Bump version to 0.1.21, add Copilot instructions, and update parquet test for key metadata handling

Author: ChristK

.github/copilot-instructions.md

@@ -0,0 +1,90 @@
+# CKutils Copilot Instructions
+
+## Project Overview
+
+CKutils is an R package providing high-performance utility functions for simulation modelling, with C++ backends via Rcpp. Key domains: data.table operations, statistical distributions (GAMLSS-compatible), and package management utilities.
+
+## Architecture
+
+- **R files** ([R/](R/)): Public API with roxygen2 documentation
+- **C++ files** ([src/](src/)): Performance-critical implementations using Rcpp
+- **Tests** ([inst/tinytest/](inst/tinytest/)): tinytest framework, validated against gamlss.dist
+
+### Key Modules
+| File | Purpose |
+|------|---------|
+| `R/rng_distr.R` | `fr*` random generation functions (BCPEo, BCT, NBI, SICHEL, etc.) |
+| `R/lookup_dt.R` | Fast key-based table lookups with `lookup_dt()`, `absorb_dt()` |
+| `R/dt_ops.R` | data.table utilities: `clone_dt()`, `del_dt_rows()` |
+| `src/distr_*.cpp` | Vectorised distribution functions (`fd*`, `fp*`, `fq*` for PDF/CDF/quantile) |
+| `src/lookup_dt.cpp` | C++ backend for `lookup_dt()` using data.table API |
+
+## Development Workflow
+
+### Pre-push validation (REQUIRED)
+```bash
+./check-before-push.sh
+```
+This runs: roxygen2 → R CMD build → R CMD INSTALL → R CMD check --as-cran → tinytest suite
+
+### Quick iteration during development
+```r
+# Regenerate documentation and install
+roxygen2::roxygenise()
+# Run tests
+tinytest::test_package("CKutils")
+```
+
+### Build requirements
+- C++17 (set in `src/Makevars`)
+- Dependencies: `data.table >= 1.18.0`, `Rcpp`, `dqrng`, `arrow >= 22.0.0`
+- Tests require: `gamlss.dist` for distribution validation
+
+## Coding Conventions
+
+### R Functions
+- Use **roxygen2** for all exports with `@export` tag
+- Document all parameters with types and defaults in `@param`
+- Include runnable `@examples` for every exported function
+- Prefix fast C++ wrappers with `f` (e.g., `fdBCPEo` vs gamlss.dist's `dBCPEo`)
+
+### Distribution Functions Pattern
+Each distribution (e.g., BCPEo) follows naming convention:
+- `fd*` - density (PDF), accepts `log_` parameter
+- `fp*` - probability (CDF), accepts `lower_tail` parameter
+- `fq*` - quantile (inverse CDF)
+- `fr*` - random generation, uses `dqrng::dqrunif` for high-quality RNG
+
+### C++ Code
+- Use `recycling_helpers.h` for parameter recycling (all vectors same length)
+- Include SIMD hints where applicable: `#pragma GCC ivdep`
+- Validate inputs early with `Rcpp::stop()` for clear error messages
+- For data.table integration, include `<datatableAPI.h>`
+
+### data.table Operations
+- Functions modify tables **in place** by default (data.table semantics)
+- Use `copy()` explicitly when non-destructive behaviour is needed
+- Factor columns in `lookup_dt()` must have consecutive integer levels starting from 1
+
+## Testing Pattern
+
+Tests in `inst/tinytest/test-*.R` follow this structure:
+```r
+# Skip if optional dependency unavailable
+if (!requireNamespace("gamlss.dist", quietly = TRUE)) {
+  exit_file("gamlss.dist not available")
+}
+
+# Generate test data with seed for reproducibility
+data <- generate_test_data(n = 100, seed = 123)
+
+# Compare against reference implementation
+expect_equal(fdBCPEo(...), gamlss.dist::dBCPEo(...), info = "PDF matches reference")
+```
+
+## Common Gotchas
+
+1. **Lookup tables**: Keys must be factors OR consecutive 1-based integers; use `is_valid_lookup_tbl()` to validate
+2. **Random number generation**: Always use `dqrng::dqrunif` instead of base R's `runif` for reproducibility
+3. **Windows compatibility**: C++ code must handle in-place modifications carefully; prefer `clone()` in Rcpp
+4. **R CMD check**: Avoid non-portable compiler flags; `check-before-push.sh` sets safe defaults

DESCRIPTION

@@ -1,8 +1,8 @@
 Package: CKutils
 Type: Package
 Title: Some Utility Functions I Use Regularly
-Version: 0.1.20
-Date: 2026-01-19
+Version: 0.1.21
+Date: 2026-01-20
 Authors@R: c(
     person("Chris", "Kypridemos", email = "christodoulosk@gmail.com", role = c("aut", "cre"), comment = c(ORCID = "0000-0002-0746-9229")),
     person("Max", "Birkett", email = "pp0u8134@liverpool.ac.uk", role = "ctb", comment = c(ORCID = "0000-0002-6076-6820")),

inst/tinytest/test-misc_functions.R

@@ -1528,21 +1528,12 @@ if (requireNamespace("arrow", quietly = TRUE)) {
 
   # Test 19: Create parquet with metadata and verify keys are restored
   temp_dir <- tempdir()
+
   temp_parquet_with_keys <- file.path(temp_dir, "test_with_keys.parquet")
   test_dt_keys <- data.table(id = 1:10, name = letters[1:10], value = rnorm(10))
   setkey(test_dt_keys, id)
-  # Write with key metadata
-  arrow::write_parquet(
-    test_dt_keys,
-    temp_parquet_with_keys,
-    properties = arrow::ParquetWriterProperties$create(
-      arrow::schema(test_dt_keys),
-      compression = "snappy"
-    ),
-    arrow_properties = arrow::ParquetArrowWriterProperties$create(
-      store_schema = TRUE
-    )
-  )
+  # Write initial parquet file
+  arrow::write_parquet(test_dt_keys, temp_parquet_with_keys, compression = "snappy")
   # Add metadata manually for keys
   tbl <- arrow::read_parquet(temp_parquet_with_keys, as_data_frame = FALSE)
   tbl_with_meta <- tbl$ReplaceSchemaMetadata(