Introduce transformers for:

lib/src/transform/script_compactor.dart

Index: lib/src/transform/script_compactor.dart
diff --git a/lib/src/transform/script_compactor.dart b/lib/src/transform/script_compactor.dart
new file mode 100644
index 0000000000000000000000000000000000000000..84d2032e61893d5dabc3aecf5b674e3a2f27da52
--- /dev/null
+++ b/lib/src/transform/script_compactor.dart
@@ -0,0 +1,122 @@
+// Copyright (c) 2013, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+/** Transfomer that combines multiple dart script tags into a single one. */
+library polymer.src.transform.script_compactor;
+
+import 'dart:async';
+
+import 'package:barback/barback.dart';
+import 'package:html5lib/parser.dart' show parseFragment;
+import 'package:path/path.dart' as path;
+
+import 'code_extractor.dart'; // import just for documentation.

[terry, 2013/08/20 17:03:24]

If this import isn't used the editor will flag a warning.

[Siggi Cherem (dart-lang), 2013/08/21 20:35:42]

the good news is that they also recognize that the import is used for the dart
doc comment, so they don't complain =), yay!

(I just doubled checked)

+import 'common.dart';
+
+/**
+ * Combines Dart script tags into a single script tag, and creates a new Dart
+ * file that calls the main function of each of the original script tags.
+ *
+ * This transformer assumes that all script tags point to external files. To
+ * support script tags with inlined code, use this transformer after running
+ * [InlineCodeExtractor] on an earlier phase.
+ *
+ * Internally, this transformer will convert each script tag into an import
+ * statement to a library, and then uses `initPolymer` (see polymer.dart)  to
+ * invoke the main method on each of these libraries and register any polymer
+ * elements annotated with `@CustomTag`.
+ */
+class ScriptCompactor extends Transformer {
+  Future<bool> isPrimary(Asset input) =>
+      new Future.value(input.id.extension == ".html");
+
+  Future apply(Transform transform) {
+    var id = transform.primaryId;
+    var logger = transform.logger;
+    return getPrimaryContent(transform).then((content) {
+      var document = parseHtml(content, id.path, logger);
+      var libraries = [];
+      bool changed = false;
+      for (var tag in document.queryAll('script')) {
+        if (tag.attributes['type'] != 'application/dart') continue;
+        tag.remove();
+        changed = true;
+        var src = tag.attributes['src'];
+        if (src == null) {
+          logger.warning('unexpected script without a src url', tag.sourceSpan);

[Jennifer Messerly, 2013/08/19 22:57:26]

Hmm, why the warning here? Are we assuming the script extractor was run first?
If so, maybe it should call the other one to ensure that phase as a pre-req.
Does barback have such a concept?

worse case, you could mention the other tranformer in the warning message.

[Siggi Cherem (dart-lang), 2013/08/21 20:35:42]

Good question. The transformers should always be run in order (later I'll set
them up in phases) but I wanted to make each transformer is robust enough to
catch if their preconditions are invalidated.

I'm not sure what the right action should be in that case. Should we issue a
warning, an error, or do something else? For now I kept the warning but I
referred to the InlineCodeExtractor as you suggested.

+          continue;
+        }
+        var libraryId = resolve(id, src, logger, tag.sourceSpan);
+
+        // TODO(sigmund): should we detect/remove duplicates?

[Jennifer Messerly, 2013/08/19 22:57:26]

i think so. don't want to generate erroneous code that imports the same library
twice. however, we probably should run the main twice.

Try running foo:
<script src="foo.dart"></script>
Try running foo again:
<script src="foo.dart"></script>

foo.dart:
main() => print('FOO!');

[Siggi Cherem (dart-lang), 2013/08/21 20:35:42]

But then, wouldn't it be ok to import it twice? (we anyways import them with a
prefix)

[Jennifer Messerly, 2013/08/21 23:35:29]

good point, importing twice would be fine assuming it doesn't trigger a Dart
language level error. if they are imported with unique prefixes that should work
great :) 

+        if (libraryId == null) continue;
+        libraries.add(libraryId);
+      }
+
+      if (!changed) {
+        transform.addOutput(new Asset.fromString(id, content));
+        return;
+      }
+
+      var bootstrapId = id.addExtension('_bootstrap.dart');
+      var filename = path.basename(bootstrapId.path);
+      document.body.nodes.add(parseFragment(
+            '<script type="application/dart" src="$filename"></script>'));
+
+      var urls = libraries.map((id) => importUrlFor(id, bootstrapId, logger))
+          .where((url) => url != null).toList();
+      var buffer = new StringBuffer()..write(_header);
+      for (int i = 0; i < urls.length; i++) {
+        buffer.writeln("import '${urls[i]}' as i$i;");
+      }
+      buffer..write(_mainPrefix)
+          ..writeAll(urls.map((url) => "      '$url',\n"))
+          ..write(_mainSuffix);
+
+      transform.addOutput(new Asset.fromString(bootstrapId, buffer.toString()));
+      transform.addOutput(new Asset.fromString(id, document.outerHtml));
+    });
+  }
+
+  /**
+   * Generate the import url for a file described by [id], referenced by a file
+   * with [sourceId].
+   */
+  String importUrlFor(AssetId id, AssetId sourceId, TransformLogger logger) {
+    // use package: urls if possible
+    if (id.path.startsWith('lib/')) {
+      return 'package:${id.package}/${id.path.substring(4)}';
+    }
+    
+    // Use relative urls only if it's possible.
+    if (id.package != sourceId.package) {
+      logger.error("don't know how to import $id from $sourceId");
+      return null;
+    }
+
+    var upPath = path.joinAll(path.split(sourceId.path).map((_) => '..'));
+    return path.normalize(path.join(sourceId.path, upPath, id.path));
+  }
+}
+
+const _header = """
+library app_bootstrap;
+
+import 'package:polymer/polymer.dart';
+import 'dart:mirrors' show currentMirrorSystem;
+
+""";
+
+const _mainPrefix = """
+
+void main() {
+  initPolymer([
+""";
+
+const _mainSuffix = """
+    ],
+    currentMirrorSystem().findLibrary(const Symbol('app_bootstrap'))

[Jennifer Messerly, 2013/08/19 22:57:26]

seems unfortunate to need to use mirrors for this. couldn't we get the HTML's
base URL?

[Siggi Cherem (dart-lang), 2013/08/21 20:35:42]

good question. I opened a bug to investigate.

[Jennifer Messerly, 2013/08/21 23:35:29]

btw, I discovered another pattern using mirrors that is slightly simpler
(doesn't depend on findLibrary):

currentMirrorSystem().isolate.rootLibrary.uri

+        .first.uri.toString());
+}
+""";