mex: normalize

Author: scivision

+stdlib/normalize.m

@@ -1,11 +1,15 @@
 %% NORMALIZE remove redundant elements of path
+% optional: mex
+%
 % normalize(p) remove redundant elements of path p
 % path need not exist, normalized path is returned
 %
 %%% Inputs
 % * p: path to normalize
 %%% Outputs
 % * c: normalized path
+%
+% MEX code is about 4-5x faster than plain Matlab below
 
 function n = normalize(p)
 arguments

buildfile.m

@@ -84,7 +84,7 @@ function publishTask(context)
 function srcs = get_mex_sources()
 
 pure = "src/pure.cpp";
-normal = ["src/normalize.cpp", pure];
+normal = ["src/normalize_fs.cpp", pure];
 
 win = [pure, "src/windows.cpp"];
 
@@ -98,6 +98,7 @@ function publishTask(context)
 "src/unlink.cpp", ...
 ["src/is_admin.cpp", "src/admin_fs.cpp"] ...
 "src/is_char_device.cpp", ...
+["src/normalize.cpp", normal], ...
 ["src/parent.cpp", "src/parent_fs.cpp", normal], ...
 "src/relative_to.cpp", ...
 "src/proximate_to.cpp", ...

example/bench_normalize.m

@@ -1,12 +1,7 @@
 a1 = "c:/a///.//../////b";
 
-fno = @() stdlib.normalize(a1, false);
-fjava = @() stdlib.normalize(a1, true);
+fno = @() stdlib.normalize(a1);
 
 t_no = timeit(fno);
-t_java = timeit(fjava);
 
-disp("No Java: " + t_no + " s")
-disp("Java: " + t_java + " s")
-
-disp("Java is " + t_no/t_java + " times faster")
+disp("No Java: " + t_no + " s")

octave_build.m

@@ -15,6 +15,7 @@ function octave_build(overwrite)
 fullfile(d, "is_wsl.cpp"), ...
 fullfile(d, "is_rosetta.cpp"), ...
 fullfile(d, "is_admin.cpp"), ...
+fullfile(d, "normalize.cpp"), ...
 fullfile(d, "parent.cpp"), ...
 fullfile(d, "proximate_to.cpp"), ...
 fullfile(d, "relative_to.cpp"), ...

src/ffilesystem.h

@@ -6,6 +6,7 @@ bool fs_is_windows();
 std::size_t fs_get_max_path();
 std::string fs_as_posix(std::string_view);
 std::string fs_drop_slash(std::string_view);
+std::string fs_normalize(std::string_view path);
 std::string fs_as_posix(std::string_view);
 
 bool fs_is_url(std::string_view);

src/normalize.cpp

@@ -1,26 +1,43 @@
+#include "mex.hpp"
+#include "mexAdapter.hpp"
+
 #include <string>
-#include <string_view>
-#include <algorithm> // std::unique
 
-#include <filesystem>
+#include <vector>
+#include <memory>
 
 #include "ffilesystem.h"
 
 
-std::string fs_drop_slash(std::string_view in)
-{
-  // drop all trailing "/" and duplicated internal "/"
-
-  std::filesystem::path p(in);
-
-  std::string s = p.generic_string();
-
-  if (!fs_is_windows() || (s.length() != 3 || p != p.root_path())){
-    while(s.length() > 1 && s.back() == '/')
-      s.pop_back();
+class MexFunction : public matlab::mex::Function {
+public:
+  void operator()(matlab::mex::ArgumentList outputs, matlab::mex::ArgumentList inputs) {
+// boilerplate engine & ArrayFactory setup
+    std::shared_ptr<matlab::engine::MATLABEngine> matlabEng = getEngine();
+
+    matlab::data::ArrayFactory factory;
+
+// wrangle Inputs
+    std::string in;
+
+    if (inputs.size() != 1) {
+      matlabEng->feval(u"error", 0,
+        std::vector<matlab::data::Array>({ factory.createScalar("Mex: One input required") }));
+    }
+    if ((inputs[0].getType() == matlab::data::ArrayType::MATLAB_STRING && inputs[0].getNumberOfElements() == 1)){
+        matlab::data::TypedArray<matlab::data::MATLABString> stringArr = inputs[0];
+        in = stringArr[0];
+    } else if (inputs[0].getType() == matlab::data::ArrayType::CHAR){
+        matlab::data::CharArray charArr = inputs[0];
+        in.assign(charArr.begin(), charArr.end());
+    } else {
+        matlabEng->feval(u"error", 0,
+        std::vector<matlab::data::Array>({ factory.createScalar("Mex: First input must be a scalar string or char vector") }));
+    }
+
+// actual function algorithm / computation
+    std::string out = fs_normalize(in);
+
+    outputs[0] = factory.createScalar(out);
   }
-
-  s.erase(std::unique(s.begin(), s.end(), [](char a, char b){ return a == '/' && b == '/'; }), s.end());
-
-  return s;
-}
+};

src/normalize_fs.cpp

@@ -0,0 +1,43 @@
+#include <string>
+#include <string_view>
+#include <algorithm> // std::unique
+
+#include <filesystem>
+
+#include "ffilesystem.h"
+
+
+std::string fs_drop_slash(std::string_view in)
+{
+  // drop all trailing "/" and duplicated internal "/"
+
+  std::filesystem::path p(in);
+
+  std::string s = p.generic_string();
+
+  if (!fs_is_windows() || p != p.root_path()){
+    while(s.length() > 1 && s.back() == '/')
+      s.pop_back();
+  }
+
+  s.erase(std::unique(s.begin(), s.end(), [](char a, char b){ return a == '/' && b == '/'; }), s.end());
+
+  return s;
+}
+
+
+std::string fs_normalize(std::string_view path)
+{
+  std::filesystem::path p(path);
+
+  std::string r = p.lexically_normal().generic_string();
+
+  // no trailing slash
+  if (r.length() > 1 && r.back() == '/' && (!fs_is_windows() || p != p.root_path()))
+    r.pop_back();
+
+  if (r.empty())
+    r.push_back('.');
+
+  return r;
+}

src/octave/drop_slash.cpp

@@ -3,7 +3,7 @@
 #include <string>
 
 #include "pure.cpp"
-#include "normalize.cpp"
+#include "normalize_fs.cpp"
 
 
 DEFUN_DLD (drop_slash, args, nargout,

src/octave/normalize.cpp

@@ -0,0 +1,20 @@
+#include <octave/oct.h>
+
+#include <string>
+
+#include "pure.cpp"
+#include "normalize_fs.cpp"
+
+
+DEFUN_DLD (normalize, args, nargout,
+           "normalize path")
+{
+  if (args.length() != 1){
+    octave_stdout << "Oct: One input required\n";
+    return octave_value("");
+  }
+
+  std::string out = fs_normalize(args(0).string_value());
+
+  return octave_value(out);
+}

src/octave/parent.cpp

@@ -4,7 +4,7 @@
 #include <filesystem>
 
 #include "pure.cpp"
-#include "normalize.cpp"
+#include "normalize_fs.cpp"
 
 
 DEFUN_DLD (parent, args, nargout,