search

signaturescience / focustools

Forecasting COVID-19 in the US
https://signaturescience.github.io/focustools/
GNU General Public License v3.0
0 stars 0 forks source link

bug in focustools basic usage vignette #60

Closed vpnagraj closed 3 years ago

vpnagraj commented 3 years ago

looks like when the pkgdown website built the vignette there was an issue fitting the incident case model:

https://signaturescience.github.io/focustools/articles/focustools.html

Screen Shot 2021-04-22 at 4 19 46 PM

incident case forecasts appear to be coming through as NA

@stephenturner assigning you and myself to ivestigate

stephenturner commented 3 years ago

Not looking at code at the moment, but a thing I did in this vignette was to get the data and save as sysdata not exported, and load that data in a hidden chunk (eval FALSe on the chunk showing how to do it). Not sure if that's related.

stephenturner commented 3 years ago

...pretty sure this has something to do with it... the built vignette I have locally doesn't have these NAs.

vpnagraj commented 3 years ago

fixed. turns out the vignette data was loading fine so that was not the problem.

the issue was with the auto ARIMA processing failing to find the urca and feasts packages. apparently fable / fabletools will let you fit an ARIMA model but will issue a warning (not error) that those packages are missing. the container running the pkgdown build had now way of knowing that those packages needed to be installed.

added them as Imports (see https://github.com/signaturescience/focustools/commit/0e4702b803bd1c5febcfc0fd8c0b4e9f9950588b).

i guess we hadnt run into this previously because all of the systems running the code had explicit calls to install those deps at some point (outside of the focustools DESCRIPTION).

the package is building and check-ing fine now. and vignette looks good:

https://signaturescience.github.io/focustools/articles/focustools.html#forecast-visualization-and-summary