Access a column programmatically on a table expression

Apologies if this shouldn’t be an issue, but I have been searching everywhere and haven’t been able to understand what can I do with table expressions. Is this documented?

I want to derive columns programmatically. I have a dataset with a bunch of dates per each row, and I want to compute their relative position compared to a reference one. I know I can do:

dt.derive({
  date1_rel: d => d.date1 - d.date_ref,
  date2_rel: d => d.date2 - d.date_ref,
 ...
})

But I don’t want to write that by hand. I tried then these cells on Observable:

columnFechas = [
  "fis",
  "fecha_de_muerte",
  "fecha_reporte_web"
  "fecha_recuperado",
  "fecha_de_notificaci_n",
  "fecha_diagnostico"
]

deriveFechas = Object.fromEntries(
  columnFechas.map(f => [
    `${f}_res`,
    d =>  d[f] - d.fecha_reporte_web 
  ])
)

dt.derive(deriveFechas)

But It throws an error saying Invalid column "f": "d[f]". I tried debugging the code and noticed that the function is a table expression, so for a reason I don’t understand I cannot even console.log the attribute d. How do I get the column that is in the variable f in a table expression. Documenting this could be important whenever you have column names with spaces

Here is a notebook replicating the issue

https://observablehq.com/d/2783d075d8079564#fechas

Seems related to #5, is there a way of doing op.get(column, d)?

[arvind]

A question I continue to ask myself despite years of working with Jeff (perhaps with even more question marks now that I’m a faculty member as well 😃

[john-guerra]

BTW the new documentation for expression looks amazing! How can you do this so quickly?!