BLD: Try github actions again #2743
Conversation
|
| @@ -19,7 +19,7 @@ toolz>=0.8.2 | |||
| # Scientific Libraries | |||
| numpy>=1.11.3 | |||
| pandas>=0.18.1,<=0.22 | |||
| pandas-datareader>=0.2.1 | |||
| pandas-datareader>=0.2.1,<0.9.0 | |||
There was a problem hiding this comment.
What needed the upper bounds on this and dask?
There was a problem hiding this comment.
Pandas and pandas-datareader have some conflicts. Without this bound I could not get dependencies to resolve.
| python-dateutil==2.8.1 # via alembic, datashape, matplotlib, pandas | ||
| python-editor==1.0.4 # via alembic | ||
| python-interface==1.6.0 # via -r etc/requirements.in | ||
| pytz==2020.1 # via -r etc/requirements.in, babel, pandas, trading-calendars |
There was a problem hiding this comment.
Shall we try to keep this in sync somewhat with the py27/py35 lockfile? I notice that 3.5 still has lots of older versions, but 3.6 is now up to date. I can re-run this in a 2.7 env, if that'd be helpful.
There was a problem hiding this comment.
I don't think we should spend any time maintaining the 2.7 env.
I lean similarly for 3.5, but we could try regenerating that file from scratch as well.
There was a problem hiding this comment.
Re: maintaining 3.5, I'd rather have a PR fail zipline's 3.5 CI early, instead of our trying to pull it into the platform and failing then.
| # being applied to Series or DatetimeIndex` in trading calendar | ||
| # construction. This causes nosetest to fail. | ||
| with warnings.catch_warnings(): | ||
| warnings.simplefilter("ignore", PerformanceWarning) |
There was a problem hiding this comment.
Could we wrap just the get_calendar calls, or is more needed? I think this only matters for subclasses of WithWerror (there are only 2). Think it's worth keeping that mixin @llllllllll ?
Co-authored-by: Richard Frank <richafrank@users.noreply.github.com>
Co-authored-by: Richard Frank <richafrank@users.noreply.github.com>
|
At some point we may also want to put work into moving off |
|
Oh, is there a way for us to cache the ta-lib install? I notice it take a minute and a half on ubuntu. |
|
RE: caching - I think we could get the ubuntu caching to work if we could figure out which files were created/affected by |
| ) | ||
| # Silence numpy DeprecationWarnings which cause nosetest to fail | ||
| with warnings.catch_warnings(): | ||
| warnings.simplefilter("ignore", DeprecationWarning) |
There was a problem hiding this comment.
Looks like this is from using system-dependent np.integer instead of int64 or int32. My initial thought is we should be stricter, since we control the data format. I say we leave what you have here and open an issue.
Supersedes: