I have a Pandas DataFrame that I'm creating row-by-row (I know, I know, it's not Pandorable/Pythonic..). I'm creating elements using .loc like so
output.loc[row_id, col_id]
and I'd like to set this value to an empty list, [].
output.loc[row_id, col_id] = []
Unfortunately, I get an error saying the size of my keys and values do not match (Pandas thinks I'm trying to set values with not to an iterable).
Is there a way to do this?
Thanks!
You can use pd.at instead:
df = pd.DataFrame()
df['B'] = [1, 2, 3]
df['A'] = None
df.at[1, 'A'] = np.array([1, 2, 3])
When you use pd.loc, pandas thinks you are interacting with a set of rows. So if you try to assign an array using pd.loc, pandas will try to match each element of an array with a corresponding element accessed by pd.loc, hence the error.
You need to make sure two things:
A hacky way to do this is to use a Series with []:
In [11]: df = pd.DataFrame([[1, 2], [3, 4]], columns=['A', 'B'])
In [12]: df.loc[[0], 'A'] = pd.Series([[]])
In [13]: df
Out[13]:
A B
0 [] 2
1 3 4
pandas doesn't really want you use [] as elements because it's usually not so efficient and makes aggregations more complicated (and un-cythonisable).
In general you don't want to build up DataFrames cell-by-cell, there is (almost?) always a better way.
The answer by MishaTeplitskiy works when the index label is 0. More generally, if you want to assign an array x to an element of a DataFrame df with row r and column c, you can use:
df.loc[[r], c] = pd.Series([x], index = [r])
