Generate a series of nested switch statements to parse CSSPrimitiveValue units.

third_party/WebKit/Source/build/scripts/make_element_lookup_trie.py

Index: third_party/WebKit/Source/build/scripts/make_element_lookup_trie.py
diff --git a/third_party/WebKit/Source/build/scripts/make_element_lookup_trie.py b/third_party/WebKit/Source/build/scripts/make_element_lookup_trie.py
index 90b696d0c22b83aa0eb1de8fc69b3b559431e477..79221a17f6c3cc529f0b0602f57588a38c48a137 100755
--- a/third_party/WebKit/Source/build/scripts/make_element_lookup_trie.py
+++ b/third_party/WebKit/Source/build/scripts/make_element_lookup_trie.py
@@ -27,6 +27,7 @@
 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
+from collections import OrderedDict
 from itertools import groupby, islice
 import sys
 
@@ -36,7 +37,7 @@ import template_expander
 PARAMETER_NAME = 'data'
 
 
-def _trie(tags, index):
+def _trie(str_to_return_value_dict, index):
     """Make a trie from list of tags, starting at index.
 
     Resulting trie is partly space-optimized (semi-radix tree): once have only
@@ -44,39 +45,47 @@ def _trie(tags, index):
     However, does not compact branch nodes with a single child. (FIXME)
 
     Returns:
-        (char, subtrie, tag, conditions): (char, trie, str, list)
-            code generation differs between branch nodes and leaf nodes,
-            hence need different data for each.
+        dict(char, subtrie, string, value, conditions):

[Timothy Loh, 2016/05/03 07:40:25]

Can we return something structured like (e.g.)

{
 'a' : {
  'b' : ("cd", "abcd.impl()"..), 
  'd' : {
   'e' : {
    'f' : ("", "adef.impl()"),
    'g' : ("", "adeg.impl()"),
   },
  },
 },
}

Should be a fair amount simpler. I think you can differentiate between internal
and leaf nodes in jinja using mapping()

http://jinja.pocoo.org/docs/dev/templates/#mapping

[meade_UTC10, 2016/05/04 07:24:54]

Done.

+            char, list, string, string, string. Code generation differs between
+            branch nodes and leaf nodes, hence need different data for each.
 
     Arguments:
-        tags: sorted list
-            (sorted needed by groupby, list needed by len)
+        str_to_return_value_dict: A ordered dictionary of strings to the return
+            value strings. Ordered by length, then alphabetically.
         index: index at which to branch
-            (assumes prior to this index strings have a common prefix)
     """
     def trie_node(char, subtags_iter):
         # Pass in |char| so we can include in same tuple without unpacking
         subtags = list(subtags_iter)  # need list for len
         if len(subtags) == 1:  # terminal node, no subtrie
-            subtrie = None
             tag = subtags[0]
-            conditions = _conditions(tag, index + 1)
+            return {
+                'char': char,
+                'subtrie': None,
+                'string': tag,
+                'value': subtags_iter[tag],
+                'conditions': _conditions(tag, index + 1)
+            }
         else:
-            subtrie = _trie(subtags, index + 1)
-            tag = None
-            conditions = None
-        return char, subtrie, tag, conditions
+            return {
+                'char': char,
+                'subtrie': _trie(subtags_iter, index + 1),
+                'string': None,
+                'value': None,
+                'conditions': None
+            }
 
     # Group by char at index
-    def char_at_index(tag):
-        return tag[index].lower()
+    def char_at_index(string):
+        return string[index].lower()
 
-    char_subtags = ((k, g) for k, g in groupby(tags, char_at_index))
+    char_subtags = []
+    for k, g in groupby(str_to_return_value_dict, char_at_index):

[Timothy Loh, 2016/05/03 07:40:25]

same comment as other place about groupby/OrderedDict

[meade_UTC10, 2016/05/04 07:24:54]

Done.

+        sub_dict = OrderedDict(
+            [(inner_key, str_to_return_value_dict[inner_key]) for inner_key in g])
+        char_subtags.append((k, sub_dict))
 
-    # FIXME: if all subtags have a common prefix, merge with child
-    # and skip the switch in the generated code
-
-    return (trie_node(char, subtags) for char, subtags in char_subtags)
+    return [trie_node(char, subtags) for char, subtags in char_subtags]
 
 
 def _conditions(tag, index):

[Timothy Loh, 2016/05/03 07:40:25]

as discussed, let's move all of the logic for this into the template

[meade_UTC10, 2016/05/04 07:24:54]

Done.

@@ -86,6 +95,19 @@ def _conditions(tag, index):
             for i, c in islice(enumerate(tag), index, None)]
 
 
+def _trie_list(str_to_return_value_dict):
+    ordered_dict = OrderedDict(

[Timothy Loh, 2016/05/03 07:40:25]

OrderedDict+groupby seems weird here, probably better to do

result = defaultdict(list)
for key,value in str_.._dict.items():
    result[len(key)].append((key, value))

which makes a dict structured like

{2: [("em",value1), ("px",value2)]}

[meade_UTC10, 2016/05/04 07:24:54]

Done.

+        sorted(str_to_return_value_dict.items(), key=lambda pair: (len(pair[0]), pair[0])))
+
+    dicts = []
+    for k, g in groupby(ordered_dict, len):
+        sub_dict = OrderedDict(
+            [(inner_key, str_to_return_value_dict[inner_key]) for inner_key in g])
+        dicts.append((k, sub_dict))
+
+    return [(length, _trie(d, 0)) for length, d in dicts]
+
+
 class ElementLookupTrieWriter(in_generator.Writer):
     # FIXME: Inherit all these from somewhere.
     defaults = {
@@ -108,7 +130,9 @@ class ElementLookupTrieWriter(in_generator.Writer):
 
     def __init__(self, in_file_paths):
         super(ElementLookupTrieWriter, self).__init__(in_file_paths)
-        self._tags = [entry['name'] for entry in self.in_file.name_dictionaries]
+        self._tags = {}
+        for entry in self.in_file.name_dictionaries:
+            self._tags[entry['name']] = entry['name']
         self._namespace = self.in_file.parameters['namespace'].strip('"')
         self._outputs = {
             (self._namespace + 'ElementLookupTrie.h'): self.generate_header,
@@ -123,15 +147,9 @@ class ElementLookupTrieWriter(in_generator.Writer):
 
     @template_expander.use_jinja('ElementLookupTrie.cpp.tmpl')
     def generate_implementation(self):
-        # First sort, so groupby works
-        self._tags.sort(key=lambda tag: (len(tag), tag))
-        # Group tags by length
-        length_tags = ((k, g) for k, g in groupby(self._tags, len))
-
         return {
             'namespace': self._namespace,
-            'length_tries': ((length, _trie(tags, 0))
-                             for length, tags in length_tags),
+            'length_tries': _trie_list(self._tags)
         }