Inconsistent values while using .applyrows()

[Nanthini10]

Trying to convert string(category) into numerical values. The equivalent of applying .to_numeric() in pandas

df_left = pd.DataFrame()
nelem = 5
df_left['key1'] = np.random.randint(0, 3, nelem)
df_left['key2'] = np.random.randint(0, 5, nelem)
df_left['val1'] = ['1', '2', '5', '3', '2']
df_left['val1'] = df_left['val1'].astype('category')
df_left = DataFrame.from_pandas(df_left)

df_left

  key1 key2 val1
0    2    0    1
1    0    4    2
2    1    3    5
3    1    4    3
4    0    4    2

def kernel(val1, out1, extra1):
    for i, val in enumerate(val1):
        out1[i] = int(val)

val1 = df_left['val1']
outdf = df_left.apply_rows(kernel, incols=['val1'], 
                           outcols=dict(out1=np.int64), kwargs=dict(extra1=1))
outdf

  key1 key2 val1 out1
0    2    0    1    0
1    0    4    2    1
2    1    3    5    3
3    1    4    3    2
4    0    4    2    1

outdf['out1'] should be exactly same as val1 but be of the type int

Pandas equivalent:

df = pd.DataFrame()
nelem = 5
df['key1'] = np.random.randint(0, 3, nelem)
df['key2'] = np.random.randint(0, 5, nelem)
df['val1'] = ['1', '2', '5', '3', '2']
df['val1'] = df['val1'].astype('category')
df['out1'] = pd.to_numeric(df['val1'])
df

key1	key2	val1	out1
0	1	0	1	1
1	1	1	2	2
2	0	2	5	5
3	2	3	3	3
4	2	0	2	2

[kkraus14]

It looks like apply_rows is operating on the integer representation of the categorical column, as opposed to the string values in the dictionary.

To confirm could you test if what you're seeing matches df['val1'].cat.codes?

[Nanthini10]

Yes, df['val1'].cat.codes values match the output out the kernel.

[kkraus14]

Closing as not a bug.