Relationship of the SmartDataframe and Pandas' DataFrame

[nautics889]

Hello.
This thread is inspired by Issue #532.

---

The PandasAI's SmartDataframe class is inherited from DataframeAbstract. In it's turn, DataframeAbstract represents an interface containing a method set matching the one from Pandas DataFrame class. So, as i undestand, it's important to keep consistency between SmartDataframe and Pandas' DataFrame classes (their behaviour, signatures and side effects).

Therefore, when i started adding .to_dict() method (#532 ) it seemed kind of confusing for me: to put all implementation details "as is" from pandas library in SmartDataframe.

---

Shouldn't we consider inheriting SmartDataframe from pandas.core.frame.DataFrame?
Or the methods supposed to have their own, independent implementation within PandasAI library?

---

List of method from DataframeAbstract:

.abs()
.append()
.apply()
.applymap()
.columns()
.concat()
.corr()
.count()
.cov()
.describe()
.drop_duplicates()
.dropna()
.duplicated()
.expanding()
.fillna()
.filter()
.groupby()
.head()
.hist()
.index()
.isna()
.join()
.max()
.mean()
.median()
.merge()
.min()
.notna()
.nunique()
.pipe()
.pivot()
.plot()
.prod()
.query()
.rename()
.resample()
.reset_index()
.rolling()
.sample()
.set_index()
.std()
.sum()
.tail()
.to_csv()
.to_dict()
.to_json()
.to_markdown()
.to_numpy()
.to_sql()
.unstack()
.value_counts()

All of the above are present in pandas.core.frame.DataFrame (pandas v1.5.3) except .concat().

[nautics889]

Well, after writing this topic, i understood the obvious: SmartDataframe can be a proxy for original pandas DataFrame rather then inherit one. So, all the methods can simply do proxy-calls to the original dataframe.

Like this:

class SmartDataframe(DataframeAbstract, Shortcuts):

    ...

    def to_dict(...):
        return self._df.to_dict(...)

    ...

---

@gventuri correct me if i'm wrong pls :)