Reads a column name written by [csfmt_var] back into its parts. It strips the trailing coordinates, then a role, then a `_vs_<denom>` segment, and whatever is left is the measure.
Value
Named list with the components that were present (e.g. `measure`, `role`, `q`, `denom`, `per`).
Where it does not invert csfmt_var
The parse is a right-to-left strip against a fixed role vocabulary, so it cannot tell which of several role-looking segments was the role. On the package's own rate name it gets the denominator wrong:
csfmt_var("numerator_nowcasted", denom = "denominator_nowcasted", per = 100)
#> "numerator_nowcasted_vs_denominator_nowcasted_pr100"
csfmt_parse("numerator_nowcasted_vs_denominator_nowcasted_pr100")$denom
#> "denominator" # the denominator's own "_nowcasted" was eaten as the roleTreat it as reliable for a single-role name such as `numerator_nowcasted_q50x0`. Check the result whenever the measure or the denominator itself ends in a role word.
See also
[csfmt_var] writes these names.
vignette("pipeline", package = "csalert"), whose closing section
parses a collapsed median column with this function.
Other naming grammar functions:
csfmt_interpret(),
csfmt_var(),
q_label(),
q_value()
Examples
csfmt_parse("numerator_nowcasted_q50x0")
#> $measure
#> [1] "numerator"
#>
#> $role
#> [1] "nowcasted"
#>
#> $q
#> [1] 0.5
#>
# the documented limit: a denominator that itself ends in a role word is
# truncated, because the role is stripped before the _vs_ segment is read
csfmt_parse("numerator_nowcasted_vs_denominator_nowcasted_pr100")
#> $measure
#> [1] "numerator_nowcasted"
#>
#> $denom
#> [1] "denominator"
#>
#> $role
#> [1] "nowcasted"
#>
#> $per
#> [1] 100
#>