Config proposal (API Simplification)

[DBusAI]

This is number 3 from #83 , @mhaas , what do you think?

import json
class Feature:
    def __init__(self,columns:list=[]):
        self.columns = columns
        self.repr = self._get_def()
    @property
    def featuretype(self):
        pass
    def _get_def(self):
        cols = []
        for col in self.columns:
            cols.append({"label": col, "type": self.featuretype})
        return cols
    def __add__(self,other):
        self.repr=self.repr+other.repr
        return self
    def __repr__(self):
        return str(self.repr)

class Category(Feature):
    def __init__(self,columns:list=[]):
        super(Category,self).__init__(columns)
    @property
    def featuretype(self):
        return "CATEGORY"
    
class Text(Feature):
    def __init__(self,columns:list=[]):
        super(Text,self).__init__(columns)
    @property
    def featuretype(self):
        return "TEXT"

class Number(Feature):
    def __init__(self,columns:list=[]):
        super(Number,self).__init__(columns)
    @property
    def featuretype(self):
        return "NUMBER"

class Schema:
    def __init__(self,features=None,labels=None,name=None):
        self.features = features
        self.labels = labels
        self.name = name
        self.repr = self._get_def()
    def _get_def(self):
        return json.dumps({
                            "features": self.features.repr,
                            "labels": self.labels.repr,
                            "name": self.name,
                            },indent=2)
    def __repr__(self):
        return str(self.repr)

features = Text(['MYTEXT'])+Number(['ANumeric'])+Category(['Manufacturee','description'])
labels = Category(['Label1','Label2'])+Number(['Label3'])
schema = Schema(features,labels,"bestbuy-category-prediction")
print(schema)

result:
{
"features": [
{
"label": "MYTEXT",
"type": "TEXT"
},
{
"label": "ANumeric",
"type": "NUMBER"
},
{
"label": "Manufacturee",
"type": "CATEGORY"
},
{
"label": "description",
"type": "CATEGORY"
}
],
"labels": [
{
"label": "Label1",
"type": "CATEGORY"
},
{
"label": "Label2",
"type": "CATEGORY"
}
],
"name": "bestbuy-category-prediction"
}

[mhaas]

Hi @DBusAI, this is very cool!

Can you open a PR with this change? You can also install pip install pre-commit and run pre-commit -a to automatically format your code according to our standards and also run all the tests.

Perhaps some preliminary feedback: the syntactic sugar you create here by overloading the + operator is nice, but it might be a bit too magic. My main concern here is that Text(['MYTEXT']) + Number(['ANumeric']) would permanently change one object. Perhaps it could return a new object instead?

[mhaas]

Thinking about this a bit more, I also like the following, very simply approach:

features = [Category("foo"), Text("bar")]
labels = [Category("MyLabel")]

schema = Schema(features, labels, "bestbuy-category-prediction")

[DBusAI]

New code version:

class Feature:
    def __init__(self,column:str):
        self.column = column
        self.repr = self._get_def()
    @property
    def featuretype(self):
        pass
    def _get_def(self):
        return {"label": self.column, "type": self.featuretype}
    def __repr__(self):
        return str(self.repr)

class Category(Feature):
    def __init__(self,column:str):
        super(Category,self).__init__(column)
    @property
    def featuretype(self):
        return "CATEGORY"
    
class Text(Feature):
    def __init__(self,column:str):
        super(Text,self).__init__(column)
    @property
    def featuretype(self):
        return "TEXT"

class Number(Feature):
    def __init__(self,column:str):
        super(Number,self).__init__(column)
    @property
    def featuretype(self):
        return "NUMBER"

class Schema:
    def __init__(self,features=None,labels=None,name=None):
        self.features = features
        self.labels = labels
        self.name = name
        self.repr = self._get_def()
    def _get_def(self):
        return { "features": [i.repr for i in self.features],
                 "labels": [i.repr for i in self.labels],
                 "name": self.name
               }
    def __repr__(self):
        return self.repr
    
    def __call__(self):
        return self.repr
    
features = [Category("manufacturer"), Text("description"),Number("price")]
labels = [Category("level1_category"),Category("level2_category"),Category("level3_category")]

schema = Schema(features, labels, "bestbuy-category-prediction")()