Closed barrettk closed 2 years ago
Not sure why drone was failing here in the previous commit. typed.missing = TRUE
is a valid argument, and prevents users from passing NA
and stuff like that (assert_numeric
will allow that by default). Its not necessary (cant imagine a user actually passing that value), so I just removed it to appease drone.
@seth127 I requested Kyle to review this PR, but tagging you in case you'd like to take a look/be involved
Not sure why drone was failing here in the previous commit.
The mpn:oldest
build was failing. That uses the 2020-06-08 snapshot, which has checkmate 2.0.0:
$ curl -fSsL https://mpn.metworx.com/snapshots/stable/2020-06-08/src/contrib/PACKAGES | grep -A1 'Package: checkmate'
Package: checkmate
Version: 2.0.0
The typed.missing
argument wasn't added to checkmate until 434a2cf (Flag to control handling of missing values, 2020-10-14), which was a part of the v2.1.0 release.
So, dropping the argument (as you've already done) is the way to go, unless we bump the minimum MPN that we test against.
Thanks @barrettk . From a quick glance, this all looks good to me, but I'll let @kyleam do a real review.
Previously, test_threads had not been tested with complex models (models with multiple estimation methods). Updating
MAXEVAL/NITER
, would cause the second estimation method to inherit the first. Additionally, in cases where the first estimation method usedMAXEVAL
, and the second one usedNITER
, both would haveMAXEVAL
(not illustrated)The below example illustrates the behavior, where the second
$EST
line became SAEM instead of IMPOriginal ctl file:
The ctl file generated by the test_threads function: