sub Folder""", @"/users:""abcdefg@hijkl.com""", @"tasks:""SomeTask,Some Other Task""", @"-someParam");

Note that the implementation has the extra feature that it will remove quotes around an argument if that makes sense (thanks to the TrimMatchingQuotes function). I believe that's part of the normal command-line interpretation.

Currently, this is the code that I have:

    private String[] SplitCommandLineArgument(String argumentString)
{
StringBuilder translatedArguments = new StringBuilder(argumentString);
bool escaped = false;
for (int i = 0; i < translatedArguments.Length; i++)
{
if (translatedArguments[i] == '"')
{
escaped = !escaped;
}
if (translatedArguments[i] == ' ' && !escaped)
{
translatedArguments[i] = '\n';
}
}


string[] toReturn = translatedArguments.ToString().Split(new char[] { '\n' }, StringSplitOptions.RemoveEmptyEntries);
for(int i = 0; i < toReturn.Length; i++)
{
toReturn[i] = RemoveMatchingQuotes(toReturn[i]);
}
return toReturn;
}


public static string RemoveMatchingQuotes(string stringToTrim)
{
int firstQuoteIndex = stringToTrim.IndexOf('"');
int lastQuoteIndex = stringToTrim.LastIndexOf('"');
while (firstQuoteIndex != lastQuoteIndex)
{
stringToTrim = stringToTrim.Remove(firstQuoteIndex, 1);
stringToTrim = stringToTrim.Remove(lastQuoteIndex - 1, 1); //-1 because we've shifted the indicies left by one
firstQuoteIndex = stringToTrim.IndexOf('"');
lastQuoteIndex = stringToTrim.LastIndexOf('"');
}
return stringToTrim;
}

It doesn't work with escaped quotes, but it works for the cases that I've come up against so far.

This is a reply to Anton's code, which do not work with escaped quotes. I modified 3 places.

  1. The constructor for StringBuilder in SplitCommandLineArguments, replacing any \" with \r
  2. In the for-loop in SplitCommandLineArguments, I now replace the \r character back to \".
  3. Changed the SplitCommandLineArgument method from private to public static.

public static string[] SplitCommandLineArgument( String argumentString )
{
StringBuilder translatedArguments = new StringBuilder( argumentString ).Replace( "\\\"", "\r" );
bool InsideQuote = false;
for ( int i = 0; i < translatedArguments.Length; i++ )
{
if ( translatedArguments[i] == '"' )
{
InsideQuote = !InsideQuote;
}
if ( translatedArguments[i] == ' ' && !InsideQuote )
{
translatedArguments[i] = '\n';
}
}


string[] toReturn = translatedArguments.ToString().Split( new char[] { '\n' }, StringSplitOptions.RemoveEmptyEntries );
for ( int i = 0; i < toReturn.Length; i++ )
{
toReturn[i] = RemoveMatchingQuotes( toReturn[i] );
toReturn[i] = toReturn[i].Replace( "\r", "\"" );
}
return toReturn;
}


public static string RemoveMatchingQuotes( string stringToTrim )
{
int firstQuoteIndex = stringToTrim.IndexOf( '"' );
int lastQuoteIndex = stringToTrim.LastIndexOf( '"' );
while ( firstQuoteIndex != lastQuoteIndex )
{
stringToTrim = stringToTrim.Remove( firstQuoteIndex, 1 );
stringToTrim = stringToTrim.Remove( lastQuoteIndex - 1, 1 ); //-1 because we've shifted the indicies left by one
firstQuoteIndex = stringToTrim.IndexOf( '"' );
lastQuoteIndex = stringToTrim.LastIndexOf( '"' );
}
return stringToTrim;
}

In addition to the good and pure managed solution by Earwicker, it may be worth mentioning, for sake of completeness, that Windows also provides the CommandLineToArgvW function for breaking up a string into an array of strings:

LPWSTR *CommandLineToArgvW(
LPCWSTR lpCmdLine, int *pNumArgs);

Parses a Unicode command line string and returns an array of pointers to the command line arguments, along with a count of such arguments, in a way that is similar to the standard C run-time argv and argc values.

An example of calling this API from C# and unpacking the resulting string array in managed code can be found at, “Converting Command Line String to Args[] using CommandLineToArgvW() API.” Below is a slightly simpler version of the same code:

[DllImport("shell32.dll", SetLastError = true)]
static extern IntPtr CommandLineToArgvW(
[MarshalAs(UnmanagedType.LPWStr)] string lpCmdLine, out int pNumArgs);


public static string[] CommandLineToArgs(string commandLine)
{
int argc;
var argv = CommandLineToArgvW(commandLine, out argc);
if (argv == IntPtr.Zero)
throw new System.ComponentModel.Win32Exception();
try
{
var args = new string[argc];
for (var i = 0; i < args.Length; i++)
{
var p = Marshal.ReadIntPtr(argv, i * IntPtr.Size);
args[i] = Marshal.PtrToStringUni(p);
}


return args;
}
finally
{
Marshal.FreeHGlobal(argv);
}
}

I took the answer from Jeffrey L Whitledge and enhanced it a little.

It now supports both single and double quotes. You can use quotes in the parameters itself by using other typed quotes.

It also strips the quotes from the arguments since these do not contribute to the argument information.

    public static string[] SplitArguments(string commandLine)
{
var parmChars = commandLine.ToCharArray();
var inSingleQuote = false;
var inDoubleQuote = false;
for (var index = 0; index < parmChars.Length; index++)
{
if (parmChars[index] == '"' && !inSingleQuote)
{
inDoubleQuote = !inDoubleQuote;
parmChars[index] = '\n';
}
if (parmChars[index] == '\'' && !inDoubleQuote)
{
inSingleQuote = !inSingleQuote;
parmChars[index] = '\n';
}
if (!inSingleQuote && !inDoubleQuote && parmChars[index] == ' ')
parmChars[index] = '\n';
}
return (new string(parmChars)).Split(new[] { '\n' }, StringSplitOptions.RemoveEmptyEntries);
}

I like iterators, and nowadays LINQ makes IEnumerable<String> as easily usable as arrays of string, so my take following the spirit of Jeffrey L Whitledge's answer is (as a extension method to string):

public static IEnumerable<string> ParseArguments(this string commandLine)
{
if (string.IsNullOrWhiteSpace(commandLine))
yield break;


var sb = new StringBuilder();
bool inQuote = false;
foreach (char c in commandLine) {
if (c == '"' && !inQuote) {
inQuote = true;
continue;
}


if (c != '"' && !(char.IsWhiteSpace(c) && !inQuote)) {
sb.Append(c);
continue;
}


if (sb.Length > 0) {
var result = sb.ToString();
sb.Clear();
inQuote = false;
yield return result;
}
}


if (sb.Length > 0)
yield return sb.ToString();
}

You can have a look at the code I've posted yesterday:

[C#] Path & arguments strings

It splits a filename + arguments into string[]. Short paths, environment variables, and missing file extensions are handled.

(Initially it was for UninstallString in Registry.)

In your question you asked for a regex, and I am a big fan and user of them, so when I needed to do this same argument split as you, I wrote my own regex after googling around and not finding a simple solution. I like short solutions, so I made one and here it is:

            var re = @"\G(""((""""|[^""])+)""|(\S+)) *";
var ms = Regex.Matches(CmdLine, re);
var list = ms.Cast<Match>()
.Select(m => Regex.Replace(
m.Groups[2].Success
? m.Groups[2].Value
: m.Groups[4].Value, @"""""", @"""")).ToArray();

It handles blanks and quotes inside quotation marks, and converts enclosed "" to ". Feel free to use the code!

Use:

public static string[] SplitArguments(string args) {
char[] parmChars = args.ToCharArray();
bool inSingleQuote = false;
bool inDoubleQuote = false;
bool escaped = false;
bool lastSplitted = false;
bool justSplitted = false;
bool lastQuoted = false;
bool justQuoted = false;


int i, j;


for(i=0, j=0; i<parmChars.Length; i++, j++) {
parmChars[j] = parmChars[i];


if(!escaped) {
if(parmChars[i] == '^') {
escaped = true;
j--;
} else if(parmChars[i] == '"' && !inSingleQuote) {
inDoubleQuote = !inDoubleQuote;
parmChars[j] = '\n';
justSplitted = true;
justQuoted = true;
} else if(parmChars[i] == '\'' && !inDoubleQuote) {
inSingleQuote = !inSingleQuote;
parmChars[j] = '\n';
justSplitted = true;
justQuoted = true;
} else if(!inSingleQuote && !inDoubleQuote && parmChars[i] == ' ') {
parmChars[j] = '\n';
justSplitted = true;
}


if(justSplitted && lastSplitted && (!lastQuoted || !justQuoted))
j--;


lastSplitted = justSplitted;
justSplitted = false;


lastQuoted = justQuoted;
justQuoted = false;
} else {
escaped = false;
}
}


if(lastQuoted)
j--;


return (new string(parmChars, 0, j)).Split(new[] { '\n' });
}

Based on Vapour in the Alley's answer, this one also supports ^ escapes.

Examples:

  • this is a test
    • this
    • is
    • a
    • test
  • this "is a" test
    • this
    • is a
    • test
  • this ^"is a^" test
    • this
    • "is
    • a"
    • test
  • this "" "is a ^^ test"
    • this
    • is a ^ test

It also supports multiple spaces (breaks arguments just one time per block of spaces).

A purely managed solution might be helpful. There are too many "problem" comments for the WINAPI function and it's not available on other platforms. Here's my code that has a well-defined behaviour (that you can change if you like).

It should do the same as what .NET/Windows do when providing that string[] args parameter, and I've compared it with a number of "interesting" values.

This is a classic state-machine implementation that takes each single character from the input string and interprets it for the current state, producing output and a new state. The state is defined in the variables escape, inQuote, hadQuote and prevCh, and the output is collected in currentArg and args.

Some of the specialties that I've discovered by experiments on a real command prompt (Windows 7): \\ produces \, \" produces ", "" within a quoted range produces ".

The ^ character seems to be magical, too: it always disappears when not doubling it. Otherwise it has no effect on a real command line. My implementation does not support this, as I haven't found a pattern in this behaviour. Maybe somebody knows more about it.

Something that doesn't fit in this pattern is the following command:

cmd /c "argdump.exe "a b c""

The cmd command seems to catch the outer quotes and take the rest verbatim. There must be some special magic sauce in this.

I've done no benchmarks on my method, but consider it reasonably fast. It doesn't use Regex and doesn't do any string concatenation but instead uses a StringBuilder to collect the characters for an argument and puts them in a list.

/// <summary>
/// Reads command line arguments from a single string.
/// </summary>
/// <param name="argsString">The string that contains the entire command line.</param>
/// <returns>An array of the parsed arguments.</returns>
public string[] ReadArgs(string argsString)
{
// Collects the split argument strings
List<string> args = new List<string>();
// Builds the current argument
var currentArg = new StringBuilder();
// Indicates whether the last character was a backslash escape character
bool escape = false;
// Indicates whether we're in a quoted range
bool inQuote = false;
// Indicates whether there were quotes in the current arguments
bool hadQuote = false;
// Remembers the previous character
char prevCh = '\0';
// Iterate all characters from the input string
for (int i = 0; i < argsString.Length; i++)
{
char ch = argsString[i];
if (ch == '\\' && !escape)
{
// Beginning of a backslash-escape sequence
escape = true;
}
else if (ch == '\\' && escape)
{
// Double backslash, keep one
currentArg.Append(ch);
escape = false;
}
else if (ch == '"' && !escape)
{
// Toggle quoted range
inQuote = !inQuote;
hadQuote = true;
if (inQuote && prevCh == '"')
{
// Doubled quote within a quoted range is like escaping
currentArg.Append(ch);
}
}
else if (ch == '"' && escape)
{
// Backslash-escaped quote, keep it
currentArg.Append(ch);
escape = false;
}
else if (char.IsWhiteSpace(ch) && !inQuote)
{
if (escape)
{
// Add pending escape char
currentArg.Append('\\');
escape = false;
}
// Accept empty arguments only if they are quoted
if (currentArg.Length > 0 || hadQuote)
{
args.Add(currentArg.ToString());
}
// Reset for next argument
currentArg.Clear();
hadQuote = false;
}
else
{
if (escape)
{
// Add pending escape char
currentArg.Append('\\');
escape = false;
}
// Copy character from input, no special meaning
currentArg.Append(ch);
}
prevCh = ch;
}
// Save last argument
if (currentArg.Length > 0 || hadQuote)
{
args.Add(currentArg.ToString());
}
return args.ToArray();
}

The good and pure managed solution by Earwicker failed to handle arguments like this:

Test("\"He whispered to her \\\"I love you\\\".\"", "He whispered to her \"I love you\".");

It returned 3 elements:

"He whispered to her \"I
love
you\"."

So here is a fix to support the "quoted \"escape\" quote":

public static IEnumerable<string> SplitCommandLine(string commandLine)
{
bool inQuotes = false;
bool isEscaping = false;


return commandLine.Split(c => {
if (c == '\\' && !isEscaping) { isEscaping = true; return false; }


if (c == '\"' && !isEscaping)
inQuotes = !inQuotes;


isEscaping = false;


return !inQuotes && Char.IsWhiteSpace(c)/*c == ' '*/;
})
.Select(arg => arg.Trim().TrimMatchingQuotes('\"').Replace("\\\"", "\""))
.Where(arg => !string.IsNullOrEmpty(arg));
}

Tested with 2 additional cases:

Test("\"C:\\Program Files\"", "C:\\Program Files");
Test("\"He whispered to her \\\"I love you\\\".\"", "He whispered to her \"I love you\".");

Also noted that the accepted answer by Atif Aziz which uses CommandLineToArgvW also failed. It returned 4 elements:

He whispered to her \
I
love
you".

Hope this helps someone looking for such a solution in the future.

Try this code:

    string[] str_para_linha_comando(string str, out int argumentos)
{
string[] linhaComando = new string[32];
bool entre_aspas = false;
int posicao_ponteiro = 0;
int argc = 0;
int inicio = 0;
int fim = 0;
string sub;


for(int i = 0; i < str.Length;)
{
if (entre_aspas)
{
// Está entre aspas
sub = str.Substring(inicio+1, fim - (inicio+1));
linhaComando[argc - 1] = sub;


posicao_ponteiro += ((fim - posicao_ponteiro)+1);
entre_aspas = false;
i = posicao_ponteiro;
}
else
{
tratar_aspas:
if (str.ElementAt(i) == '\"')
{
inicio = i;
fim = str.IndexOf('\"', inicio + 1);
entre_aspas = true;
argc++;
}
else
{
// Se não for aspas, então ler até achar o primeiro espaço em branco
if (str.ElementAt(i) == ' ')
{
if (str.ElementAt(i + 1) == '\"')
{
i++;
goto tratar_aspas;
}


// Pular os espaços em branco adiconais
while(str.ElementAt(i) == ' ') i++;


argc++;
inicio = i;
fim = str.IndexOf(' ', inicio);
if (fim == -1) fim = str.Length;
sub = str.Substring(inicio, fim - inicio);
linhaComando[argc - 1] = sub;
posicao_ponteiro += (fim - posicao_ponteiro);


i = posicao_ponteiro;
if (posicao_ponteiro == str.Length) break;
}
else
{
argc++;
inicio = i;
fim = str.IndexOf(' ', inicio);
if (fim == -1) fim = str.Length;


sub = str.Substring(inicio, fim - inicio);
linhaComando[argc - 1] = sub;
posicao_ponteiro += fim - posicao_ponteiro;
i = posicao_ponteiro;
if (posicao_ponteiro == str.Length) break;
}
}
}
}


argumentos = argc;


return linhaComando;
}

It's written in Portuguese.

Here's a one liner that gets the job done (see the one line that does all of the work inside the BurstCmdLineArgs(...) method).

Not what I'd call the most readable line of code, but you can break it out for readability's sake. It's simple on purpose and does not work well for all argument cases (like file name arguments that contain the split string character delimiter in them).

This solution has worked well in my solutions that use it. Like I said, it gets the job done without a rat's nest of code to handle every possible argument format n-factorial.

using System;
using System.Collections.Generic;
using System.Linq;


namespace CmdArgProcessor
{
class Program
{
static void Main(string[] args)
{
// test switches and switches with values
// -test1 1 -test2 2 -test3 -test4 -test5 5


string dummyString = string.Empty;


var argDict = BurstCmdLineArgs(args);


Console.WriteLine("Value for switch = -test1: {0}", argDict["test1"]);
Console.WriteLine("Value for switch = -test2: {0}", argDict["test2"]);
Console.WriteLine("Switch -test3 is present? {0}", argDict.TryGetValue("test3", out dummyString));
Console.WriteLine("Switch -test4 is present? {0}", argDict.TryGetValue("test4", out dummyString));
Console.WriteLine("Value for switch = -test5: {0}", argDict["test5"]);


// Console output:
//
// Value for switch = -test1: 1
// Value for switch = -test2: 2
// Switch -test3 is present? True
// Switch -test4 is present? True
// Value for switch = -test5: 5
}


public static Dictionary<string, string> BurstCmdLineArgs(string[] args)
{
var argDict = new Dictionary<string, string>();


// Flatten the args in to a single string separated by a space.
// Then split the args on the dash delimiter of a cmd line "switch".
// E.g. -mySwitch myValue
//  or -JustMySwitch (no value)
//  where: all values must follow a switch.
// Then loop through each string returned by the split operation.
// If the string can be split again by a space character,
// then the second string is a value to be paired with a switch,
// otherwise, only the switch is added as a key with an empty string as the value.
// Use dictionary indexer to retrieve values for cmd line switches.
// Use Dictionary::ContainsKey(...) where only a switch is recorded as the key.
string.Join(" ", args).Split('-').ToList().ForEach(s => argDict.Add(s.Split()[0], (s.Split().Count() > 1 ? s.Split()[1] : "")));


return argDict;
}
}
}

I don't think there are single quotes or ^ quotes for C# applications. The following function is working fine for me:

public static IEnumerable<String> SplitArguments(string commandLine)
{
Char quoteChar = '"';
Char escapeChar = '\\';
Boolean insideQuote = false;
Boolean insideEscape = false;


StringBuilder currentArg = new StringBuilder();


// needed to keep "" as argument but drop whitespaces between arguments
Int32 currentArgCharCount = 0;


for (Int32 i = 0; i < commandLine.Length; i++)
{
Char c = commandLine[i];
if (c == quoteChar)
{
currentArgCharCount++;


if (insideEscape)
{
currentArg.Append(c);       // found \" -> add " to arg
insideEscape = false;
}
else if (insideQuote)
{
insideQuote = false;        // quote ended
}
else
{
insideQuote = true;         // quote started
}
}
else if (c == escapeChar)
{
currentArgCharCount++;


if (insideEscape)   // found \\ -> add \\ (only \" will be ")
currentArg.Append(escapeChar + escapeChar);


insideEscape = !insideEscape;
}
else if (Char.IsWhiteSpace(c))
{
if (insideQuote)
{
currentArgCharCount++;
currentArg.Append(c);       // append whitespace inside quote
}
else
{
if (currentArgCharCount > 0)
yield return currentArg.ToString();


currentArgCharCount = 0;
currentArg.Clear();
}
}
else
{
currentArgCharCount++;
if (insideEscape)
{
// found non-escaping backslash -> add \ (only \" will be ")
currentArg.Append(escapeChar);
currentArgCharCount = 0;
insideEscape = false;
}
currentArg.Append(c);
}
}


if (currentArgCharCount > 0)
yield return currentArg.ToString();
}

Oh heck. It's all ... Eugh. But this is legit official. From Microsoft in C# for .NET Core, maybe windows only, maybe cross-platform, but MIT licensed.

Select tidbits, method declarations and notable comments;

internal static unsafe string[] InternalCreateCommandLine(bool includeArg0)
private static unsafe int SegmentCommandLine(char * pCmdLine, string[] argArray, bool includeArg0)
private static unsafe int ScanArgument0(ref char* psrc, char[] arg)
private static unsafe int ScanArgument(ref char* psrc, ref bool inquote, char[] arg)

-

// First, parse the program name (argv[0]). Argv[0] is parsed under special rules. Anything up to
// the first whitespace outside a quoted subtring is accepted. Backslashes are treated as normal
// characters.

-

// Rules: 2N backslashes + " ==> N backslashes and begin/end quote
//      2N+1 backslashes + " ==> N backslashes + literal "
//         N backslashes     ==> N backslashes

This is code ported to .NET Core from .NET Framework from what I assume is either the MSVC C library or CommandLineToArgvW.

Here's my half-hearted attempt at handling some of the shenanigans with Regular Expressions, and ignoring the argument zero bit. It's a little bit wizardy.

private static readonly Regex RxWinArgs
= new Regex("([^\\s\"]+\"|((?<=\\s|^)(?!\"\"(?!\"))\")+)(\"\"|.*?)*\"[^\\s\"]*|[^\\s]+",
RegexOptions.Compiled
| RegexOptions.Singleline
| RegexOptions.ExplicitCapture
| RegexOptions.CultureInvariant);


internal static IEnumerable<string> ParseArgumentsWindows(string args) {
var match = RxWinArgs.Match(args);


while (match.Success) {
yield return match.Value;
match = match.NextMatch();
}
}

Tested it a fair bit on wacky generated output. It's output matches a fair percentage of what the monkeys typed up and ran through CommandLineToArgvW.

Couldn't find anything I liked here. I hate to mess up the stack with yield magic for a small command-line (if it were a stream of a terabyte, it would be another story).

Here's my take, it supports quote escapes with double quotes like these:

param="a 15"" screen isn't bad" param2='a 15" screen isn''t bad' param3="" param4= /param5

result:

param="a 15" screen isn't bad"

param2='a 15" screen isn't bad'

param3=""

param4=

/param5

public static string[] SplitArguments(string commandLine)
{
List<string> args         = new List<string>();
List<char>   currentArg   = new List<char>();
char?        quoteSection = null; // Keeps track of a quoted section (and the type of quote that was used to open it)
char[]       quoteChars   = new[] {'\'', '\"'};
char         previous     = ' '; // Used for escaping double quotes


for (var index = 0; index < commandLine.Length; index++)
{
char c = commandLine[index];
if (quoteChars.Contains(c))
{
if (previous == c) // Escape sequence detected
{
previous = ' '; // Prevent re-escaping
if (!quoteSection.HasValue)
{
quoteSection = c; // oops, we ended the quoted section prematurely
continue;         // don't add the 2nd quote (un-escape)
}


if (quoteSection.Value == c)
quoteSection = null; // appears to be an empty string (not an escape sequence)
}
else if (quoteSection.HasValue)
{
if (quoteSection == c)
quoteSection = null; // End quoted section
}
else
quoteSection = c; // Start quoted section
}
else if (char.IsWhiteSpace(c))
{
if (!quoteSection.HasValue)
{
args.Add(new string(currentArg.ToArray()));
currentArg.Clear();
previous = c;
continue;
}
}


currentArg.Add(c);
previous = c;
}


if (currentArg.Count > 0)
args.Add(new string(currentArg.ToArray()));


return args.ToArray();
}

I have implemented state machine to have same parser results as if args would be passed into .NET application and processed in static void Main(string[] args) method.

    public static IList<string> ParseCommandLineArgsString(string commandLineArgsString)
{
List<string> args = new List<string>();


commandLineArgsString = commandLineArgsString.Trim();
if (commandLineArgsString.Length == 0)
return args;


int index = 0;
while (index != commandLineArgsString.Length)
{
args.Add(ReadOneArgFromCommandLineArgsString(commandLineArgsString, ref index));
}


return args;
}


private static string ReadOneArgFromCommandLineArgsString(string line, ref int index)
{
if (index >= line.Length)
return string.Empty;


var sb = new StringBuilder(512);
int state = 0;
while (true)
{
char c = line[index];
index++;
switch (state)
{
case 0: //string outside quotation marks
if (c == '\\') //possible escaping character for quotation mark otherwise normal character
{
state = 1;
}
else if (c == '"') //opening quotation mark for string between quotation marks
{
state = 2;
}
else if (c == ' ') //closing arg
{
return sb.ToString();
}
else
{
sb.Append(c);
}


break;
case 1: //possible escaping \ for quotation mark or normal character
if (c == '"') //If escaping quotation mark only quotation mark is added into result
{
state = 0;
sb.Append(c);
}
else // \ works as not-special character
{
state = 0;
sb.Append('\\');
index--;
}


break;
case 2: //string between quotation marks
if (c == '"') //quotation mark in string between quotation marks can be escape mark for following quotation mark or can be ending quotation mark for string between quotation marks
{
state = 3;
}
else if (c == '\\') //escaping \ for possible following quotation mark otherwise normal character
{
state = 4;
}
else //text in quotation marks
{
sb.Append(c);
}


break;
case 3: //quotation mark in string between quotation marks
if (c == '"') //Quotation mark after quotation mark - that means that this one is escaped and can added into result and we will stay in string between quotation marks state
{
state = 2;
sb.Append(c);
}
else //we had two consecutive quotation marks - this means empty string but the following chars (until space) will be part of same arg result as well
{
state = 0;
index--;
}


break;
case 4: //possible escaping \ for quotation mark or normal character in string between quotation marks
if (c == '"') //If escaping quotation mark only quotation mark added into result
{
state = 2;
sb.Append(c);
}
else
{
state = 2;
sb.Append('\\');
index--;
}


break;
}


if (index == line.Length)
return sb.ToString();
}
}

Here is the solution which treats space(s) (single or multiple spaces) as command line parameter separator and returns the real command line arguments:

static string[] ParseMultiSpacedArguments(string commandLine)
{
var isLastCharSpace = false;
char[] parmChars = commandLine.ToCharArray();
bool inQuote = false;
for (int index = 0; index < parmChars.Length; index++)
{
if (parmChars[index] == '"')
inQuote = !inQuote;
if (!inQuote && parmChars[index] == ' ' && !isLastCharSpace)
parmChars[index] = '\n';


isLastCharSpace = parmChars[index] == '\n' || parmChars[index] == ' ';
}


return (new string(parmChars)).Split('\n');
}

Because I wanted the same behavior as OP (split a string exactly the same as windows cmd would do it) I wrote a bunch of test cases and tested the here posted answers:

    Test( 0, m, "One",                    new[] { "One" });
Test( 1, m, "One ",                   new[] { "One" });
Test( 2, m, " One",                   new[] { "One" });
Test( 3, m, " One ",                  new[] { "One" });
Test( 4, m, "One Two",                new[] { "One", "Two" });
Test( 5, m, "One  Two",               new[] { "One", "Two" });
Test( 6, m, "One   Two",              new[] { "One", "Two" });
Test( 7, m, "\"One Two\"",            new[] { "One Two" });
Test( 8, m, "One \"Two Three\"",      new[] { "One", "Two Three" });
Test( 9, m, "One \"Two Three\" Four", new[] { "One", "Two Three", "Four" });
Test(10, m, "One=\"Two Three\" Four", new[] { "One=Two Three", "Four" });
Test(11, m, "One\"Two Three\" Four",  new[] { "OneTwo Three", "Four" });
Test(12, m, "One\"Two Three   Four",  new[] { "OneTwo Three   Four" });
Test(13, m, "\"One Two\"",            new[] { "One Two" });
Test(14, m, "One\" \"Two",            new[] { "One Two" });
Test(15, m, "\"One\"  \"Two\"",       new[] { "One", "Two" });
Test(16, m, "One\\\"  Two",           new[] { "One\"", "Two" });
Test(17, m, "\\\"One\\\"  Two",       new[] { "\"One\"", "Two" });
Test(18, m, "One\"",                  new[] { "One" });
Test(19, m, "\"One",                  new[] { "One" });
Test(20, m, "One \"\"",               new[] { "One", "" });
Test(21, m, "One \"",                 new[] { "One", "" });
Test(22, m, "1 A=\"B C\"=D 2",        new[] { "1", "A=B C=D", "2" });
Test(23, m, "1 A=\"B \\\" C\"=D 2",   new[] { "1", "A=B \" C=D", "2" });
Test(24, m, "1 \\A 2",                new[] { "1", "\\A", "2" });
Test(25, m, "1 \\\" 2",               new[] { "1", "\"", "2" });
Test(26, m, "1 \\\\\" 2",             new[] { "1", "\\\"", "2" });
Test(27, m, "\"",                     new[] { "" });
Test(28, m, "\\\"",                   new[] { "\"" });
Test(29, m, "'A B'",                  new[] { "'A", "B'" });
Test(30, m, "^",                      new[] { "^" });
Test(31, m, "^A",                     new[] { "A" });
Test(32, m, "^^",                     new[] { "^" });
Test(33, m, "\\^^",                   new[] { "\\^" });
Test(34, m, "^\\\\",                  new[] { "\\\\" });
Test(35, m, "^\"A B\"",               new[] { "A B" });


// Test cases Anton


Test(36, m, @"/src:""C:\tmp\Some Folder\Sub Folder"" /users:""abcdefg@hijkl.com"" tasks:""SomeTask,Some Other Task"" -someParam foo", new[] { @"/src:C:\tmp\Some Folder\Sub Folder", @"/users:abcdefg@hijkl.com", @"tasks:SomeTask,Some Other Task", @"-someParam", @"foo" });


// Test cases Daniel Earwicker


Test(37, m, "",            new string[] { });
Test(38, m, "a",           new[] { "a" });
Test(39, m, " abc ",       new[] { "abc" });
Test(40, m, "a b ",        new[] { "a", "b" });
Test(41, m, "a b \"c d\"", new[] { "a", "b", "c d" });


// Test cases Fabio Iotti


Test(42, m, "this is a test ",    new[] { "this", "is", "a", "test" });
Test(43, m, "this \"is a\" test", new[] { "this", "is a", "test" });


// Test cases Kevin Thach


Test(44, m, "\"C:\\Program Files\"",                       new[] { "C:\\Program Files" });
Test(45, m, "\"He whispered to her \\\"I love you\\\".\"", new[] { "He whispered to her \"I love you\"." });

the "expected" value comes from directly testing it with cmd.exe on my machine (Win10 x64) and a simple print program:

static void Main(string[] args) => Console.Out.WriteLine($"Count := {args.Length}\n{string.Join("\n", args.Select((v,i) => $"[{i}] => '{v}'"))}");

These are the results:


Solution                      | Failed Tests
------------------------------|-------------------------------------
Atif Aziz (749653)            | 2, 3, 10, 11, 12, 14, 16, 17, 18, 26, 28, 31, 32, 33, 34, 35, 36, 37, 39, 45
Jeffrey L Whitledge (298968)  | 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 37, 39, 40, 41, 42, 43, 44, 45
Daniel Earwicker (298990)     | 10, 11, 12, 14, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 45
Anton (299795)                | 12, 16, 17, 18, 19, 21, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 45
CS. (467313)                  | 12, 18, 19, 21, 27, 31, 32, 33, 34, 35
Vapour in the Alley (2132004) | 10, 11, 12, 14, 16, 17, 20, 21, 22, 23, 25, 26, 27, 28, 29, 31, 32, 33, 34, 35, 36, 45
Monoman (7774211)             | 14, 16, 17, 20, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 45
Thomas Petersson (19091999)   | 2, 3, 10, 11, 12, 14, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 39, 45
Fabio Iotti (19725880)        | 1, 2, 3, 7, 10, 11, 12, 13, 14, 15, 16, 17, 19, 21, 22, 23, 25, 26, 28, 29, 30, 35, 36, 37, 39, 40, 42, 44, 45
ygoe (23961658)               | 26, 31, 32, 33, 34, 35
Kevin Thach (24829691)        | 10, 11, 12, 14, 18, 19, 20, 21, 22, 23, 26, 27, 31, 32, 33, 34, 35, 36
Lucas De Jesus (31621370)     | 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45
HarryP (48008872)             | 24, 26, 31, 32, 33, 34, 35
TylerY86 (53290784)           | 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 41, 43, 44, 45
Louis Somers (55903304)       | 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 29, 31, 32, 33, 34, 35, 36, 39, 41, 43, 44, 45
user2126375 (58233585)        | 5, 6, 15, 16, 17, 31, 32, 33, 34, 35
DilipNannaware (59131568)     | 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 37, 39, 40, 41, 42, 43, 44, 45
Mikescher (this)              | -

Because no answer seemed correct (at least based on my use case) here is my solution, it currently passes all test cases (but if anyone has additional (failing) corner cases please comment):

public static IEnumerable<string> SplitArgs(string commandLine)
{
var result = new StringBuilder();


var quoted = false;
var escaped = false;
var started = false;
var allowcaret = false;
for (int i = 0; i < commandLine.Length; i++)
{
var chr = commandLine[i];


if (chr == '^' && !quoted)
{
if (allowcaret)
{
result.Append(chr);
started = true;
escaped = false;
allowcaret = false;
}
else if (i + 1 < commandLine.Length && commandLine[i + 1] == '^')
{
allowcaret = true;
}
else if (i + 1 == commandLine.Length)
{
result.Append(chr);
started = true;
escaped = false;
}
}
else if (escaped)
{
result.Append(chr);
started = true;
escaped = false;
}
else if (chr == '"')
{
quoted = !quoted;
started = true;
}
else if (chr == '\\' && i + 1 < commandLine.Length && commandLine[i + 1] == '"')
{
escaped = true;
}
else if (chr == ' ' && !quoted)
{
if (started) yield return result.ToString();
result.Clear();
started = false;
}
else
{
result.Append(chr);
started = true;
}
}


if (started) yield return result.ToString();
}

The code I used to generate the test results can be found here

There's a NuGet package which contains exactly the functionality you need:

Microsoft.CodeAnalysis.Common contains the class CommandLineParser with the method SplitCommandLineIntoArguments.

You use it like this:

using Microsoft.CodeAnalysis;
// [...]
var cli = @"/src:""C:\tmp\Some Folder\Sub Folder"" /users:""abcdefg@hijkl.com"" tasks:""SomeTask,Some Other Task"" -someParam foo";
var cliArgs = CommandLineParser.SplitCommandLineIntoArguments(cli, true);


Console.WriteLine(string.Join('\n', cliArgs));
// prints out:
// /src:"C:\tmp\Some Folder\Sub Folder"
// /users:"abcdefg@hijkl.com"
// tasks:"SomeTask,Some Other Task"
// -someParam
// foo

I wrote a method to separate a file name from its arguments, for use with ProcessStartInfo which requires separating the file name and argument string.

For instance "C:\Users\Me\Something.exe" -a would give { "C:\Users\Me\Something.exe", "-a" } as a result

Code below:

    public static string[] SplitCommandFromArgs(string commandLine)
{
commandLine = commandLine.Trim();
if (commandLine[0] == '"')
{
bool isEscaped = false;
for (int c = 1; c < commandLine.Length; c++)
{
if (commandLine[c] == '"' && !isEscaped)
{
return new string[] { commandLine.Substring(1, c - 1), commandLine.Substring(c + 1).Trim() };
}
isEscaped = commandLine[c] == '\\';
}
}
else
{
for (int c = 1; c < commandLine.Length; c++) {
if (commandLine[c] == ' ')
{
return new string[] { commandLine.Substring(0, c), commandLine.Substring(c).Trim() };
}
}
}
return new string[] { commandLine, "" };
}