Setting values in Pandas DataFrame with .at, incomprehensible type error

Question

I'm constructing a pandas Dataframe for some ML. The X Dataframe has a date index composed of all existing dates from my various data files:

all_index=set()
for table in data:
    for date in table.index.values:
        all_index.add(date)

Then I construct my datavariable where I want to consolidate every data I have:

temp2= np.empty((len(all_index),1,))
temp2[:]=np.nan
X=pd.DataFrame(temp2, all_index)

And, of course, now I want to fill it with the data (data is 1 DF, later on, it will be a list of DF):

for i in X.index.values:
    for j in data[0].index.values:
        if(j==i):
            X.at[i, 0] = data['Column Name'][i]

The error is :

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-73-73562c8b1e98> in <module>
      8             #X[i]=data[0]['BCH-USD'][i]
      9             elem = data[0]['BCH-USD'][str(i)]
---> 10             X.at[i, 0] = elem
     11             #print(X[0][i])
     12             print(data[0]['BCH-USD'][i])

~\AppData\Local\Programs\Python\Python37\lib\site-packages\pandas\core\indexing.py in __setitem__(self, key, value)
   2157         key = list(self._convert_key(key, is_setter=True))
   2158         key.append(value)
-> 2159         self.obj._set_value(*key, takeable=self._takeable)
   2160 
   2161 

~\AppData\Local\Programs\Python\Python37\lib\site-packages\pandas\core\frame.py in _set_value(self, index, col, value, takeable)
   2580             series = self._get_item_cache(col)
   2581             engine = self.index._engine
-> 2582             engine.set_value(series._values, index, value)
   2583             return self
   2584         except (KeyError, TypeError):

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.set_value()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.set_value()

pandas/_libs/src\util.pxd in util.set_value_at()

pandas/_libs/src\util.pxd in util.set_value_at_unsafe()

ValueError: setting an array element with a sequence.

What I tried :

This error is weird since set_value is Deprecated. And the doc page says to use .at. And .at uses set_value...

I also tried to see the type of the variables type(data['Column Name'][i]) -> it's float64

I also tried to convert with pd.is_numeric. Same error

I try to print out data['Column Name'][i] in the loop, no error. If I try to print out X, also no error.

If I try without loop : X.at['2018-11-24', 0] = data['Column Name'][0] It works...

I expect to get:

A DataFrame with as index all dates in my multiple csv files, as column the values (if available) from my csv files. If not available, just nan.

score 0 · Answer 1 · answered Jan 13 '19 at 12:33

0

Probably the error comes from the fact that you are using chain indexing, which is something you should avoid. Try using:

for i in X.index.values:
    for j in data[0].index.values:
        if(j==i):
            X.at[i, 0] = data['Column Name'].at[i]

Also, look at this documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html

answered Jan 13 '19 at 12:33

danie

155
1
10

thanks. but the error is still the same. Something I also do not understand: why is my variable X filled with the correct values with the error appearing? – Yves Jan 13 '19 at 18:31
Do you have any duplicates in the index? – danie Jan 13 '19 at 19:09
No since the index has been build from a set (contains only unique values) – Yves Jan 13 '19 at 19:30
I see in the traceback that you are assigning a string to the dataframe, maybe this answer can help you? https://stackoverflow.com/a/4675383/9668634 – danie Jan 13 '19 at 19:40

score 0 · Accepted Answer · answered Jan 16 '19 at 14:51

I finally managed to solve my problem by using join(), which I didn't manage before.

First I build a temp DataFrame that will have all Dates possible :

all_dates=set()
for table in data:
    for ind in table.index.values:
        all_dates.add(table['Date'][ind])

dates_list=list(all_dates)
Data={'Date': dates_list}
temp=pd.DataFrame(Data)
temp.sort_values(by=['Date'], inplace=True, ascending=True)
temp=temp.reset_index(drop=True)

Then, I joined that temp Dataframe to my list of Dataframe extracted from CSVs (on top since it has most indexes):

data.insert(0,temp)
dfs = [df.set_index('Date') for df in data]
df_final=dfs[0].join(dfs[1:])

So, df_final has the Dates as an index (sorted) and the columns are the columns of the extracted Dataframes only.

The advantage of this method is that when one source of data is not complete, df_final will have "nan" there instead of a row less and losing all values from the other sources for that date.

Setting values in Pandas DataFrame with .at, incomprehensible type error

2 Answers2