BUG: Timestamp.__new__ doesnt preserve nanosecond

[sinhrks]

When Timestamp.__new__ accepts a timestamp with nanosecond, nanosecond is not preserved properly.

# create Timestamp with ns (OK)
t = pd.Timestamp('2011-01-01') + pd.offsets.Nano(5)
t, t.value
# (Timestamp('2011-01-01 00:00:00.000000005'), 1293840000000000005)

# If it is passed to Timestamp.__init__, ns is not displayed even though internal value includes it. (NG)
t = pd.Timestamp(t)
t, t.value
# (Timestamp('2011-01-01 00:00:00'), 1293840000000000005)

# If offset is added to above result, ns is displayed properly (OK)
t = t + pd.offsets.Nano(5)
t, t.value
# (Timestamp('2011-01-01 00:00:00.000000010'), 1293840000000000010)

NOTE: Unrelated to this issue, test_tslib had test_timedelta_ns_arithmetic method duplicatelly. Thus renamed.

[sinhrks]

Modified the subject. The problem is not __init__ , but __new__

[jreback]

I think this might be a repr issue (and not a calc issue). as these tests all pass (for .value), but only Timestamp(t) doesn't have the correcvt repr (but the right value)

[sinhrks]

__repr__ refers to nanosecond, not value.
https://github.com/pydata/pandas/blob/master/pandas/tslib.pyx#L252
And nanosecond set from dst.ps, not value. Thus dst.ps must be set properly before setting nanosecond.
https://github.com/pydata/pandas/blob/master/pandas/tslib.pyx#L213

t = pd.Timestamp('2011-01-01') + pd.offsets.Nano(5)
t = pd.Timestamp(t)
t.nanosecond
# 0 (NG)

[jreback]

ok then...thanks