handle comments in Java text blocks, #805

Unix/cloc

@@ -1002,6 +1002,7 @@ $opt_git = 1 if defined($opt_git_diff_all) or
                 defined($opt_git_diff_rel) or
                 (defined($opt_vcs) and ($opt_vcs eq "git"));
 $opt_by_file  = 1 if defined  $opt_by_file_by_lang;
+$opt_fmt = 0 unless defined $opt_fmt;
 if ($opt_fmt) {
     $opt_by_file = 1;
     $opt_json = 1;
@@ -7898,6 +7899,38 @@ sub docstring_to_C {                         # {{{1
     print "<- docstring_to_C\n" if $opt_v > 2;
     return @{$ra_lines};
 } # 1}}}
+sub docstring_rm_comments {                  # {{{1
+    my ($ra_lines, ) = @_;
+    # Remove embedded C/C++ style comments in docstrings.
+
+    print "-> docstring_rm_comments()\n" if $opt_v > 2;
+
+    my $in_docstring = 0;
+    foreach (@{$ra_lines}) {
+        if (/((""")|('''))(.*?)\1/) {
+            # single line docstring
+            my ($i_start, $i_end) = ($-[0]+3, $+[0]-3);
+            # replace /*, */, // with xx
+            substr($_, $i_start, $i_end-$i_start) =~ s{(/\*|\*/|//)}{xx}g;
+            next;
+        } elsif (/^(.*?)((""")|('''))/ and  $in_docstring) {
+            $in_docstring = 0;
+            my $i_end = length $1;
+            if ($i_end) {
+                substr($_, 0, $i_end) =~ s{(/\*|\*/|//)}{xx}g;
+            }
+        } elsif (/((""")|('''))(.*?)$/ and !$in_docstring) {
+            $in_docstring = 1;
+            my $i_start = $-[0]+3;
+            substr($_, $i_start) =~ s{(/\*|\*/|//)}{xx}g;
+        } elsif ($in_docstring) {
+            s{(/\*|\*/|//)}{xx}g;
+        }
+    }
+
+    print "<- docstring_rm_comments\n" if $opt_v > 2;
+    return @{$ra_lines};
+} # 1}}}
 sub jupyter_nb {                             # {{{1
     my ($ra_lines, ) = @_;
     # Translate .ipynb file content into an equivalent set of code
@@ -9814,6 +9847,7 @@ sub set_constants {                          # {{{1
                                 [ 'call_regexp_common'  , 'C++'    ],
                             ],
     'Java'               => [
+								[ 'docstring_rm_comments', ],
                                 [ 'replace_regex', '\\\\$', ' '],
                                 # Java seems to have more path globs in strings
                                 # than other languages.  The variations makes

Unix/t/01_opts.t

@@ -767,6 +767,13 @@ my @Tests = (
                     'ref'  => '../tests/outputs/issues/804/infoSQL.java.yaml',
                 },
 
+                {
+                    'name' => 'comments in Java text blocks #805',
+                    'cd'   => '../tests/inputs/issues/805',
+                    'args' => 'text_block.java',
+                    'ref'  => '../tests/outputs/issues/805/text_block.java.yaml',
+                },
+
             );
 
 # Special cases:

cloc

@@ -1017,6 +1017,7 @@ $opt_git = 1 if defined($opt_git_diff_all) or
                 defined($opt_git_diff_rel) or
                 (defined($opt_vcs) and ($opt_vcs eq "git"));
 $opt_by_file  = 1 if defined  $opt_by_file_by_lang;
+$opt_fmt = 0 unless defined $opt_fmt;
 if ($opt_fmt) {
     $opt_by_file = 1;
     $opt_json = 1;
@@ -7913,6 +7914,38 @@ sub docstring_to_C {                         # {{{1
     print "<- docstring_to_C\n" if $opt_v > 2;
     return @{$ra_lines};
 } # 1}}}
+sub docstring_rm_comments {                  # {{{1
+    my ($ra_lines, ) = @_;
+    # Remove embedded C/C++ style comments in docstrings.
+
+    print "-> docstring_rm_comments()\n" if $opt_v > 2;
+
+    my $in_docstring = 0;
+    foreach (@{$ra_lines}) {
+        if (/((""")|('''))(.*?)\1/) {
+            # single line docstring
+            my ($i_start, $i_end) = ($-[0]+3, $+[0]-3);
+            # replace /*, */, // with xx
+            substr($_, $i_start, $i_end-$i_start) =~ s{(/\*|\*/|//)}{xx}g;
+            next;
+        } elsif (/^(.*?)((""")|('''))/ and  $in_docstring) {
+            $in_docstring = 0;
+            my $i_end = length $1;
+            if ($i_end) {
+                substr($_, 0, $i_end) =~ s{(/\*|\*/|//)}{xx}g;
+            }
+        } elsif (/((""")|('''))(.*?)$/ and !$in_docstring) {
+            $in_docstring = 1;
+            my $i_start = $-[0]+3;
+            substr($_, $i_start) =~ s{(/\*|\*/|//)}{xx}g;
+        } elsif ($in_docstring) {
+            s{(/\*|\*/|//)}{xx}g;
+        }
+    }
+
+    print "<- docstring_rm_comments\n" if $opt_v > 2;
+    return @{$ra_lines};
+} # 1}}}
 sub jupyter_nb {                             # {{{1
     my ($ra_lines, ) = @_;
     # Translate .ipynb file content into an equivalent set of code
@@ -9829,6 +9862,7 @@ sub set_constants {                          # {{{1
                                 [ 'call_regexp_common'  , 'C++'    ],
                             ],
     'Java'               => [
+								[ 'docstring_rm_comments', ],
                                 [ 'replace_regex', '\\\\$', ' '],
                                 # Java seems to have more path globs in strings
                                 # than other languages.  The variations makes

tests/inputs/issues/805/text_block.java

@@ -0,0 +1,21 @@
+// regular comment
+public String getBlockOfHtml() {
+String request = """
+            GET /*cho/foo HTT*/1.1
+            Host: local
+            Accept: */*
+            Co//ection: closed
+
+            """;
+
+    return """
+            <html>
+              /* 
+               * NOT comment
+               */
+                <body>
+                    <span>example text</span>
+                </body>
+            </html>
+            """;
+    }

tests/outputs/issues/805/text_block.java.yaml

@@ -0,0 +1,21 @@
+---
+# github.com/AlDanial/cloc
+header : 
+  cloc_url           : github.com/AlDanial/cloc
+  cloc_version       : 1.99
+  elapsed_seconds    : 0.00261998176574707
+  n_files            : 1
+  n_lines            : 21
+  files_per_second   : 381.682045682046
+  lines_per_second   : 8015.32295932296
+  report_file        : ../../../outputs/issues/805/text_block.java.yaml
+'Java' :
+  nFiles: 1
+  blank: 2
+  comment: 1
+  code: 18
+SUM: 
+  blank: 2
+  comment: 1
+  code: 18
+  nFiles: 1