DataFrame creation incorrect error message

[sergei-sh]

The problem was already mentioned as part of other issues, but still persists in 0.22

#8020
blaze/blaze#466

Reported both expected shape and input data shape are both transposed which causes a lot of confusion. In my opinion, the reference value should be DataFrame.shape.

my_arr = np.array([1, 2, 3])
print("my_arr.shape: {}".format(my_arr.shape))
df = pd.DataFrame(index=[0], columns=range(0, 4), data=my_arr)

my_arr.shape: (3,)
Traceback (most recent call last):
...
ValueError: Shape of passed values is (1, 3), indices imply (4, 1)

Below are shapes which are expected to be reported:

my_arr = np.array([[0, 1, 2, 3]])
print("my_arr.shape: {}".format(my_arr.shape))
df = pd.DataFrame(index=[0], columns=range(0, 4), data=my_arr)
print(df.shape)`

my_arr.shape: (1, 4)
(1, 4)

I'm not sure, whether this is another issue, but in the first example, the error cause is 1-dimensional data while constructor expects 2-dimensional data. The user gets no hint about this from the error message.

[jreback]

the first example is wrong. The block manager reports this, but doesn't flip the dim (like we do for everything else), so would welcome a PR to correct that.

I don't see a problem with the 2nd. You gave a 1, 4 array. That's the same as the dim of the frame, so it constructs.