iterate over pyspark dataframe columns

Question

I have the following pyspark.dataframe:

age  state  name    income
21    DC    john    30-50K
NaN   VA    gerry   20-30K

I'm trying to achieve the equivalent of df.isnull().sum() (from pandas) which produces:

age      1
state    0
name     0
income   0

At first I tried something along the lines of:

null_counter = [df[c].isNotNull().count() for c in df.columns]

but this produces the following error:

TypeError: Column is not iterable

Similarly, this is how I'm currently iterating over columns to get the minimum value:

class BaseAnalyzer:
    def __init__(self, report, struct):
        self.report = report
        self._struct = struct
        self.name = struct.name
        self.data_type = struct.dataType
        self.min = None
        self.max = None

    def __repr__(self):
        return '<Column: %s>' % self.name


class BaseReport:
    def __init__(self, df):
        self.df = df
        self.columns_list = df.columns
        self.columns = {f.name: BaseAnalyzer(self, f) for f in df.schema.fields}

    def calculate_stats(self):
        find_min = self.df.select([fn.min(self.df[c]).alias(c) for c in self.df.columns]).collect()
        min_row = find_min[0]
        for column, min_value in min_row.asDict().items():
            self[column].min = min_value

    def __getitem__(self, name):
        return self.columns[name]

    def __repr__(self):
        return '<Report>'

report = BaseReport(df)
calc = report.calculate_stats()

for column in report1.columns.values():
if hasattr(column, 'min'):
    print("{}:{}".format(column, column.min))

which allows me to 'iterate over the columns'

<Column: age>:1
<Column: name>: Alan
<Column: state>:ALASKA
<Column: income>:0-1k

I think this method has become way to complicated, how can I properly iterate over ALL columns to provide vaiour summary statistcs (min, max, isnull, notnull, etc..) The distinction between pyspark.sql.Row and pyspark.sql.Column seems strange coming from pandas.

Answer 1

Have you tried something like this:

names = df.schema.names
for name in names:
    print(name + ': ' + df.where(df[name].isNull()).count())

You can see how this could be modified to put the information into a dictionary or some other more useful format.