BUG: Timestamp constructor not handling timezone conversions correctly

[stephenwlin]

Timestamp constructor raises an exception when given iso8601 string with UTC offset and a desired timezone. When given datetime object with timezone and a desired timezone, it gives a correct string but incorrect internal timestamp value.

In [1]: from pandas import Timestamp as ts, DataFrame as df
In [2]: from datetime import datetime as dt
In [3]: from pytz import timezone as tz
In [4]: UTC='UTC'
In [5]: EST='US/Eastern'
In [6]: CST='US/Central'
In [7]: # all the following should have the same timestamp value
In [8]: # (commented lines throw exceptions!!)
In [9]: tss=[ts("2012-01-02 01:00:00+00:00"),
   ...: #    ts("2012-01-02 01:00:00+00:00", tz=EST),
   ...:      ts("2012-01-01 20:00:00", tz=EST),
   ...:      ts(dt(2012,1,1,20,0,0,tzinfo=tz(EST))),
   ...:      ts(dt(2012,1,1,19,0,0,tzinfo=tz(CST))),
   ...:      ts(dt(2012,1,1,20,0,0,tzinfo=tz(EST)), tz=EST),
   ...:      ts(dt(2012,1,1,19,0,0,tzinfo=tz(CST)), tz=EST),
   ...:      ts("2012-01-01 20:00:00-05:00"),
   ...:      ts("2012-01-01 19:00:00-06:00"),
   ...: #    ts("2012-01-01 20:00:00-05:00", tz=EST),
   ...: #    ts("2012-01-01 19:00:00-06:00", tz=EST),
   ...:      ts(1325466000000000000, tz=UTC),
   ...:      ts(1325466000000000000, tz=EST)]

In [10]: strs = [str(ts) for ts in tss]
In [11]: vals = [(ts.value/1000000000000) for ts in tss]
In [12]: df({'strs' : strs, 'vals' : vals})
Out[12]: 
                        strs     vals
0  2012-01-02 01:00:00+00:00  1325466
1  2012-01-01 20:00:00-05:00  1325466
2  2012-01-01 20:00:00-05:00  1325466
3  2012-01-01 19:00:00-06:00  1325466
4  2012-01-01 20:00:00-05:00  1325448 <-- bad value!!
5  2012-01-01 20:00:00-05:00  1325448 <-- bad value!!
6  2012-01-01 20:00:00-05:00  1325466
7  2012-01-01 19:00:00-06:00  1325466
8  2012-01-02 01:00:00+00:00  1325466
9  2012-01-01 20:00:00-05:00  1325466

[stephenwlin]

after fix:

In [1]: from pandas import Timestamp as ts, DataFrame as df
In [2]: from datetime import datetime as dt
In [3]: from pytz import timezone as tz
In [4]: UTC='UTC'
In [5]: EST='US/Eastern'
In [6]: CST='US/Central'
In [7]: # all the following should have the same timestamp value
In [8]: tss=[ts("2012-01-02 01:00:00+00:00"),
   ...:      ts("2012-01-02 01:00:00+00:00", tz=EST),
   ...:      ts("2012-01-01 20:00:00", tz=EST),
   ...:      ts(dt(2012,1,1,20,0,0,tzinfo=tz(EST))),
   ...:      ts(dt(2012,1,1,19,0,0,tzinfo=tz(CST))),
   ...:      ts(dt(2012,1,1,20,0,0,tzinfo=tz(EST)), tz=EST),
   ...:      ts(dt(2012,1,1,19,0,0,tzinfo=tz(CST)), tz=EST),
   ...:      ts("2012-01-01 20:00:00-05:00"),
   ...:      ts("2012-01-01 19:00:00-06:00"),
   ...:      ts("2012-01-01 20:00:00-05:00", tz=EST),
   ...:      ts("2012-01-01 19:00:00-06:00", tz=EST),
   ...:      ts(1325466000000000000, tz=UTC),
   ...:      ts(1325466000000000000, tz=EST)]

In [9]: strs = [str(ts) for ts in tss]
In [10]: vals = [(ts.value/1000000000000) for ts in tss]
In [11]: df({'strs' : strs, 'vals' : vals})
Out[11]: 
                         strs     vals
0   2012-01-02 01:00:00+00:00  1325466
1   2012-01-01 20:00:00-05:00  1325466
2   2012-01-01 20:00:00-05:00  1325466
3   2012-01-01 20:00:00-05:00  1325466
4   2012-01-01 19:00:00-06:00  1325466
5   2012-01-01 20:00:00-05:00  1325466
6   2012-01-01 20:00:00-05:00  1325466
7   2012-01-01 20:00:00-05:00  1325466
8   2012-01-01 19:00:00-06:00  1325466
9   2012-01-01 20:00:00-05:00  1325466
10  2012-01-01 20:00:00-05:00  1325466
11  2012-01-02 01:00:00+00:00  1325466
12  2012-01-01 20:00:00-05:00  1325466