Add winsomely that converts args to a command line (#33465)

Converts the collection of strings 'args' into a Windows command line.

Co-authored-by:    Mustafa M. <[email protected]>
Co-authored-by:    Markus Kuhn <[email protected]>
Co-authored-by:    Jameson Nash <[email protected]>

Author: 3 people

base/cmd.jl

@@ -102,6 +102,8 @@ shell_escape(cmd::Cmd; special::AbstractString="") =
     shell_escape(cmd.exec..., special=special)
 shell_escape_posixly(cmd::Cmd) =
     shell_escape_posixly(cmd.exec...)
+shell_escape_winsomely(cmd::Cmd) =
+    shell_escape_winsomely(cmd.exec...)
 
 function show(io::IO, cmd::Cmd)
     print_env = cmd.env !== nothing

base/shell.jl

@@ -253,3 +253,62 @@ julia> Base.shell_escape_posixly("echo", "this", "&&", "that")
 """
 shell_escape_posixly(args::AbstractString...) =
     sprint(print_shell_escaped_posixly, args...)
+
+
+function print_shell_escaped_winsomely(io::IO, args::AbstractString...)
+    first = true
+    for arg in args
+        first || write(io, ' ')
+        first = false
+        # Quote any arg that contains a whitespace (' ' or '\t') or a double quote mark '"'.
+        # It's also valid to quote an arg with just a whitespace,
+        # but the following may be 'safer', and both implementations are valid anyways.
+        quotes = any(c -> c in (' ', '\t', '"'), arg) || isempty(arg)
+        quotes && write(io, '"')
+        backslashes = 0
+        for c in arg
+            if c == '\\'
+                backslashes += 1
+            else
+                # escape all backslashes and the following double quote
+                c == '"' && (backslashes = backslashes * 2 + 1)
+                for j = 1:backslashes
+                    # backslashes aren't special here
+                    write(io, '\\')
+                end
+                backslashes = 0
+                write(io, c)
+            end
+        end
+        # escape all backslashes, letting the terminating double quote we add below to then be interpreted as a special char
+        quotes && (backslashes *= 2)
+        for j = 1:backslashes
+            write(io, '\\')
+        end
+        quotes && write(io, '"')
+    end
+    return nothing
+end
+
+
+"""
+     shell_escaped_winsomely(args::Union{Cmd,AbstractString...})::String
+
+Convert the collection of strings `args` into single string suitable for passing as the argument
+string for a Windows command line. Windows passes the entire command line as a single string to
+the application (unlike POSIX systems, where the list of arguments are passed separately).
+Many Windows API applications (including julia.exe), use the conventions of the [Microsoft C
+runtime](https://docs.microsoft.com/en-us/cpp/c-language/parsing-c-command-line-arguments) to
+split that command line into a list of strings. This function implements the inverse of such a
+C runtime command-line parser. It joins command-line arguments to be passed to a Windows console
+application into a command line, escaping or quoting meta characters such as space,
+double quotes and backslash where needed. This may be useful in concert with the `windows_verbatim`
+flag to [`Cmd`](@ref) when constructing process pipelines.
+
+# Example
+```jldoctest
+julia> println(shell_escaped_winsomely("A B\\", "C"))
+"A B\\" C
+"""
+shell_escape_winsomely(args::AbstractString...) =
+    sprint(print_shell_escaped_winsomely, args..., sizehint=(sum(length, args)) + 3*length(args))

test/spawn.jl

@@ -675,3 +675,88 @@ end
 if Sys.iswindows()
     rm(busybox, force=true)
 end
+
+
+# shell escaping on Windows
+@testset "shell_escape_winsomely" begin
+    # Note  argument A can be parsed both as A or "A".
+    # We do not test that the parsing satisfies either of these conditions.
+    # In other words, tests may fail even for valid parsing.
+    # This is done to avoid overly verbose tests.
+
+    # input :
+    # output: ""
+    @test Base.shell_escape_winsomely("") == "\"\""
+
+    @test Base.shell_escape_winsomely("A") == "A"
+
+    @test Base.shell_escape_winsomely(`A`) == "A"
+
+    # input : hello world
+    # output: "hello world"
+    @test Base.shell_escape_winsomely("hello world") == "\"hello world\""
+
+    # input : hello  world
+    # output: "hello  world"
+    @test Base.shell_escape_winsomely("hello\tworld") == "\"hello\tworld\""
+
+    # input : hello"world
+    # output: "hello\"world" (also valid) hello\"world
+    @test Base.shell_escape_winsomely("hello\"world") == "\"hello\\\"world\""
+
+    # input : hello""world
+    # output: "hello\"\"world" (also valid) hello\"\"world
+    @test Base.shell_escape_winsomely("hello\"\"world") == "\"hello\\\"\\\"world\""
+
+    # input : hello\world
+    # output: hello\world
+    @test Base.shell_escape_winsomely("hello\\world") == "hello\\world"
+
+    # input : hello\\world
+    # output: hello\\world
+    @test Base.shell_escape_winsomely("hello\\\\world") == "hello\\\\world"
+
+    # input : hello\"world
+    # output: "hello\"world" (also valid) hello\"world
+    @test Base.shell_escape_winsomely("hello\\\"world") == "\"hello\\\\\\\"world\""
+
+    # input : hello\\"world
+    # output: "hello\\\\\"world" (also valid) hello\\\\\"world
+    @test Base.shell_escape_winsomely("hello\\\\\"world")  == "\"hello\\\\\\\\\\\"world\""
+
+    # input : hello world\
+    # output: "hello world\\"
+    @test Base.shell_escape_winsomely("hello world\\") == "\"hello world\\\\\""
+
+    # input : A\B
+    # output: A\B"
+    @test Base.shell_escape_winsomely("A\\B") == "A\\B"
+
+    # input : [A\, B]
+    # output: "A\ B"
+    @test Base.shell_escape_winsomely("A\\", "B") == "A\\ B"
+
+    # input : A"B
+    # output: "A\"B"
+    @test Base.shell_escape_winsomely("A\"B") ==  "\"A\\\"B\""
+
+    # input : [A B\, C]
+    # output: "A B\\" C
+    @test Base.shell_escape_winsomely("A B\\", "C") == "\"A B\\\\\" C"
+
+    # input : [A "B, C]
+    # output: "A \"B" C
+    @test Base.shell_escape_winsomely("A \"B", "C") == "\"A \\\"B\" C"
+
+    # input : [A B\, C]
+    # output: "A B\\" C
+    @test Base.shell_escape_winsomely("A B\\", "C") == "\"A B\\\\\" C"
+
+    # input :[A\ B\, C]
+    # output: "A\ B\\" C
+    @test Base.shell_escape_winsomely("A\\ B\\", "C") == "\"A\\ B\\\\\" C"
+
+    # input : [A\ B\, C, D K]
+    # output: "A\ B\\" C "D K"
+    @test Base.shell_escape_winsomely("A\\ B\\", "C", "D K") == "\"A\\ B\\\\\" C \"D K\""
+end