Perform usual types and values checks on columns of a pandas.DataFrame

pandas-checks.py

# Create Dataframe with fake data
df = pd.util.testing.makeMissingDataframe()
df['index1'] = df.index # create a text column by replicating index
df['A'] = 0 # create a zero column

# Helper function
def check_df_sanity(df, verbose=False):
    """Perform usual types and values checks on columns of a pandas.DataFrame"""
    
    for col in df:
        # Types checks
        is_numeric = pd.api.types.is_numeric_dtype(df[col])
        is_string = pd.api.types.is_string_dtype(df[col])
        is_date = pd.api.types.is_datetime64_any_dtype(df[col])
        is_re = pd.api.types.is_re_compilable(col)
        
        # NaN checks
        nulls = df[col].isnull()
        is_null = nulls.values.any()
        nulls = nulls.sum()
        
        # Values checks
        average = 0
        std = 0
        has_zeros = False
        zeros = 0
        has_negatives = False
        negatives = 0
        
        num_elem = df[col].size
        
        if(is_numeric):
	    # average ± standard deviation
            average = np.mean(df[col])
            std = np.std(df[col])
			
	    # zero values
            zeros = (df[col] == 0)
            has_zeros = zeros.any()
            zeros = zeros.sum()
			
	    # negative values
            negatives = (df[col] < 0)
            has_negatives = negatives.any()
            negatives = negatives.sum()
        
        # Report
        print("column %s :" % col)

        if(is_numeric):
            print("\tis numeric")
            print("\thas average = (%f ± %f)" % (average, std))
            if(has_zeros):
                print("\thas %i zero values (%.2f %%)" % (zeros, 100. * zeros / num_elem))
            if(has_negatives):
                print("\thas %i negative values (%.2f %%)" % (negatives, 100. * negatives / num_elem))

        if(is_string):
            print("\tis string")

        if(is_date):
            print("\tis date")
            
        if(is_null):
            print("\thas %i missing entries (%.2f %%)" % (nulls, 100. * nulls / num_elem))

        if(verbose):
            # less used/useful checks go here
            if(is_re):
                print("\thas a title that can be used in regex")
            
# test
check_df_sanity(df)

# example output
"""
column A :
	is numeric
	has average = (0.000000 ± 0.000000)
	has 30 zero values (100.00 %)
column B :
	is numeric
	has average = (0.030225 ± 1.107637)
	has 13 negative values (43.33 %)
	has 6 missing entries (20.00 %)
column C :
	is numeric
	has average = (-0.169237 ± 0.999756)
	has 14 negative values (46.67 %)
	has 3 missing entries (10.00 %)
column D :
	is numeric
	has average = (-0.141427 ± 1.015646)
	has 18 negative values (60.00 %)
column index1 :
	is string
"""