From 22be892635a030a997d1f7ea79127ae612617eea Mon Sep 17 00:00:00 2001
From: Al <albarrentine@gmail.com>
Date: Wed, 1 Jun 2016 12:35:48 -0400
Subject: [PATCH] [dictionaries] Updates to German dictionaries

---
 .../dictionaries/de/ambiguous_expansions.txt  |  2 +
 resources/dictionaries/de/entrances.txt       |  1 +
 .../dictionaries/de/level_types_numbered.txt  |  4 +-
 resources/dictionaries/de/staircases.txt      |  3 +-
 resources/dictionaries/de/unit_directions.txt |  2 +
 .../dictionaries/de/unit_types_numbered.txt   |  4 +-
 scripts/geodata/addresses/components.py       | 58 ++++++++++---------
 7 files changed, 43 insertions(+), 31 deletions(-)
 create mode 100644 resources/dictionaries/de/entrances.txt
 create mode 100644 resources/dictionaries/de/unit_directions.txt

diff --git a/resources/dictionaries/de/ambiguous_expansions.txt b/resources/dictionaries/de/ambiguous_expansions.txt
index 45c3d48b..838ea42f 100644
--- a/resources/dictionaries/de/ambiguous_expansions.txt
+++ b/resources/dictionaries/de/ambiguous_expansions.txt
@@ -5,8 +5,10 @@ g
 h
 i
 k
+l
 n
 o
+r
 s
 u
 v
diff --git a/resources/dictionaries/de/entrances.txt b/resources/dictionaries/de/entrances.txt
new file mode 100644
index 00000000..06f2e515
--- /dev/null
+++ b/resources/dictionaries/de/entrances.txt
@@ -0,0 +1 @@
+eingang
\ No newline at end of file
diff --git a/resources/dictionaries/de/level_types_numbered.txt b/resources/dictionaries/de/level_types_numbered.txt
index 58f99d00..535d0b4f 100644
--- a/resources/dictionaries/de/level_types_numbered.txt
+++ b/resources/dictionaries/de/level_types_numbered.txt
@@ -1 +1,3 @@
-obergeschoss|og|o g
\ No newline at end of file
+etage
+obergeschoss|og|o g
+stock
\ No newline at end of file
diff --git a/resources/dictionaries/de/staircases.txt b/resources/dictionaries/de/staircases.txt
index b3ba5503..29f69507 100644
--- a/resources/dictionaries/de/staircases.txt
+++ b/resources/dictionaries/de/staircases.txt
@@ -1 +1,2 @@
-stiege|stg
\ No newline at end of file
+stiege|stg
+treppe
\ No newline at end of file
diff --git a/resources/dictionaries/de/unit_directions.txt b/resources/dictionaries/de/unit_directions.txt
new file mode 100644
index 00000000..f5a2dd16
--- /dev/null
+++ b/resources/dictionaries/de/unit_directions.txt
@@ -0,0 +1,2 @@
+links|l
+rechts|r
\ No newline at end of file
diff --git a/resources/dictionaries/de/unit_types_numbered.txt b/resources/dictionaries/de/unit_types_numbered.txt
index f7b3fdbf..986c8dcf 100644
--- a/resources/dictionaries/de/unit_types_numbered.txt
+++ b/resources/dictionaries/de/unit_types_numbered.txt
@@ -1,5 +1,5 @@
-abteilung|abt
+appartement|apt|app
 büro|buro|buero
 top
-wohnung|whg|w
+wohnung|whg|w|/ w|/ / w
 zimmer|zi
\ No newline at end of file
diff --git a/scripts/geodata/addresses/components.py b/scripts/geodata/addresses/components.py
index 014fa0c4..da0c4c20 100644
--- a/scripts/geodata/addresses/components.py
+++ b/scripts/geodata/addresses/components.py
@@ -399,40 +399,44 @@ class AddressComponents(object):
 
     def combine_fields(self, address_components, language, country=None, generated_components=None):
         combo_config = address_config.get_property('components.combinations', language, country=country, default={})
-        values = []
-        probs = []
+        combos = []
+        probs = {}
         generated_components = generated_components or set()
 
-        for k, v in six.iteritems(combo_config):
-            values.append(v)
-            probs.append(v['probability'])
-
-        if not isclose(sum(probs), 1.0):
-            values.append(None)
-            probs.append(1.0 - sum(probs))
-
-        probs = cdf(probs)
-
-        combo = weighted_choice(values, probs)
-        if combo is not None:
+        for k, combo in six.iteritems(combo_config):
             components = OrderedDict.fromkeys(combo['components']).keys()
             if not all((c in address_components and (c in generated_components or self.is_numeric(address_components[c])) for c in components)):
-                return None
+                continue
 
-            values = []
-            probs = []
-            for s in combo['separators']:
-                values.append(s['separator'])
-                probs.append(s['probability'])
+            combos.append((len(components), combo))
 
-            probs = cdf(probs)
-            separator = weighted_choice(values, probs)
+        if not combos:
+            return None
 
-            new_label = combo['label']
-            new_value = separator.join([address_components.pop(c) for c in components])
-            address_components[new_label] = new_value
-            return new_label
-        return None
+        combos.sort(key=operator.itemgetter(0), reverse=True)
+
+        for num_components, combo in combos:
+            prob = combo['probability']
+            if random.random() < prob:
+                break
+        else:
+            return None
+
+        components = OrderedDict.fromkeys(combo['components']).keys()
+
+        values = []
+        probs = []
+        for s in combo['separators']:
+            values.append(s['separator'])
+            probs.append(s['probability'])
+
+        probs = cdf(probs)
+        separator = weighted_choice(values, probs)
+
+        new_label = combo['label']
+        new_value = separator.join([address_components.pop(c) for c in components])
+        address_components[new_label] = new_value
+        return new_label
 
     def generated_type(self, component, existing_components, language, country=None):
         component_config = address_config.get_property('components.{}'.format(component), language, country=country)