featureforge

In [6]:

    

from featureforge.feature import input_schema, output_schema

@input_schema({"body": str})
@output_schema(int, lambda i: i >= 0)
def body_length(message):
    return len(message["body"])


    

In [7]:

    

record = {}
record['body'] = 'this is an example'
body_length(record)


    

    
Out[7]:





18

In [8]:

    

record = {}
record['body'] = 555
body_length(record)


    

    




---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-8-20cbaa6cff4b> in <module>()
      1 record = {}
      2 record['body'] = 555
----> 3 body_length(record)

<ipython-input-6-bb77d45f5a28> in body_length(message)
      4 @output_schema(int, lambda i: i >= 0)
      5 def body_length(message):
----> 6     return len(message["body"])

TypeError: object of type 'int' has no len()

In [17]:

    

%%writefile english-badwords.txt
bad
word


    

    




Writing english-badwords.txt

In [29]:

    

from featureforge.feature import Feature, soft_schema
from schema import Schema

class SubjectHasBadWord(Feature):
    input_schema = soft_schema(subject=str)
    # The above is equivalent to
    #   input_schema = Schema({"subject": str, str: Optional(object)})
    output_schema = Schema(bool)

    def __init__(self, bad_words_filename):
        self.bad_words = set(open(bad_words_filename).readlines())

    def _evaluate(self, message):
        subject_words = set(message["subject"].split())
        return bool(subject_words & self.bad_words)

has_bad_word_english = SubjectHasBadWord("english-badwords.txt")


    

In [30]:

    

record = {}
record['subject'] = 'this is a bad word'
has_bad_word_english(record)


    

    
Out[30]:





True

In [32]:

    

record = {}
record['subject'] = 'this is another one'
has_bad_word_english(record)


    

    
Out[32]:





False