Generate a dataset config from a delivery file

Sniffs the delivery (sniff_dataset) and writes a fully-optioned, self-documenting YAML config: every config key appears, detected values set live, optional settings commented out with their defaults shown, each key documented with the same description the validator's vocabulary uses. Duplicate header names arrive renamed positionally (with # was "..." annotations and csv_skip: 1); fixed-width files get a character-position ruler comment above fwf_widths, and a packed file (no detectable boundaries) gets explicit TODO widths that validate_config refuses to run until filled in.

Usage

generate_dataset_config(path, config_dir = ".", dataset_name = NULL)

Arguments

path

Character. Path to the delivery file to sniff.

config_dir

Character. Directory to write <dataset_name>.yml into (created if absent). Defaults to ".".

dataset_name

Character or NULL. Config identity; defaults to the delivery filename without extension, sanitised to [A-Za-z0-9_-].

Value

Invisibly, the path of the written config file.

Details

Create-only. An existing config for the dataset is never overwritten – generation aborts with dqcheckr_config_exists and the file is untouched. Once generated, a config is owned by hand edits; to re-sniff a changed delivery, generate under a different name (or into a scratch directory) and diff. The write is crash-safe (emitted to a temp file, renamed into place once complete); see the Accepted design decisions section of the specification vignette for the residual concurrent-generation caveat.

Examples

tmp <- file.path(tempdir(), "gen-demo")
dir.create(tmp, showWarnings = FALSE)
f <- file.path(tmp, "orders.csv")
writeLines(c("id,amount", "A1,10", "A2,20"), f)
cfg <- generate_dataset_config(f, config_dir = tmp)
#> [dqcheckr] Config written: /tmp/RtmpXCCxT9/gen-demo/orders.yml
#>   Review the commented options, then check with validate_config("orders", config_dir = "/tmp/RtmpXCCxT9/gen-demo").
# A deployment needs the global config too: validate_config() (like
# run_dq_check()) requires dqcheckr.yml to be present.
generate_global_config(tmp)
#> [dqcheckr] Global config written: /tmp/RtmpXCCxT9/gen-demo/dqcheckr.yml
#>   Relative snapshot_db/report_output_dir resolve against the working directory -- run from the deployment root.
validate_config("orders", config_dir = tmp)$valid
#> [1] TRUE