How to separate values from pandas series into dictionary?-CodePudding

I have a pandas series like this:

LIST  0     ITEM1
1           Element1
2           Element2
3           Element3           
4           Element4
5           Element5
6           Element6
7           Element7
8           ITEM2
9           Element8
10          ELEMENT9
11          ELEMENT10
12          Element11
13          Element12      
14          Element13
15          Element14
16          Element2
17          Element24
18          Element25
19          Element26
20          ITEM3
21          Element28
Name: Items, dtype: object

I would like to separate the Items from the element objects. In the real example, the elements aren't all called 'Elements' and the same happens for 'Items', so I cannot connect the code to the naming (as in contains 'elements' and contains 'items'). I would need to access the values by dictionary keys or by dataframe columns. For example:

df['ITEMS1'] should give the first elements: Element1 to Element7.
or dict['ITEMS'] should be connected to the first 7 elements as well.

How can I separate the elements from the items?

CodePudding user response：

Create groups using str.contains and cumsum. For each group create a list of tuple before exporting as a list. Finally convert the list of tuple to a dict.

d = dict(df.groupby(df['LIST'].str.contains('^ITEM').cumsum())['LIST'] \
           .apply(lambda x: (x.iloc[0], x.iloc[1:].tolist())).tolist())

print(d)

# Output:
{'ITEM1': ['Element1',
  'Element2',
  'Element3',
  'Element4',
  'Element5',
  'Element6',
  'Element7'],
 'ITEM2': ['Element8',
  'ELEMENT9',
  'ELEMENT10',
  'Element11',
  'Element12',
  'Element13',
  'Element14',
  'Element2',
  'Element24',
  'Element25',
  'Element26'],
 'ITEM3': ['Element28']}

Note: you have to find a way to distinguish Item from Element but the idea remains the same.

CodePudding user response：

You can use dict comprehension:

print ({i.iloc[0]: i.iloc[1:].tolist() for _, i in df.groupby(df["Items"].str.startswith("ITEM").cumsum())["Items"]})

{'ITEM1': ['Element1', 'Element2', 'Element3', 'Element4', 'Element5', 'Element6', 'Element7'],
 'ITEM2': ['Element8', 'ELEMENT9', 'ELEMENT10', 'Element11', 'Element12', 'Element13', 'Element14',
           'Element2', 'Element24', 'Element25', 'Element26'],
 'ITEM3': ['Element28']}

CodePudding user response：

You can also use:

item_index = df[df["col"].str.startswith("ITEM")].index.to_list()
data = np.split(df['col'].to_numpy(), item_index)
df = pd.DataFrame(data[1:]).set_index(0).T

Output:

0      ITEM1      ITEM2      ITEM3
1   Element1   Element8  Element28
2   Element2   ELEMENT9       None
3   Element3  ELEMENT10       None
4   Element4  Element11       None
5   Element5  Element12       None
6   Element6  Element13       None
7   Element7  Element14       None
8       None   Element2       None
9       None  Element24       None
10      None  Element25       None
11      None  Element26       None