/pandas_notes_interview_prep

Pandas Shenanigans

1) Dataframe Manipulation

single column

apply/lambda
df['col1'].apply(lambda x : x.strip().capitalize())
mapping - more efficient for applying methods to a single column (series vs. entire row)
df['col1'].map(lambda x: x + 1)

#Ternary if statement mapped to single col
df['Col1'].map(lambda x: 1500 if x =='Music' else 800)

multiple columns

same function
df[['colA', 'colD']] = df[['colA', 'colD']].apply(lambda x: x + 1)
different functions
def col_func(x):
    x['col1'] = 1
    x['col2'] = x['col2'] + 'x
    return x

df.apply(col_func, axis=1)
iterating
for index, row in df.iterrows():
    print(row['col1'], row['col2'])
    
 for ind in df.index:
     print(df['name'][ind], df['age'][ind])
conditional new column
# Ternary if statement
df['Price'] = [1500 if x =='Music' else 800 for x in df['Event']]

# Apply/Map
df['Price'] = df['Event'].map(lambda x: 1500 if x =='Music' else 800)
filtering
# Boundary filtering
dataframe[dataframe['Percentage'] > 80]

# In a list
options = ['Math', 'Commerce', 'Shaurya']
rslt_df = dataframe[~dataframe['Stream'].isin(options)]

# Both conditions
rslt_df = dataframe[(dataframe['Age'] == 21) & dataframe['Stream'].isin(options)]

# Sorting 
df.sort_values(by='col1', ascending=False)
aggregations
# Simple
df.groupby(['a','b'])['c'].sum()

# Multiple aggregations
df.groupby(['Gender','State'])[['col_1', 'col_2']].agg([min, max])

# Aggregations:
#'sum', 'min', 'max', 'count', 'mean'

aggregations = {'shortcode' :'count',
                'is_video' : np.mean,
                'is_sponsored': np.mean}

# as_index so you don't have to .reset_index() later
df.groupby(company_info, as_index = False).agg(aggregations)

# With lambda function
df.groupby('Sex').Age.agg(
    ['max', 'mean', lambda x: True if x.mean() > 50 else False]
)
drop
df = df.dropna(how = 'all')
df.drop(df[df['Age'] < 25].index, inplace = True)
Column Naming
# Change all the column names
df.columns =['Col_1', 'Col_2']

# Change all the row indexes
df.index = ['Row_1', 'Row_2', 'Row_3', 'Row_4']

# Change some of the columns 
df = df.rename(columns = {"Col_1":"Mod_col"})

## add x to every column name
df = df.rename(columns=lambda x: x+'x')
types
df["somecolumn"].astype(int)
Text?
df1 = df['Position'].str.contains("PG")

2) create dataframe

  • list
  • dict
  • list of dicts
pd.DataFrame(lst, columns = ["col1", "col2"])
pd.DataFrame(dict)
pd.DataFrame(lst_of_dicts)
list
lst = [['Geek', 25], ['is', 30],
       ['for', 26], ['Geeksforgeeks', 22]]
       
pd.DataFrame(lst, columns = ["word", "number"])
dict
dict = {'col1':['val1', 'val2', 'val3'],
        'col2':[0, 1, 3]}
pd.DataFrame(dict)
list of dicts
lst_of_dicts = [{'Geeks': 'dataframe', 'For': 'using', 'geeks': 'list'},
        {'Geeks':10, 'For': 20, 'geeks': 30}]
pd.DataFrame(lst_of_dicts)
series
author = ['Jitender', 'Purnima', 'Arpit', 'Jyoti']
article = [210, 211, 114, 178]

auth_series = pd.Series(author)
article_series = pd.Series(article)

frame = { 'Author': auth_series, 'Article': article_series }

result = pd.DataFrame(frame)

3) Combinging Data

mapping new data
new_data = { "col1":"val1",
             "col2":"val2",
             "col5":"val5",
             "col3":"val3",
             "col4":"val4"
              }

# combine this new data with existing DataFrame
df["new_col"] = df["old_col_to_match"].map(new_data)
replacing new data
df[col].map({'yes':True, 'no':False})
merge/concat
df1.merge(df2, how='left', on='a')

df1.concat(df2)