Data corruption in DataFrame when creating new columns while iterating over the index

[rafaljozefowicz]

That's the shortest example I could find. Data gets somehow corrupted when creating new columns while iterating over the index. Pre-generating initial values for columns solved my problem (commented out part)
I am using pandas 0.9.0

df = pd.DataFrame(index=[0,1])
df[0] = nan
wasCol = {}
# uncommenting these makes the results match
#for col in xrange(100, 200):
#    wasCol[col] = 1
#    df[col] = nan

for i, dt in enumerate(df.index):
    for col in xrange(100, 200):
        if not col in wasCol:
            wasCol[col] = 1
            df[col] = nan
        df[col][dt] = i

myid = 100
print(len(df.ix[isnan(df[myid]), [myid]]))
print(len(df.ix[isnan(df[myid]), [myid]]))

Outputs:

0
1

[wesm]

Looks fine on the development version. I'll see if I can figure out what was going wrong so I can add a unit test

[wesm]

Well, I can verify this was fixed in 5d6e7c8-- the bug is still present in v0.9.1. I would suggest upgrading to 0.10.0 or the development version as soon as practical. I'll add a unit test for your example