file_checksum: use Java to compute file checksum rather than system() call

this can be slightly slower, but seems to be a better practice.

Author: scivision

+stdlib/+fileio/file_checksum.m

@@ -0,0 +1,33 @@
+function hash = file_checksum(file, method)
+%% file_checkum compute checksum of file
+% read in chunks to avoid excessive RAM use
+%
+% method: md5, sha1, sha256, etc.
+%
+% Reference: https://docs.oracle.com/javase/8/docs/api/java/security/MessageDigest.html
+arguments
+    file (1,1) string {mustBeFile}
+    method (1,1) string {mustBeNonzeroLengthText}
+end
+
+if any(method == ["sha256", "SHA256"])
+  method = "SHA-256";
+end
+
+file_chunk = 10e6;  % arbitrary (bytes) didn't seem to be very sensitive for speed
+
+inst = java.security.MessageDigest.getInstance(method);
+
+fid = fopen(file, 'r');
+assert(fid > 0, "could not open " + file)
+
+while ~feof(fid)
+  inst.update(fread(fid, file_chunk, '*uint8'))
+end
+fclose(fid);
+
+hash = typecast(inst.digest, 'uint8');
+
+hash = string(sprintf('%.2x', hash));
+
+end

+stdlib/+test/TestFileImpure.m

@@ -67,18 +67,20 @@ function test_is_exe_which_fullpath(tc)
 
 function test_hash(tc)
 import matlab.unittest.constraints.IsFile
+import matlab.unittest.fixtures.TemporaryFolderFixture
 
-fn = tempname;
+fixture = tc.applyFixture(TemporaryFolderFixture);
+
+fn = fullfile(fixture.Folder, "hello");
 fid = fopen(fn, "w");
 tc.assumeGreaterThan(fid, 0);
 fprintf(fid, "hello");
 fclose(fid);
 tc.assumeThat(fn, IsFile)
 
-tc.verifyEqual(stdlib.fileio.md5sum(fn), "5d41402abc4b2a76b9719d911017c592")
-tc.verifyEqual(stdlib.fileio.sha256sum(fn), "2cf24dba5fb0a30e26e83b2ac5b9e29e1b161e5c1fa7425e73043362938b9824")
+tc.verifyEqual(stdlib.fileio.file_checksum(fn, "md5"), "5d41402abc4b2a76b9719d911017c592")
+tc.verifyEqual(stdlib.fileio.file_checksum(fn, "sha256"), "2cf24dba5fb0a30e26e83b2ac5b9e29e1b161e5c1fa7425e73043362938b9824")
 
-delete(fn)
 end
 
 end

+stdlib/md5sum.m

@@ -5,6 +5,6 @@
   file (1,1) string {mustBeFile}
 end
 
-hash = stdlib.fileio.md5sum(file);
+hash = stdlib.fileio.file_checksum(file, "md5");
 
 end

+stdlib/sha256sum.m

@@ -5,6 +5,6 @@
   file (1,1) string {mustBeFile}
 end
 
-hash = stdlib.fileio.sha256sum(file);
+hash = stdlib.fileio.file_checksum(file, "SHA-256");
 
 end

+stdlib/+fileio/md5sum.m renamed to archive/md5sum.m

@@ -1,5 +1,6 @@
 function hash = md5sum(file)
 %% md5sum(file)
+% DEPRECATED: use file_checksum(file, "md5") instead
 % compute MD5 hash of file
 arguments
   file (1,1) string {mustBeFile}
@@ -12,7 +13,7 @@
 elseif ispc
   [stat, hash] = system("CertUtil -hashfile " + file + " MD5");
 else
-  error("no sha256sum method for your OS")
+  error("no method for your OS")
 end
 
 assert(stat == 0, hash, "failed to compute md5sum of " + file)

+stdlib/+fileio/sha256sum.m renamed to archive/sha256sum.m

@@ -1,5 +1,7 @@
 function hash = sha256sum(file)
-
+%% sha256sum
+% DEPRECATED: use file_checksum(file, "SHA-256") instead
+% compute sha256 checksum of filetemp
 arguments
   file (1,1) string {mustBeFile}
 end
@@ -11,7 +13,7 @@
 elseif ispc
   [stat, hash] = system("CertUtil -hashfile " + file + " SHA256");
 else
-  error("no sha256sum method for your OS")
+  error("no method for your OS")
 end
 
 assert(stat == 0, hash, "failed to compute SHA256 hash of " + file)