Do not emit some unnecessary whitespaces in unparser.

lib/compiler/implementation/tree/unparser.dart

Index: lib/compiler/implementation/tree/unparser.dart
diff --git a/lib/compiler/implementation/tree/unparser.dart b/lib/compiler/implementation/tree/unparser.dart
index 3e35c289b6f917e498b31e83cd6ae32a150e8dde..e0b3306abc8c3d3ca4a2686f5d9d76e7f3879319 100644
--- a/lib/compiler/implementation/tree/unparser.dart
+++ b/lib/compiler/implementation/tree/unparser.dart
@@ -168,7 +168,8 @@ class Unparser implements Visitor {
     visit(node.condition);
     visit(node.thenPart);
     if (node.hasElsePart) {
-      addToken(node.elseToken);
+      add(node.elseToken.value);
+      if (node.elsePart is !Block) sb.add(' ');

[ahe, 2012/08/30 10:42:45]

This could probably be generalized if we didn't use a string buffer, but used a
custom class to collect "tokens". When these tokens are concatenated to produce
the final string, you'd be able to easily see if a space is needed or not.

       visit(node.elsePart);
     }
   }
@@ -209,7 +210,7 @@ class Unparser implements Visitor {
   }
 
   visitLiteralList(LiteralList node) {
-    addToken(node.constKeyword);
+    if (node.constKeyword !== null) add(node.constKeyword.value);
     visit(node.typeArguments);
     visit(node.elements);
   }
@@ -221,7 +222,7 @@ class Unparser implements Visitor {
    */
   unparseNodeListFrom(NodeList node, Link<Node> from) {
     if (from.isEmpty()) return;
-    String delimiter = (node.delimiter === null) ? " " : "${node.delimiter}";
+    String delimiter = (node.delimiter === null) ? "" : "${node.delimiter}";
     visit(from.head);
     for (Link link = from.tail; !link.isEmpty(); link = link.tail) {
       sb.add(delimiter);
@@ -230,7 +231,7 @@ class Unparser implements Visitor {
   }
 
   visitNodeList(NodeList node) {
-    if (node.beginToken !== null) addToken(node.beginToken);
+    addToken(node.beginToken);
     if (node.nodes !== null) {
       unparseNodeListFrom(node, node.nodes);
     }
@@ -243,10 +244,10 @@ class Unparser implements Visitor {
 
   visitReturn(Return node) {
     add(node.beginToken.value);
-    if (node.hasExpression) {
+    if (node.hasExpression && node.getBeginToken().slowToString() != '=>') {

[ahe, 2012/08/30 10:42:45]

Please do not use slowToString. That's why it has "slow" in its name.

It should be:

node.getBeginToken().stringValue !== '=>'

Furthermore, getBeginToken() is not the right abstraction here. It is for
computing the range of an AST node. This is why the previous line uses
node.beginToken directly. A related remark: beginToken is a bad name. It should
probably be called "returnKeyword", like LiteralList's "constKeyword".

Anyways, I think it would be much nicer if the Return class had a bool getter
named something like "isCompactReturn".

[Roman, 2012/08/30 11:38:48]

In the last patch I changed this to:
node.beginToken.kind != FUNCTION_TOKEN

[ahe, 2012/08/30 11:51:07]

Please don't use the _TOKEN variables. You really should use stringValue and
string literals.

[Roman, 2012/08/30 11:54:56]

Changed to:
node.beginToken.stringValue != '=>'

       sb.add(' ');
-      visit(node.expression);
     }
+    visit(node.expression);

[ahe, 2012/08/30 10:42:45]

Is this change intentional?

[Roman, 2012/08/30 11:38:48]

Yes. visit() checks for null anyway and I didn't want to place another if inside
of existing if for adding whitespace.

[ahe, 2012/08/30 11:51:07]

Of course. I wasn't thinking.

     if (node.endToken !== null) add(node.endToken.value);
   }
 
@@ -339,10 +340,10 @@ class Unparser implements Visitor {
   }
 
   visitDoWhile(DoWhile node) {
-    addToken(node.doKeyword);
+    add(node.doKeyword.value);
+    if (node.body is !Block) sb.add(' ');
     visit(node.body);
-    sb.add(' ');
-    addToken(node.whileKeyword);
+    add(node.whileKeyword.value);
     visit(node.condition);
     sb.add(node.endToken.value);
   }
@@ -350,7 +351,6 @@ class Unparser implements Visitor {
   visitWhile(While node) {
     addToken(node.whileKeyword);
     visit(node.condition);
-    sb.add(' ');
     visit(node.body);
   }
 
@@ -395,12 +395,12 @@ class Unparser implements Visitor {
 
   visitForIn(ForIn node) {
     add(node.forToken.value);
-    sb.add(' (');
+    sb.add('(');
     visit(node.declaredIdentifier);
     sb.add(' ');
     addToken(node.inToken);
     visit(node.expression);
-    sb.add(') ');
+    sb.add(')');
     visit(node.body);
   }
 
@@ -415,9 +415,7 @@ class Unparser implements Visitor {
   }
 
   visitLiteralMap(LiteralMap node) {
-    if (node.constKeyword !== null) {
-      add(node.constKeyword.value);
-    }
+    if (node.constKeyword !== null) add(node.constKeyword.value);
     if (node.typeArguments !== null) visit(node.typeArguments);
     visit(node.entries);
   }
@@ -425,21 +423,18 @@ class Unparser implements Visitor {
   visitLiteralMapEntry(LiteralMapEntry node) {
     visit(node.key);
     add(node.colonToken.value);
-    sb.add(' ');
     visit(node.value);
   }
 
   visitNamedArgument(NamedArgument node) {
     visit(node.name);
     add(node.colonToken.value);
-    sb.add(' ');
     visit(node.expression);
   }
 
   visitSwitchStatement(SwitchStatement node) {
     addToken(node.switchKeyword);
     visit(node.parenthesizedExpression);
-    sb.add(' ');
     visit(node.cases);
   }
 
@@ -458,7 +453,7 @@ class Unparser implements Visitor {
     visit(node.argument);
     if (node.prefixIdentifier !== null) {
       visit(node.prefixIdentifier);
-      sb.add(': ');
+      sb.add(':');
       visit(node.prefix);
     }
     sb.add(')');
@@ -466,11 +461,10 @@ class Unparser implements Visitor {
   }
 
   visitTryStatement(TryStatement node) {
-    addToken(node.tryKeyword);
+    add(node.tryKeyword.value);
     visit(node.tryBlock);
     visit(node.catchBlocks);
     if (node.finallyKeyword !== null) {
-      sb.add(' ');
       addToken(node.finallyKeyword);
       visit(node.finallyBlock);
     }
@@ -485,11 +479,12 @@ class Unparser implements Visitor {
 
   visitCatchBlock(CatchBlock node) {
     addToken(node.onKeyword);
-    visit(node.type);
-    sb.add(' ');
-    addToken(node.catchKeyword);
+    if (node.type !== null) {
+      visit(node.type);
+      sb.add(' ');
+    }
+    add(node.catchKeyword.value);
     visit(node.formals);
-    sb.add(' ');
     visit(node.block);
   }