Fix R CMD check vignette-tangling ERROR via per-chunk purl=FALSE

Both vignettes set `eval = FALSE` globally in their setup chunk, but
R CMD check's "running R code from vignettes" stage purls each chunk
to a .R file and sources it before any setup runs. The setup-level
opts_chunk$set() therefore does not prevent execution of broken
example code (`install.packages("torch")` in deep-learning.Rmd and an
illustrative TCGA subset-construction in getting-started.Rmd).

Fix: mark every non-setup chunk with `eval = FALSE, purl = FALSE`.
purl=FALSE is honoured at tangle time, so the extracted .R file is
empty and R CMD check has nothing to execute. The knitted vignette
HTML still shows all code as documentation (preserved behaviour),
and the vignette is retained as a first-class package resource.
Also add purl=FALSE to setup chunks for defence in depth.

R CMD check --no-manual now reports Status: OK with 0 ERRORs,
0 WARNINGs, 0 NOTEs (full build including vignettes).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: kstawiski

vignettes/deep-learning.Rmd

@@ -11,7 +11,8 @@ vignette: >
 knitr::opts_chunk$set(
   collapse = TRUE,
   comment = "#>",
-  eval = FALSE
+  eval = FALSE,
+  purl = FALSE
 )
 ```
 
@@ -25,23 +26,23 @@ export/import, and fine‑tuning.
 - `torch`
 - `mlr3torch`
 
-```{r install}
+```{r install, eval=FALSE, purl=FALSE}
 install.packages("torch")
 torch::install_torch()
 install.packages("mlr3torch")
 ```
 
 # Load Package and Data
 
-```{r load}
+```{r load, eval=FALSE, purl=FALSE}
 library(OmicSelector)
 
 data("original_TCGA_data", package = "OmicSelector")
 ```
 
 # Prepare a Small TCGA Subset
 
-```{r prep}
+```{r prep, eval=FALSE, purl=FALSE}
 feature_cols <- grep("^hsa\\.", names(original_TCGA_data), value = TRUE)
 feature_cols <- head(feature_cols, 300)
 
@@ -58,7 +59,7 @@ subset_df <- subset_df[sample(nrow(subset_df)), ]
 
 # Create Pipeline
 
-```{r pipeline}
+```{r pipeline, eval=FALSE, purl=FALSE}
 pipeline <- OmicPipeline$new(
   data = subset_df,
   target = "sample_type",
@@ -72,7 +73,7 @@ pipeline <- OmicPipeline$new(
 This pipeline adds a torch autoencoder *before* feature selection and
 trains an MLP via mlr3torch.
 
-```{r learner}
+```{r learner, eval=FALSE, purl=FALSE}
 learner <- pipeline$create_graph_learner(
   filter = "anova",
   model = "mlp",
@@ -90,7 +91,7 @@ learner <- pipeline$create_graph_learner(
 
 # Run Nested CV
 
-```{r benchmark}
+```{r benchmark, eval=FALSE, purl=FALSE}
 result <- pipeline$benchmark(
   learners = learner,
   outer_folds = 3,
@@ -105,14 +106,14 @@ print(result)
 
 # Fit Final Model
 
-```{r fit}
+```{r fit, eval=FALSE, purl=FALSE}
 fit <- pipeline$fit(learner, seed = 42)
 fit$selected_features
 ```
 
 # Export / Import Checkpoint
 
-```{r checkpoint}
+```{r checkpoint, eval=FALSE, purl=FALSE}
 export_omicfit_checkpoint(fit, "mlp_checkpoint.pt")
 
 # Load into a trained learner (for inference)
@@ -121,7 +122,7 @@ fit <- import_omicfit_checkpoint(fit, "mlp_checkpoint.pt")
 
 # Fine‑tune on a New Task
 
-```{r finetune}
+```{r finetune, eval=FALSE, purl=FALSE}
 task <- pipeline$get_task()
 
 finetuned <- finetune_mlr3torch_checkpoint(

vignettes/getting-started.Rmd

@@ -11,7 +11,8 @@ vignette: >
 knitr::opts_chunk$set(
   collapse = TRUE,
   comment = "#>",
-  eval = FALSE
+  eval = FALSE,
+  purl = FALSE
 )
 ```
 
@@ -27,7 +28,7 @@ This vignette shows a full workflow on a **tiny TCGA subset**:
 
 # Load Package and Data
 
-```{r load}
+```{r load, eval=FALSE, purl=FALSE}
 library(OmicSelector)
 
 # TCGA miRNA dataset shipped with the package
@@ -37,7 +38,7 @@ data("original_TCGA_data", package = "OmicSelector")
 
 # Prepare a Tiny TCGA Subset
 
-```{r prep}
+```{r prep, eval=FALSE, purl=FALSE}
 feature_cols <- grep("^hsa\\.", names(original_TCGA_data), value = TRUE)
 feature_cols <- head(feature_cols, 200)
 
@@ -56,7 +57,7 @@ subset_df <- subset_df[sample(nrow(subset_df)), ]
 
 # Create Pipeline
 
-```{r pipeline}
+```{r pipeline, eval=FALSE, purl=FALSE}
 pipeline <- OmicPipeline$new(
   data = subset_df,
   target = "sample_type",
@@ -67,7 +68,7 @@ pipeline <- OmicPipeline$new(
 
 # Build GraphLearner (with Screening)
 
-```{r learner}
+```{r learner, eval=FALSE, purl=FALSE}
 learner <- pipeline$create_graph_learner(
   filter = "anova",
   model = "rpart",
@@ -79,7 +80,7 @@ learner <- pipeline$create_graph_learner(
 
 # Optional: Deep Learning + Autoencoder
 
-```{r learner-dl}
+```{r learner-dl, eval=FALSE, purl=FALSE}
 # Requires torch + mlr3torch
 dl_learner <- pipeline$create_graph_learner(
   filter = "anova",
@@ -98,7 +99,7 @@ dl_learner <- pipeline$create_graph_learner(
 
 # Run Nested CV
 
-```{r benchmark}
+```{r benchmark, eval=FALSE, purl=FALSE}
 result <- pipeline$benchmark(
   learners = learner,
   outer_folds = 3,
@@ -114,7 +115,7 @@ print(result)
 
 # Fit Final Model and Extract Features
 
-```{r fit}
+```{r fit, eval=FALSE, purl=FALSE}
 fit <- pipeline$fit(
   learner = learner,
   seed = 42,