BUG: Transpose construction/block error with certain dtypes

[RobinFiveWords]

Code Sample, a copy-pastable example if possible

import pandas as pd
from io import StringIO

data = "dt1,obj1,obj2,obj3,num1\n,a,c,e,0\n,b,d,f,1"

df1 = pd.read_csv(StringIO(data), dtype={'num1': float}, parse_dates=['dt1'])
df2 = pd.read_csv(StringIO(data), dtype={'num1': int}, parse_dates=['dt1'])

print(df1.T)
print(df2.T)  # ValueError in v0.20.1 but not v0.19.2

Problem description

Tried to provide a minimal example here. I originally encountered the error with a dataframe containing datetime, object, and int dtypes, and triggered it with this...

df.iloc[[0, -1]].T

...and even found things like this:

df.iloc[[0, 1], [2, 3, 4, 9]].T  # okay
df.iloc[[0, 1], [2, 3, 4, 4, 9]].T  # ValueError
df.iloc[[0, 1], [2, 3, 4, 9, 4]].T  # okay (?!)

Expected Output

    0    1

dt1 NaT NaT
obj1 a b
obj2 c d
obj3 e f
num1 0 1
0 1
dt1 NaT NaT
obj1 a b
obj2 c d
obj3 e f
num1 0 1

Output of pd.show_versions()

# Paste the output here pd.show_versions() here INSTALLED VERSIONS ------------------ commit: None python: 3.6.0.final.0 python-bits: 64 OS: Windows OS-release: 7 machine: AMD64 processor: Intel64 Family 6 Model 61 Stepping 4, GenuineIntel byteorder: little LC_ALL: None LANG: None LOCALE: None.None

pandas: 0.20.1
pytest: 3.0.5
pip: 9.0.1
setuptools: 27.2.0
Cython: 0.25.2
numpy: 1.12.1
scipy: 0.19.0
xarray: None
IPython: 5.1.0
sphinx: 1.5.1
patsy: 0.4.1
dateutil: 2.6.0
pytz: 2016.10
blosc: None
bottleneck: 1.2.0
tables: 3.2.2
numexpr: 2.6.2
feather: None
matplotlib: 2.0.1
openpyxl: 2.4.1
xlrd: 1.0.0
xlwt: 1.2.0
xlsxwriter: 0.9.6
lxml: 3.7.2
bs4: 4.5.3
html5lib: None
sqlalchemy: 1.1.5
pymysql: None
psycopg2: None
jinja2: 2.9.4
s3fs: None
pandas_gbq: None
pandas_datareader: None

[chris-b1]

This also repros with direct construction

pd.DataFrame(np.array([[pd.NaT, 'a', 'b', 0],
                       [pd.NaT, 'b', 'c', 1]]))
# IndexError: tuple index out of range

But doesn't if only one string column

pd.DataFrame(np.array([[pd.NaT, 'a', 0],
                       [pd.NaT, 'b', 1]]))
# fine!

[RobinFiveWords]

I see some inconsistency in the maybe_infer_to_datetimelike function in pandas/core/dtypes/cast.py. v is first raveled but is only unraveled in the cases reaching line 830 or 838. When I added .reshape(shape) to end of line 840, Chris's test ran fine. What's the best approach here, to remove the reshapes in lines 830 and 838 and apply it just once, at the end of the return statement in line 865?

[chris-b1]

@RobinFiveWords - easiest to review if you would submit a PR with that change - main thing will be that existing tests still pass.