o'clock fix (#13)

* Implement extra function to fix and enhance o'clock behaviour, 9 tests failing

* Bug fix to solve 9 previous test failures.

* Implement function to convert word numbers to digit numbers + add tests

* Modify default value + add tests to improve coverage

timefhuman/categorize.py

@@ -27,6 +27,7 @@ def categorize(tokens, now):
     [7/17/2018, 3 pm, '-', 7/19/2018, 2 pm]
     """
     tokens = list(tokens)
+    tokens = convert_words_to_numbers(tokens)
     tokens = convert_day_of_week(tokens, now)
     tokens = convert_relative_days(tokens, now)
     tokens = convert_time_of_day(tokens)
@@ -37,6 +38,24 @@ def categorize(tokens, now):
     return tokens
 
 
+# TODO: add conversions for thirty, fourty-five
+# TODO: maybe set default for seven o'clock to 7 pm not am?
+def convert_words_to_numbers(tokens):
+    """
+    Converts numbers in word format into number format
+    >>> convert_words_to_numbers(['five', "o'clock"])
+    ['5', "o'clock"]
+    >>> convert_words_to_numbers(['seven', "o'clock"])
+    ['7', "o'clock"]
+    """
+    number_words = ["zero", "one", "two", "three", "four", "five", "six",
+                    "seven", "eight", "nine", "ten", "eleven", "twelve"]
+    for index, token in enumerate(tokens):
+        if token.lower() in number_words:
+            tokens[index] = str(number_words.index(token.lower()))
+    return tokens
+
+
 # TODO: "monday next week"
 def convert_day_of_week(tokens, now=datetime.datetime.now()):
     """Convert day-of-week vernacular into date-like string.
@@ -90,7 +109,6 @@ def convert_relative_days(tokens, now=datetime.datetime.now()):
     return tokens
 
 
-
 # TODO: convert to new token-based system
 def extract_weeks_offset(tokens, end=None, key_tokens=(
         'next', 'previous', 'last', 'upcoming', 'past', 'prev')):
@@ -316,6 +334,35 @@ def extract_hour_minute(string, time_of_day=None):
     return TimeToken(relative_hour=hour, minute=minute, time_of_day=time_of_day)
 
 
+def extract_hour_minute_token(tokens, time_of_day=None):
+    """
+        Attempt to extract the exact token which contains the hour and minute and convert it into a number.
+        This will either be 1 before or 2 before the am/pm token.
+        12:00 is the default token to prevent failure
+        Tests for this helper function are included in maybe_substitute_hour_minute
+        >>> extract_hour_minute_token(["3", "o'clock"])
+        -2, 3
+        >>> extract_hour_minute_token(["Gibberish", "twice"])
+        -1, 12
+        >>> extract_hour_minute_token(["only one value"])
+        -1 12
+    """
+
+    # look at previous n tokens
+    n = 2
+    for i in range(1, n+1):
+        try:
+            return -i, extract_hour_minute(tokens[-i], time_of_day)
+        # if nothing is returned from extract_hour_minute
+        except ValueError:
+            pass
+        # if the tokens list is only of length 1
+        except IndexError:
+            pass
+    # default return value
+    return -1, 12
+
+
 def maybe_substitute_hour_minute(tokens):
     """Attempt to extract hour and minute.
 
@@ -340,15 +387,19 @@ def maybe_substitute_hour_minute(tokens):
     ['7/17/18', 3 pm]
     >>> maybe_substitute_hour_minute(['3', 'p.m.', '-', '4', 'p.m.'])
     [3 pm, '-', 4 pm]
+    >>> maybe_substitute_hour_minute(['5', "o'clock", 'pm'])
+    [5 pm]
+    >>> maybe_substitute_hour_minute(['12', "o'clock", 'pm'])
+    [12 pm]
     """
     remove_dots = lambda token: token.replace('.', '')
     temp_tokens = clean_tokens(tokens, remove_dots)
 
     for time_of_day in ('am', 'pm'):
         while time_of_day in temp_tokens:
             index = temp_tokens.index(time_of_day)
-            time_token = extract_hour_minute(temp_tokens[index-1], time_of_day)
-            tokens = tokens[:index-1] + [time_token] + tokens[index+1:]
+            (unchanged_index, time_token) = extract_hour_minute_token(temp_tokens[:index], time_of_day)
+            tokens = tokens[:index+unchanged_index] + [time_token] + tokens[index+1:]
             temp_tokens = clean_tokens(tokens, remove_dots)
 
     tokens = [extract_hour_minute(token, None)