How to query a numerical column name in pandas?

Question

Lets suppose I create a dataframe with columns and query i.e

pd.DataFrame([[1,2],[3,4],[5,6]],columns=['a','b']).query('a>1')

This will give me

   a  b
1  3  4
2  5  6

But when dataframe values are too large and I don’t have column names, how can I query a column by its index?

I tried querying by passing a number, but it’s not the way of doing it.

pd.DataFrame([[1,2],[3,4],[5,6]]).query('0>1') # This is what I tried.

How to denote 0 is the column name in query?

Expected Output:

   0  1
1  3  4
2  5  6

Asked By: Bharath

||

Source

Answer 1

Since the query is under development one possible solution is creating a monkey patch for pd.DataFrame to evaluate self i.e :

def query_cols(self,expr):
    if 'self' in expr:
        return self[eval(expr)]
    else:
        return self.query(expr)

pd.DataFrame.query_cols = query_cols

pd.DataFrame([[1,2],[3,4],[5,6]]).query_cols('self[1] > 3')

   0  1
1  3  4
2  5  6

pd.DataFrame([[1,2],[3,4],[5,6]]).query_cols('self[1] == 4')

   0  1
1  3  4

pd.DataFrame([[1,2],[3,4],[5,6]],columns=['a','b']).query_cols('a > 3')

   a  b
2  5  6

This is a simple trick and doesn’t suit all the cases, answer will be updated when the issue with query is resolved.

Answered By: Bharath

Answer 2

Solution

An option without any monkey patching is to use @ to define a variable and do this as follows.

# If you are fond of one-liners
df = pd.DataFrame([[1,2],[3,4],[5,6]]); df.query('@df[0] > 1')

# Otherwise this is the same as
df = pd.DataFrame([[1,2],[3,4],[5,6]])
df.query('@df[0] > 1') # @df refers to the variable df

Output:

   0  1
1  3  4
2  5  6

References

You can find more ways of dealing with this here.

Docs: pandas.DataFrame.query

Answered By: CypherX

Answer 3

You can create an intermediate column with assign + a lambda function:

pd.DataFrame([[1, 2], [3, 4], [5, 6]]).assign(col=lambda x: x[0]).query("col>1")

Answered By: 00schneider

How to query a numerical column name in pandas?

Question:

Answers:

Solution

References