bugfix: convert index to int so that the pandas indexing doesnt fail #1683
Conversation
|
Hi @shadiakiki1986 thanks for the PR. I was wondering if you could explain/show some test where this is an issue, so I could try and reproduce it, or how you caught this |
zipline/data/minute_bars.py
Outdated
|
|
||
| # cast to int for pandas | ||
| # https://github.com/pandas-dev/pandas/blob/master/pandas/tseries/base.py#L247 | ||
| num_days = int(num_days) |
There was a problem hiding this comment.
I think that if we're going to check for an int, we should be doing this earlier, before we even check for num_days == 0
We can do that on line 552; using the // operator which does integer division in both Python 2 and 3: num_days = data['shape'][0] // self._minutes_per_day
There was a problem hiding this comment.
That's a better idea. I just pushed a new commit amending the PR
zipline/data/minute_bars.py
Outdated
| @@ -549,7 +549,9 @@ def last_date_in_output_for_sid(self, sid): | |||
| with open(sizes_path, mode='r') as f: | |||
| sizes = f.read() | |||
| data = json.loads(sizes) | |||
| num_days = data['shape'][0] / self._minutes_per_day | |||
| # use integer division so that the result is an int for pandas index later | |||
| # https://github.com/pandas-dev/pandas/blob/master/pandas/tseries/base.py#L247 | |||
There was a problem hiding this comment.
Looks like the flake8 tests failed because line 552 is too long (and probably 553 as well)
There was a problem hiding this comment.
shortened the lines
zipline/data/minute_bars.py
Outdated
| @@ -549,7 +549,9 @@ def last_date_in_output_for_sid(self, sid): | |||
| with open(sizes_path, mode='r') as f: | |||
| sizes = f.read() | |||
| data = json.loads(sizes) | |||
| num_days = data['shape'][0] / self._minutes_per_day | |||
| # use integer division so that the result is an int | |||
| # for pandas index later https://goo.gl/VAW649 | |||
There was a problem hiding this comment.
Ok great. Just one last thing; we should keep this link expanded (so people know it's a safe link to click). You can make the flake8 tests ignore the length of this line by adding # noqa line 553. So it would look like:
# for pandas index later <link> # noqa
Should've mentioned that before, sorry! (after that it should be good to merge)
There was a problem hiding this comment.
Also you should squash your commits into 1, and change the commit message to say BUG: <description> (to be consistent with our dev guidelines https://github.com/quantopian/zipline/wiki/Zipline-development-guidelines)
There was a problem hiding this comment.
No problem :) Done and rebased
shadiakiki1986
force-pushed
the
bugfix_index_should_be_int
branch
from
e29aeb3
to
045d95f
Compare
shadiakiki1986
force-pushed
the
bugfix_index_should_be_int
branch
from
045d95f
to
5bea92f
Compare
|
Btw, here is my travis build which demonstrates this issue |
|
Looks good to merge 👍 thanks for the PR |
otherwise, the index could come out as a float because of the division ahead of it