Correct Yahoo period end timestamp calculation #365
Conversation
pandas_datareader/yahoo/daily.py
Outdated
| @@ -97,7 +98,9 @@ def yurl(symbol): | |||
|
|
|||
| def _get_params(self, symbol): | |||
| unix_start = int(time.mktime(self.start.timetuple())) | |||
| unix_end = int(time.mktime(self.end.timetuple())) | |||
| day_end = self.end + datetime.timedelta(days=1) \ | |||
There was a problem hiding this comment.
Is it safe to assume self.end is a whole day only with time 0:00:00?
There was a problem hiding this comment.
It is a datetime created from year/month/day (as discussed in #356 there still might be some timezone issues ...)
The yahoo API becoming very sensitive ...
There was a problem hiding this comment.
does this need any additional tests?
There was a problem hiding this comment.
This approach is not correct.
import pandas_datareader as pdr
r = pdr.yahoo.daily.YahooDailyReader(['IBM'],'1/1/1960','1/1/2017 12:30:00')
r.end
Out[6]: Timestamp('2017-01-01 12:30:00')
Since it accepts times as well as dates you will need to create a new timestamp using the year, day and month and setting the hour to 23, minute to 59, and second to 59.
There was a problem hiding this comment.
I would add a test like this, to assert the end-points of the received data are correct (should fail before and pass after this PR)
import pandas_datareader as pdr
r = pdr.yahoo.daily.YahooDailyReader(['IBM'],'1/1/1960','1/1/2017 12:30:00')
r.end
There was a problem hiding this comment.
@jreback As discussed above, the failure was intermittent, sometimes not returning the data for 12/31. I'M m not sure what value a test on r.end adds ... Thanks
Signed-off-by: Gábor Lipták <[email protected]>
|
@jreback Green after updates |
|
thanks @gliptak |
Signed-off-by: Gábor Lipták [email protected]
xref #356