DragenTool Object

Intermediate base class for all tidydragen tools. Inherits nemo::Tool and adds DRAGEN-specific shared logic e.g. the metrics parser/tidier used by every *_metrics.csv table. Tools like DragenCov inherit from DragenTool rather than nemo::Tool directly. The dragen-metrics ftype is registered here (parse via extra_ftypes(), tidy via the tidy_file override), so a plain metrics table needs no methods, just ftype: 'dragen-metrics' in its schema. A table needing a normalise/drop_constant still declares an explicit tidy_<tbl>() that delegates to tidy_metrics.

Details

DRAGEN *_metrics.csv files share a headerless 4/5-column shape: section, rg, variable, count[, pct]. rg (read-group / sample) is empty for most sections but populated for e.g. fastqc and per-sample variant sections. The value column (count) is read as character since some metrics carry string values (e.g. Ploidy estimation = "XX") and counts can exceed 32-bit range; each tidy column is coerced to its schema type after the wide pivot.

Super class

nemo::Tool -> DragenTool

Public fields

on_unmapped

(character(1))
Policy for metrics present in a *_metrics.csv file but absent from the schema. error refuses to drop data and aborts, so schema drift (e.g. a new DRAGEN version adds a metric) fails loudly rather than silently omitting a column. Set to warn to instead drop the unmapped metrics with a warning.

on_coerce_fail

(character(1))
Policy for a non-empty metric value that fails coercion to its schema numeric type (becomes NA). error aborts rather than losing the value; usually a wrong schema type (should be char) or an unexpected string. Set to warn to keep the NA with a warning. Genuine missings (NA/empty in the file) never trigger this.

on_unexpected_col

(character(1))
Policy for an id column named in tidy_metrics(drop_constant=) that was expected to be constant (and so dropped) but instead varies across rows. error aborts rather than letting a surprise extra column appear in the output table (schema drift into a data lake). Set to warn to keep the unexpected column with a warning. Tables that legitimately vary in section/rg opt out via drop_constant = character() and never trip this.

Examples

# Abstract base. Normally you use a subclass (e.g. DragenVar). The parser/tidier
# are private; here we just inspect the inherited surface + policy fields.
indir <- system.file("extdata/dragenmap", package = "tidydragen")
tool <- DragenTool$new(name = "dragenmap", pkg = "tidydragen", path = indir)
tool$list_files()      # inherited nemo::Tool file discovery
#> # A tibble: 6 × 9
#>   tool_parser       parser bname   size lastmodified        path  pattern prefix
#>   <chr>             <chr>  <chr> <fs::> <dttm>              <chr> <chr>   <chr> 
#> 1 dragenmap_metrics metri… samp… 27.03K 2026-09-15 12:58:09 /hom… "\\.ma… sampl…
#> 2 dragenmap_time    time   samp…   1.3K 2026-09-15 12:58:09 /hom… "\\.ti… sampl…
#> 3 dragenmap_fragle… fragl… samp… 19.39K 2026-09-15 12:58:09 /hom… "\\.fr… sampl…
#> 4 dragenmap_trimmer trimm… samp…  1.35K 2026-09-15 12:58:09 /hom… "\\.tr… sampl…
#> 5 dragenmap_umimain umima… samp…  5.22K 2026-09-15 12:58:09 /hom… "\\.um… sampl…
#> 6 dragenmap_gcmain  gcmain samp… 10.52K 2026-09-15 12:58:09 /hom… "\\.gc… sampl…
#> # ℹ 1 more variable: prefix_suffix <chr>
tool$on_unmapped       # "error": unmapped metrics abort rather than drop
#> [1] "error"
tool$on_coerce_fail    # "error": values that fail type coercion abort
#> [1] "error"

# flip a policy to be lenient (drop/keep-NA with a warning instead)
tool$on_unmapped <- "warn"