Pandas 0.18 doesn't like having null-ish values in categoricals. Fixing
this properly requires re-thinking the semantics for missing_value on
pipeline terms, so we're punting on that until after we've upgraded to
0.18.
Pandas 0.18 deprecated passing "null-ish" values to pd.categorical. The
expectation, instead, is that you use categorical's native support for
missing data, which means the user will always get NaN's for missing
entries of the categorical.
A follow-up to this change should probably drop support for custom
missing values entirely and to use LabelArray/categorical for integer
data.
Remove module scope invocations of `get_calendar('NYSE')`, which cuts
zipline import time in half on my machine. This make the zipline CLI
noticeably more responsive, and it reduces memory consumed at import
time from 130MB to 90MB.
Before:
$ time python -c 'import zipline'
real 0m1.262s
user 0m1.128s
sys 0m0.120s
After:
$ time python -c 'import zipline'
real 0m0.676s
user 0m0.536s
sys 0m0.132s
This reverts commit 5b1aa5ec55.
The paradigm is: we're calculating a new capital base for the
performance period. We are therefore using the total
portfolio_value, not just the cash, to calculate the
difference from the specified target as the algorithm
has meaningful holdings.
They're not meaningful, and they cause warnings from numpy.
Implemented in terms of a new preprocessor, `expect_bounded`, which
takes a tuple of `upper_bound` and `lower_bound`.
test_resample now fully covers the resample module.
Fix a bug exposed by increased coverage, where daily aggregation on
`high` would return `nan` for an asset instead of 1) during the
course of day `1d` history was called on non-consecutive minutes and 2)
either, a) the value for the previously inspected dt was `nan` or b)
there were only `nan`s between the previous and current dt.
`low` had a similar bug which was only triggered if the value for the
previously inspected dt was `nan`.
If subclasses have additional capital change information that
is required to correctly calculate the target values for
cash capital changes, it can now be provided via
"portfolio_value_adjustment".
Increase coverage on `ReindexSessionBarReader` so that all methods which
are considered part of the interface are covered by `test_resample`.
Fix bug in `get_value`, exposed by increased coverage, where the
`NoDataOnDate` exception was bubbling from the bcolz reader all the way
up when a session which was a holidy on the underlying reader was passed
to the reindex reader. (The reindex reader should return nan/0 in that
case.)
Also, move location of data index exceptions so that they are agnostic
to bcolz/us_equity_pricing; since the exception is now used by the
resample module to fix aforementioned bug.
BarData now takes the trading calendar as a parameter.
can_trade now checks if the asset’s exchange is open at the current or
next market minute (defined by the given trading calendar).
Remove special handling for the last session of an asset, which was
moving the last traded back a session.
If the asset has data on a session, `get_last_traded_dt` should always
return that session if it is the parameter to the method.
No longer auto-updates its internal as-of date, instead requires an explicit
as-of date from the consumer.
Take a static list of contracts (instead of needing an assetfinder).
Instead of the as_of method, the user-facing API now lets you pass in an
offset, which is defined as an integral number of sessions.
Add direct coverage on last_available_dt.
Also move reader creation into the instance fixture.
This patch attempted to add coverage on `get_last_traded_dt`, but in doing
so, revealed a bug in `BcolzDailyBarReader.get_last_traded_dt` when
requesting the last trading session of an asset.
When that is fixed, the skip can be removed.
Scott Sanderson