CODEWITHSUNDEEP
pythonpandaslogicseriescase-sensitive
npross
npross

Reputation: 1888

Case sensitive Pandas Series matching and clean Panda Series Logic

I have a pandas dataFrame of fruits:: 

df = pd.read_csv(newfile, header=None)
df
             0        1        2             3        4        5    6   7
0        Apple  Bananas      Fig    Elderberry   Cherry    Honeydew NaN NaN 
1      Bananas   Cherry   Dragon    Elderberry      NaN         NaN NaN NaN
2       Cherry    Grape      NaN           NaN      NaN         NaN NaN NaN
3       Dragon      NaN    Apple        Bananas  Cherry  Elderberry NaN NaN
4   Elderberry    Apple  Bananas            Fig   Grape         NaN NaN NaN
5          Fig   Cherry Honeydew          Apple     NaN         NaN NaN NaN
6        Grape      NaN      NaN            NaN     NaN         NaN NaN NaN
7     Honeydew    Grape      Fig     Elderberry  Dragon      Cherry Bananas Apple

And I'm trying to find "fruit pairings", e.g. in the first row, Apple and Fig are a pair, and 6th row Fig and Apple. Likewise for Apple-Elderberry and Elderberry-Apple, but not Apple and Bananas (there are no Apples in the row starting with Bananas).

I've got the following code working, and that does this::

fruits = df[0]
stock  = df.drop(0, axis=1)

for i in range(len(fruits)):
    string1 = str(fruits[i])
    full_line = (stock.iloc[i])
    line = np.array(full_line.dropna(axis=0))
    if len(line) > 0 : 
        for j in range(len(stock)):
            iind = (fruits[fruits == line[j]].index[0])
            this_line = stock.iloc[iind]
            logic_out = this_line.str.match(string1)
            print(logic_out)

BUT!! (1) It breaks at the fruits == line[j] due the Pandas Series being case sensitive and (2) the boolean out put is a mixture of True's, Falses and NaNs. Ideally, I just want to count the Trues. Any and all help v. much appreciated!!

Upvotes: 0

Views: 198

Answers (1)

piRSquared
piRSquared

Reputation: 294358

I'm going to use set logic, pandas stacking, and numpy broadcasting

f = lambda x: x.title() if isinstance(x, str) else x

s = df.applymap(f).set_index('0').rename_axis(None).stack().groupby(level=0).apply(set)

f = s.index
p = s.values

one_way = (p[:, None] & [{x} for x in f]).astype(bool)
[(f[i], f[j]) for i, j in zip(*np.where(one_way & one_way.T))]

[('Apple', 'Elderberry'),
 ('Apple', 'Fig'),
 ('Apple', 'Honeydew'),
 ('Bananas', 'Dragon'),
 ('Bananas', 'Elderberry'),
 ('Dragon', 'Bananas'),
 ('Elderberry', 'Apple'),
 ('Elderberry', 'Bananas'),
 ('Fig', 'Apple'),
 ('Fig', 'Honeydew'),
 ('Honeydew', 'Apple'),
 ('Honeydew', 'Fig')]

Upvotes: 1

Related Questions