add normalize, canonical-distinct from absolute_path

Author: scivision

+stdlib/+fileio/absolute_path.m

@@ -11,20 +11,16 @@
   return
 end
 
-if ~stdlib.fileio.is_absolute_path(abspath)
-  % otherwise the default is Documents/Matlab, which is probably not wanted.
-  abspath = fullfile(pwd, abspath);
-end
-
 if ispc && startsWith(abspath, "\\")
   % UNC path is not canonicalized
   return
 end
 
-% similar benchmark time as java method
-% REQUIRES path to exist, while java method does not.
-% abspath = builtin('_canonicalizepath', abspath);
+if ~stdlib.fileio.is_absolute_path(abspath)
+  % .toAbsolutePath() default is Documents/Matlab, which is probably not wanted.
+  abspath = fullfile(pwd, abspath);
+end
 
-abspath = string(java.io.File(abspath).getCanonicalPath());
+abspath = string(java.io.File(abspath).getAbsolutePath());
 
 end % function

+stdlib/+fileio/canonical.m

@@ -0,0 +1,30 @@
+function c = canonical(p)
+
+arguments
+  p string {mustBeScalarOrEmpty}
+end
+
+% have to expand ~ first (like C++ filesystem::path::absolute)
+c = stdlib.fileio.expanduser(p);
+
+if isempty(c)
+  return
+end
+
+if ispc && startsWith(c, "\\")
+  % UNC path is not canonicalized
+  return
+end
+
+if ~stdlib.fileio.is_absolute_path(c)
+  % .toAbsolutePath() default is Documents/Matlab, which is probably not wanted.
+  c = fullfile(pwd, c);
+end
+
+% similar benchmark time as java method
+% REQUIRES path to exist, while java method does not.
+% abspath = builtin('_canonicalizepath', abspath);
+
+c = string(java.io.File(c).getCanonicalPath());
+
+end % function

+stdlib/+fileio/extract_zstd.m

@@ -5,8 +5,8 @@ function extract_zstd(archive, out_dir)
   out_dir (1,1) string {mustBeFolder}
 end
 
-archive = stdlib.fileio.absolute_path(archive);
-out_dir = stdlib.fileio.absolute_path(out_dir);
+archive = stdlib.fileio.canonical(archive);
+out_dir = stdlib.fileio.canonical(out_dir);
 
 exe = stdlib.fileio.which("cmake");
 if isempty(exe)

+stdlib/+fileio/normalize.m

@@ -0,0 +1,14 @@
+function n = normalize(p)
+% normalize(p) remove redundant elements of path p
+arguments
+  p string {mustBeScalarOrEmpty}
+end
+
+n = p;
+if isempty(p)
+  return
+end
+
+n = string(java.io.File(stdlib.fileio.expanduser(n)).toPath().normalize());
+
+end

+stdlib/+fileio/path_tail.m

@@ -9,7 +9,7 @@
   return
 end
 
-[~, name, ext] = fileparts(stdlib.fileio.absolute_path(apath));
+[~, name, ext] = fileparts(stdlib.fileio.canonical(apath));
 
 last = append(name, ext);
 

+stdlib/+fileio/samepath.m

@@ -10,10 +10,9 @@
    return
 end
 
-% absolute_path because unlike c++ filesystem, isSameFile does not
-% canonicalize first
-p1 = java.io.File(stdlib.fileio.absolute_path(path1)).toPath();
-p2 = java.io.File(stdlib.fileio.absolute_path(path2)).toPath();
+% unlike c++ filesystem, isSameFile does not canonicalize first
+p1 = java.io.File(stdlib.fileio.canonical(path1)).toPath();
+p2 = java.io.File(stdlib.fileio.canonical(path2)).toPath();
 
 try
   issame = java.nio.file.Files.isSameFile(p1, p2);

+stdlib/+test/TestFilePure.m

@@ -65,14 +65,12 @@ function test_absolute_path(tc)
 
 par1 = absolute_path("../2foo");
 tc.verifyNotEmpty(par1)
-tc.verifyFalse(contains(par1, ".."))
 
 par2 = absolute_path("../4foo");
 tc.verifyTrue(strncmp(par2, pabs2, 2))
 
 pt1 = absolute_path("bar/../2foo");
 tc.verifyNotEmpty(pt1)
-tc.verifyFalse(contains(pt1, ".."))
 
 va = absolute_path("2foo");
 vb = absolute_path("4foo");
@@ -81,6 +79,40 @@ function test_absolute_path(tc)
 
 end
 
+
+function test_canonical(tc)
+
+import stdlib.canonical
+
+tc.assumeTrue(usejava("jvm"), "Java required for absolute_path")
+
+tc.verifyEmpty(canonical(string.empty))
+tc.verifyEqual(canonical(""), string(pwd))
+
+pabs = canonical('2foo');
+pabs2 = canonical('4foo');
+tc.verifyFalse(startsWith(pabs, "2"))
+tc.verifyTrue(strncmp(pabs, pabs2, 2))
+
+par1 = canonical("../2foo");
+tc.verifyNotEmpty(par1)
+tc.verifyFalse(contains(par1, ".."))
+
+par2 = canonical("../4foo");
+tc.verifyTrue(strncmp(par2, pabs2, 2))
+
+pt1 = canonical("bar/../2foo");
+tc.verifyNotEmpty(pt1)
+tc.verifyFalse(contains(pt1, ".."))
+
+va = canonical("2foo");
+vb = canonical("4foo");
+tc.verifyFalse(startsWith(va, "2"))
+tc.verifyTrue(strncmp(va, vb, 2))
+
+end
+
+
 function test_with_suffix(tc)
 import stdlib.with_suffix
 

+stdlib/+test/TestWindowsCOM.m

@@ -14,7 +14,7 @@ function test_shortname(tc)
 tc.verifyTrue(endsWith(short, 'notepad.exe'), "Short name should end with 'notepad.exe'")
 tc.verifyTrue(contains(short, "MICROS~1"))
 
-tc.verifyEqual(stdlib.absolute_path(short), test_file)
+tc.verifyEqual(stdlib.canonical(short), test_file)
 
 end
 

+stdlib/absolute_path.m

@@ -1,22 +1,20 @@
-function abspath = absolute_path(p)
+function a = absolute_path(p)
 %% absolute_path(p)
 % path need not exist, but absolute path is returned
 %
 % NOTE: some network file systems are not resolvable by Matlab Java
 % subsystem, but are sometimes still valid--so return
 % unmodified path if this occurs.
 %
-% This also resolves Windows short paths to full long paths.
-%
 %%% Inputs
 % * p: path to make absolute
 %%% Outputs
-% * abspath: absolute path, if determined
+% * a: absolute path, if determined
 
 arguments
   p string {mustBeScalarOrEmpty}
 end
 
-abspath = stdlib.fileio.absolute_path(p);
+a = stdlib.fileio.absolute_path(p);
 
 end

+stdlib/canonical.m

@@ -1,7 +1,6 @@
-function abspath = canonical(p)
+function c = canonical(p)
 %% canonical(p)
-% yes this is the same as absolute_path
-% path need not exist, but absolute path is returned
+% path need not exist, but canonical path is returned
 %
 % NOTE: some network file systems are not resolvable by Matlab Java
 % subsystem, but are sometimes still valid--so return
@@ -10,14 +9,14 @@
 % This also resolves Windows short paths to full long paths.
 %
 %%% Inputs
-% * p: path to make absolute
+% * p: path to make canonical
 %%% Outputs
-% * abspath: absolute path, if determined
+% * c: canonical path, if determined
 
 arguments
   p string {mustBeScalarOrEmpty}
 end
 
-abspath = stdlib.fileio.absolute_path(p);
+c = stdlib.fileio.canonical(p);
 
 end