From fc94753481337b42e14dc4ab388d2107d6b75935 Mon Sep 17 00:00:00 2001
From: Al <albarrentine@gmail.com>
Date: Wed, 18 May 2016 00:43:01 -0400
Subject: [PATCH] [po boxes] random PO box generation

---
 scripts/geodata/addresses/floors.py    |  3 +-
 scripts/geodata/addresses/numbering.py |  3 +-
 scripts/geodata/addresses/po_boxes.py  | 39 +++++++++++++++++++++++++-
 scripts/geodata/addresses/units.py     |  3 +-
 4 files changed, 42 insertions(+), 6 deletions(-)

diff --git a/scripts/geodata/addresses/floors.py b/scripts/geodata/addresses/floors.py
index 7dc2b87d..fd4ff87d 100644
--- a/scripts/geodata/addresses/floors.py
+++ b/scripts/geodata/addresses/floors.py
@@ -2,14 +2,13 @@ import random
 import six
 
 from geodata.addresses.config import address_config
+
 from geodata.addresses.numbering import NumberedComponent, sample_alphabet, latin_alphabet
 from geodata.encoding import safe_decode
 from geodata.math.sampling import weighted_choice, zipfian_distribution, cdf
 
 
 class Floor(NumberedComponent):
-    config_key = 'levels'
-
     # When we don't know the number of floors, use a Zipfian distribution
     # to choose randomly between 1 and max_floors with 1 being much more
     # likely than 2, etc.
diff --git a/scripts/geodata/addresses/numbering.py b/scripts/geodata/addresses/numbering.py
index b84500dc..3c43f946 100644
--- a/scripts/geodata/addresses/numbering.py
+++ b/scripts/geodata/addresses/numbering.py
@@ -107,7 +107,6 @@ class NumberedComponent(object):
     ALPHA = 'alpha'
     ALPHA_PLUS_NUMERIC = 'alpha_plus_numeric'
     NUMERIC_PLUS_ALPHA = 'numeric_plus_alpha'
-    DIRECTIONAL = 'directional'
 
     @classmethod
     def choose_alphanumeric_type(cls, key, language, country=None):
@@ -116,7 +115,7 @@ class NumberedComponent(object):
         values = []
         probs = []
 
-        for num_type in (cls.NUMERIC, cls.ALPHA, cls.ALPHA_PLUS_NUMERIC, cls.NUMERIC_PLUS_ALPHA, cls.DIRECTIONAL):
+        for num_type in (cls.NUMERIC, cls.ALPHA, cls.ALPHA_PLUS_NUMERIC, cls.NUMERIC_PLUS_ALPHA):
             key = '{}_probability'.format(num_type)
             prob = alphanumeric_props.get(key)
             if prob is not None:
diff --git a/scripts/geodata/addresses/po_boxes.py b/scripts/geodata/addresses/po_boxes.py
index f9482b1b..cd2f4703 100644
--- a/scripts/geodata/addresses/po_boxes.py
+++ b/scripts/geodata/addresses/po_boxes.py
@@ -1,8 +1,10 @@
 import random
 import six
 
+from geodata.addresses.config import address_config
 from geodata.addresses.numbering import NumberedComponent, sample_alphabet, latin_alphabet
 from geodata.encoding import safe_decode
+from geodata.math.sampling import cdf, weighted_choice
 
 
 class POBox(NumberedComponent):
@@ -24,9 +26,44 @@ class POBox(NumberedComponent):
         return six.u('').join([safe_decode(cls.random_digits(num_digits)), suffix])
 
     @classmethod
-    def random_letter(cls, alphabet=latin_alphabet):
+    def random_letter(cls, language, country=None):
+        alphabet = address_config.get_property('alphabet', language, country=country, default=latin_alphabet)
         return sample_alphabet(alphabet)
 
+    @classmethod
+    def random(cls, language, country=None):
+        num_type, num_type_props = cls.choose_alphanumeric_type('po_boxes.alphanumeric', language, country=country)
+
+        if num_type != cls.ALPHA:
+            digit_config = address_config.get_property('po_boxes.digits', language, country=country, default=[])
+            values = []
+            probs = []
+
+            for val in digit_config:
+                values.append(val['length'])
+                probs.append(val['probability'])
+
+            probs = cdf(probs)
+
+            num_digits = weighted_choice(values, probs)
+
+            number = cls.random_digits(num_digits)
+
+            if num_type == cls.NUMERIC:
+                return safe_decode(number)
+            else:
+                letter = cls.random_letter(language, country=country)
+
+                whitespace_probability = float(num_type_props.get('whitespace_probability', 0.0))
+                whitespace_phrase = six.u(' ') if whitespace_probability and random.random() < whitespace_probability else six.u('')
+
+                if num_type == cls.ALPHA_PLUS_NUMERIC:
+                    return six.u('{}{}{}').format(letter, whitespace_phrase, number)
+                elif num_type == cls.NUMERIC_PLUS_ALPHA:
+                    return six.u('{}{}{}').format(number, whitespace_phrase, letter)
+        else:
+            return cls.random_letter(language, country=country)
+
     @classmethod
     def phrase(cls, box_number, language, country=None):
         return cls.numeric_phrase('po_boxes.alphanumeric', safe_decode(box_number), language,
diff --git a/scripts/geodata/addresses/units.py b/scripts/geodata/addresses/units.py
index 5f3230b1..b3f55612 100644
--- a/scripts/geodata/addresses/units.py
+++ b/scripts/geodata/addresses/units.py
@@ -5,6 +5,7 @@ from geodata.addresses.config import address_config
 from geodata.addresses.directions import RelativeDirection
 from geodata.addresses.floors import Floor
 from geodata.addresses.numbering import NumberedComponent, sample_alphabet, latin_alphabet
+from geodata.configs.utils import nested_get
 from geodata.encoding import safe_decode
 from geodata.math.sampling import weighted_choice, zipfian_distribution, cdf
 
@@ -67,7 +68,7 @@ class Unit(NumberedComponent):
             else:
                 number = weighted_choice(cls.positive_units_letters, cls.positive_units_letters_cdf)
 
-                whitespace_probability = num_type_props.get('{}_whitespace_probability'.format(num_type))
+                whitespace_probability = nested_get(num_type_props, (num_type, 'whitespace_probability'))
                 whitespace_phrase = six.u(' ') if whitespace_probability and random.random() < whitespace_probability else six.u('')
 
                 if num_type == cls.ALPHA_PLUS_NUMERIC: