[R] Check warnings explicitly for model compatibility tests

[hcho3]

• Suppress warnings in test_model_compatibility.R, as they clutter test results.
• Furthermore, check that the correct types of warnings are thrown in test_model_compatibility.R.

@vspinu Can you review?

[codecov-commenter]

Codecov Report

Merging #6114 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #6114   +/-   ##
=======================================
  Coverage   78.52%   78.52%           
=======================================
  Files          12       12           
  Lines        3069     3069           
=======================================
  Hits         2410     2410           
  Misses        659      659           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1453bee...2c0ba92. Read the comment docs.

[trivialfis]

I think we can remove that warning in longer term. Just as an example, if I unify the single_precision_histogram for tree methods, the RDS breaks.

[hcho3]

Yes, the warning is meant to encourage users to move away from saveRDS() and use xgb.save() instead. The habit is hard to break, so we produce copious warnings in the user's terminal. Here, we suppress the warnings when running the unit tests.

[vspinu]

I still see some internal warnings. Maybe those could be gotten rid of as well?

⠏ |   0       | update trees in an existing model[10:46:30] WARNING: ../..//amalgamation/../src/gbm/gbtree.cc:81: DANGER AHEAD: You have manually specified `updater` parameter. The `tree_method` parameter will be ignored. Incorrect sequence of updaters will produce undefined behavior. For common uses, we recommend using`tree_method` parameter instead.
⠋ |   1       | update trees in an existing model[10:46:30] WARNING: ../..//amalgamation/../src/gbm/gbtree.cc:81: DANGER AHEAD: You have manually specified `updater` parameter. The `tree_method` parameter will be ignored. Incorrect sequence of updaters will produce undefined behavior. For common uses, we recommend using`tree_method` parameter instead.
[10:46:30] WARNING: ../..//amalgamation/../src/gbm/gbtree.cc:81: DANGER AHEAD: You have manually specified `updater` parameter. The `tree_method` parameter will be ignored. Incorrect sequence of updaters will produce undefined behavior. For common uses, we recommend using`tree_method` parameter instead.
[10:46:30] WARNING: ../..//amalgamation/../src/gbm/gbtree.cc:81: DANGER AHEAD: You have manually specified `updater` parameter. The `tree_method` parameter will be ignored. Incorrect sequence of updaters will produce undefined behavior. For common uses, we recommend using`tree_method` parameter instead.
⠇ |   9       | update trees in an existing model[10:46:30] WARNING: ../..//amalgamation/../src/gbm/gbtree.cc:81: DANGER AHEAD: You have manually specified `updater` parameter. The `tree_method` parameter will be ignored. Incorrect sequence of updaters will produce undefined behavior. For common uses, we recommend using`tree_method` parameter instead.
[10:46:30] WARNING: ../..//amalgamation/../src/gbm/gbtree.cc:81: DANGER AHEAD: You have manually specified `updater` parameter. The `tree_method` parameter will be ignored. Incorrect sequence of updaters will produce undefined behavior. For common uses, we recommend using`tree_method` parameter instead.
✔ |  22       | update trees in an existing model [0.3 s]

Also there is a download message:

⠏ |   0       | Models from previous versions of XGBoost can be loadedtrying URL 'https://xgboost-ci-jenkins-artifacts.s3-us-west-2.amazonaws.com/xgboost_r_model_compatibility_test.zip'
Content type 'application/zip' length 63026 bytes (61 KB)
==================================================
downloaded 61 KB

It can be disabled by quiet = TRUE argument of download.file, (optionally followed with on.exit(unlink(zip_file, force = TRUE)) in order to avoid it showing in git status).

[vspinu]

I am a bit confused by this logic. So for ver >= 1.1.1.1, predict and run_booster_check no longer throw warnings when the model is loaded from RDS? Wouldn't you want to discourage it regardless and throw warning?

[hcho3]

That's correct. When RDS is used, XGBoost package does not actually know that RDS is used. So the warning is only generated when a material change in the model representation is detected. There was a major change in 1.1.1.1, and a minor one in 1.2.0.1. The major change is covered by the R warning ("The model had been generated by XGBoost version 1.0.0 or earlier and was loaded from a RDS file") and the minor change is covered by the C++ warning ("Attempted to load internal configuration for a model file that was generated by a previous version of XGBoost").

[hcho3]

The official documentation has a clear guidance against RDS: https://www.rdocumentation.org/packages/xgboost/versions/1.2.0.1/topics/a-compatibility-note-for-saveRDS-save

[vspinu]

Version is hardcoded here, so for new models this if-else won't be triggered at all?

[hcho3]

That's also correct, since no C++ warning is currently generated for the models that were generated by version 1.2.0.1.