search

jepusto / lmeInfo

Information matrices for fitted nlme::lme and nlme::gls models
https://jepusto.github.io/lmeInfo/
4 stars 2 forks source link

Prepare for CRAN re-submission #44

Closed manchen07 closed 2 years ago

manchen07 commented 2 years ago

I've run the win-builder checks (devel, release, oldrelease) and the rhub checks on various platforms. The results look good.

I ran the devtools::check(env_vars = c(_R_CHECK_DEPENDSONLY= TRUE)) and fixed the unit tests and removed one example.

The revdep check shows that scdhlm 0.5.2 is broken. Because we imported lmeInfo to scdhlm and the removal of returnModel in g_mlm() seems causing the problem for some example and test in scdhlm package. See revdep/problems.md for more detail.

Could you please let me know what's the next step? Thanks!

manchen07 commented 2 years ago

And I'm re-installing R now...Have been using 4.1.0 for too long...

jepusto commented 2 years ago

And I'm re-installing R now...Have been using 4.1.0 for too long...

Ha. R 4.2.1 is scheduled for release next week (6/23).

manchen07 commented 2 years ago

And I'm re-installing R now...Have been using 4.1.0 for too long...

Ha. R 4.2.1 is scheduled for release next week (6/23).

Yes...I just saw that...I guess I'll stick to 4.1.0 for another week.

jepusto commented 2 years ago

For the reverse-dependency on scdhlm, I can think of two possible options:

  1. Add a ... argument to g_mlm(), but let it have no effect on the output. Then, after we've posted a new version of scdhlm to CRAN (and corrected the unit tests/examples that specify returnModel), we can take the ... out for the next version of lmeInfo.
  2. Correct the unit tests/examples in scdhlm, post the new version to CRAN, and only then post the new version of lmeInfo.
jepusto commented 2 years ago

I pushed a commit to avoid using {dplyr} and {tidyr} in the unit tests, so that we can remove them from SUGGESTS.

jepusto commented 2 years ago

For the reverse-dependency on scdhlm, I can think of two possible options:

  1. Add a ... argument to g_mlm(), but let it have no effect on the output. Then, after we've posted a new version of scdhlm to CRAN (and corrected the unit tests/examples that specify returnModel), we can take the ... out for the next version of lmeInfo.
  2. Correct the unit tests/examples in scdhlm, post the new version to CRAN, and only then post the new version of lmeInfo.

Which direction do you think we should go? (Or perhaps some other strategy?)

manchen07 commented 2 years ago

For the reverse-dependency on scdhlm, I can think of two possible options:

  1. Add a ... argument to g_mlm(), but let it have no effect on the output. Then, after we've posted a new version of scdhlm to CRAN (and corrected the unit tests/examples that specify returnModel), we can take the ... out for the next version of lmeInfo.
  2. Correct the unit tests/examples in scdhlm, post the new version to CRAN, and only then post the new version of lmeInfo.

Which direction do you think we should go? (Or perhaps some other strategy?)

We would submit a new version of scdhlm to CRAN within one month after adding the 3-level models, right? Then maybe we could go with the second direction. But I'm okay with the first direction as well. The package is ready for submission (after adding ...). I'm more than happy to submit it to CRAN this week. Please let me know what you decide.

jepusto commented 2 years ago

Let’s go with the … approach then. It would be nice to get lmeInfo up on CRAN so that we can leave it alone for a while.

On Jun 14, 2022, at 5:40 PM, Man Chen @.***> wrote:

 For the reverse-dependency on scdhlm, I can think of two possible options:

Add a ... argument to g_mlm(), but let it have no effect on the output. Then, after we've posted a new version of scdhlm to CRAN (and corrected the unit tests/examples that specify returnModel), we can take the ... out for the next version of lmeInfo. Correct the unit tests/examples in scdhlm, post the new version to CRAN, and only then post the new version of lmeInfo. Which direction do you think we should go? (Or perhaps some other strategy?)

We would submit a new version of scdhlm to CRAN within one month after adding the 3-level models, right? Then maybe we could go with the second direction. But I'm okay with the first direction as well. The package is ready for submission (after adding ...). I'm more than happy to submit it to CRAN this week. Please let me know what you decide.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because your review was requested.

manchen07 commented 2 years ago

Just added it. I'm re-running all checks.

manchen07 commented 2 years ago

I've pushed the commits adding ... argument and revdepcheck results to cran-comments. Unfortunately, the rhub checks are not running right now. I can rerun them tomorrow.