When a series contains all strings, converting to dtype "int" generates bogus int values

[lbeltrame]

This works:

In [3]: series = pandas.Series(["car", "house", "tree", 1])

In [4]: series
Out[4]: 
0      car
1    house
2     tree
3        1

In [5]: series.astype("int")
ValueError: invalid literal for long() with base 10: 'tree'

This doesn't:

In [7]: series = pandas.Series(["car", "house", "tree","1"])

In [8]: series.astype("int")
Out[8]: 
0                  0
1           44735616
2    139883221510192
3                  1

This with RC2 state from git.

[wesm]

That's no good. I'll look into it

[wesm]

Irritating NumPy brokenness:

In [3]: arr = np.array(["car", "house", "tree","1"], dtype=object)

In [4]: arr
Out[4]: array([car, house, tree, 1], dtype=object)

In [5]: arr.astype(int)
Out[5]: array([4454445056, 4454445056, 4482056920,          1])

[makmanalp]

For posterity, if you hit this message and arrive here through the helpful bug id in the backtrace, (long-fixed bug aside) it means you're trying to convert something into an int that can't be converted into an int, and the error message is telling you the exact thing that failed: invalid literal for int() with base 10: blablabla. Don't be a bonehead like me, and read the error message instead. :-)