Feature: add_predictor

[mmyrte]

Reintroduces the db$add_predictor interface that simultaneously writes to pred_meta_t and pred_data_t. Incidentally, the dev env was upgraded to R 4.6 and roxygen 8.0.0 introduced, which makes it look like this PR is huge, but actually isn't.

[Copilot]

Pull request overview

This PR reintroduces the db$add_predictor() convenience interface to write predictor metadata (pred_meta_t) and predictor values (pred_data_t) in one step, alongside an environment upgrade (R 4.6 + roxygen2 8.0.0) that regenerates documentation and workflows.

Changes:

• Add add_predictor() (and evoland_db$add_predictor()) to upsert predictor metadata + data using the active id_run.
• Tighten/coerce pred_meta_t types (incl. sources normalization) and update tinytests accordingly.
• Update CI workflows and regenerate Rd files due to roxygen changes.

Reviewed changes

Copilot reviewed 13 out of 22 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
rproject.toml	Bumps declared R version for the dev environment.
R/util_download.R	Clarifies that df_in can come from predictor/intervention metadata tables.
R/trans_models_t.R	Minor formatting/roxygen adjustments (comma cleanup, doc tweak).
R/pred_meta_t.R	Strengthens coercion for pred_meta_t fields and normalizes sources.
R/pred_data_t.R	Adds add_predictor() and adjusts printing/fill behavior for predictor columns.
R/evoland_db.R	Exposes add_predictor() as an R6 method on evoland_db.
R/evoland_db_views.R	Roxygen alias formatting cleanup.
man/util_terra.Rd	Roxygen-regenerated link target update for terra::SpatRaster.
man/util_download.Rd	Mirrors util_download.R doc change in generated Rd.
man/pred_meta_t.Rd	Mirrors pred_meta_t doc wording update in generated Rd.
man/pred_data_t.Rd	Documents newly added add_predictor() alias/usage.
man/parquet_db.Rd	Large roxygen formatting regeneration (method anchors/sections).
man/LearnerClassifGrrf.Rd	Roxygen formatting + link target regeneration.
man/evoland_db.Rd	Adds add_predictor() method to generated class docs + roxygen regeneration.
man/coords_t.Rd	Roxygen-regenerated link target update for terra::SpatExtent.
inst/tinytest/test_pred_meta_t.R	Extends pred_meta_t tests for missing sources behavior.
inst/tinytest/test_integ_trans_models_t.R	Updates plot-data assertions to match regenerated outputs.
inst/tinytest/test_db_evoland.R	Adds integration coverage for db$add_predictor() behavior.
DESCRIPTION	Updates roxygen config metadata to roxygen2 8.0.0.
.github/workflows/R-CMD-check.yaml	Adjusts matrix + cache paths; updates action versions (needs fixes).
.github/workflows/R-CMD-check-light.yml	Updates checkout action version (needs fixes).

Files not reviewed (8)

• man/LearnerClassifGrrf.Rd: Language not supported
• man/coords_t.Rd: Language not supported
• man/evoland_db.Rd: Language not supported
• man/parquet_db.Rd: Language not supported
• man/pred_data_t.Rd: Language not supported
• man/pred_meta_t.Rd: Language not supported
• man/util_download.Rd: Language not supported
• man/util_terra.Rd: Language not supported

Comments suppressed due to low confidence (1)

R/pred_data_t.R:235

• In set_pred_coltypes(), fill_value is parsed via type.convert(), but the assignment uses meta_row$fill_value (a character, because pred_meta_t$fill_value is cast to char). This can coerce numeric predictor columns to character when filling NAs, and it ignores the parsed fill_value value entirely. Use the parsed fill_value for assignment (and handle factors explicitly if you need level-safe filling).

    # if col is factor, fill_value being a character is safe
    # dt set() can add a new level if it's not already present
    fill_value <- meta_row$fill_value |> type.convert(as.is = TRUE)
    if (!is.na(fill_value)) {
      data.table::set(
        result,
        i = which(is.na(result[[col]])),
        j = col,
        value = meta_row$fill_value
      )

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.