Discard first line of multiline string if only whitespace

runtime/vm/scanner.cc

Index: runtime/vm/scanner.cc
===================================================================
--- runtime/vm/scanner.cc	(revision 27888)
+++ runtime/vm/scanner.cc	(working copy)
@@ -422,19 +422,14 @@
 
   // Entering string scanning mode.
   BeginStringLiteral(c0_);
-  string_is_multiline_ = (LookaheadChar(1) == c0_) &&
-      (LookaheadChar(2) == c0_);
+  ReadChar();
 
-  ReadChar();  // Skip opening delimiter.
-  if (string_is_multiline_) {
+  if ((c0_ == string_delimiter_) && (LookaheadChar(1) == string_delimiter_)) {
+    string_is_multiline_ = true;
     ReadChar();  // Skip two additional string delimiters.
     ReadChar();
-    if (c0_ == '\n') {
-      // Skip first character of multiline string if it is a newline.
-      ReadChar();
-    }
   }
-  ScanLiteralStringChars(is_raw);
+  ScanLiteralStringChars(is_raw, true);

[Ivan Posva, 2013/09/25 16:31:50]

Can't you pass string_is_multiline_ here? As this true only applies if we are
scanning a multiline string.

Correct would be "string_is_multiline_ ? true : false", but that seems a bit too
over the top.

[hausner, 2013/09/25 17:06:49]

Simplified as you suggested.

 }
 
 
@@ -497,7 +492,7 @@
 }
 
 
-void Scanner::ScanLiteralStringChars(bool is_raw) {
+void Scanner::ScanLiteralStringChars(bool is_raw, bool is_first_line) {
   GrowableArray<int32_t> string_chars(64);
 
   ASSERT(IsScanningString());
@@ -595,6 +590,27 @@
       }
       string_chars.Add(ch1);
     }
+    // The first line of a multi-line string is discarded if it only
+    // contains whitespace.
+    if (string_is_multiline_ &&
+        is_first_line &&
+        (string_chars.Last() == '\n')) {
+      bool whitespace_only = true;
+      // Last character is the newline, don't inspect it.
+      const intptr_t len = string_chars.length() - 1;
+      for (int i = 0; i < len; i++) {
+        int32_t ch = string_chars[i];
+        if ((ch != ' ') && (ch != '\t')) {

[Ivan Posva, 2013/09/25 16:31:50]

Please clarify the definition of white-space definition with Gilad.

[hausner, 2013/09/25 17:06:49]

Will do. I used the WHITESPACE production in the spec, although this production
is not used to formally define what whitespace means in a string literal.

+          // Non-whitespace character, keep the first line.
+          whitespace_only = false;
+          break;
+        }
+      }
+      if (whitespace_only) {
+        string_chars.Clear();  // Discard characters on first line.
+      }
+      is_first_line = false;
+    }
     ReadChar();
   }
 }
@@ -627,7 +643,7 @@
           break;
         }
       } else {
-        ScanLiteralStringChars(false);
+        ScanLiteralStringChars(false, false);
       }
       break;
     }