as_forecast_quantile() renames the specified quantile level column in its input data to "quantile_level" (in place)

[dylanhmorris]

This does not occur for the specified "forecast" and "observed" columns. This means that calling as_forecast_quantile() multiple times on the same input data.table can succeed the first time and fail subsequently.

Reprex

library(scoringutils)
library(data.table)

quantile_data <- data.table(my_quantile = c(0.25, 0.5),
                            forecast_value = c(1, 2),
                            observed_value = c(5, 5))

quantile_forecast <- quantile_data |>
    as_forecast_quantile(
        predicted = "forecast_value",
        observed = "observed_value",
        quantile_level = "my_quantile")

## the quantile_forecast output is as expected
quantile_forecast
#> Forecast type: quantile
#> Forecast unit:
#> 
#>    quantile_level predicted observed
#>             <num>     <num>    <num>
#> 1:           0.25         1        5
#> 2:           0.50         2        5

## the "my_quantile" column in the input quantile_data
## been renamed "quantile_level"; the "forecast_value"
## and "observed_value" columns retain
## their original names
quantile_data
#>    quantile_level forecast_value observed_value
#>             <num>          <num>          <num>
#> 1:           0.25              1              5
#> 2:           0.50              2              5

## this now raises an error, because the column name has changed
## from "my_quantile" to "quantile_level"
quantile_data |>
    as_forecast_quantile(
        predicted = "forecast_value",
        observed = "observed_value",
        quantile_level = "my_quantile")
#> Error in as_forecast_quantile.default(quantile_data, predicted = "forecast_value", : Assertion on 'quantile_level' failed: Must be a subset of {'quantile_level','forecast_value','observed_value'}, but has additional elements {'my_quantile'}.

^{Created on 2025-01-03 with reprex v2.1.1}

Info

scoringutils version 2.0.0

Can provide full sessionInfo() output if needed.

[dylanhmorris]

Potentially related to #972