Add ability for summary functions to refer to columns other than the output column

[travis-leith]

In my experience it is uncommon to want to have a summary row in a table that uses a univariate function for all columns (as is the case in gt now - v0.2.2).
More commonly I want to apply, for example, a sum to some columns (such as trade value) and a weighted average to others (such as average trade price. In order to do this I would need a way for those functions to reference columns other than that for which the output is for. Such as the summary for column x should be sum(x*y)/sum(y)

[rich-iannone]

This is a great idea, thanks for posting it here. Totally agree that summarizing is a bit simplistic right now.

Yes, I need this, for example, for calculating a rate. I have a column with "number of cases", and another column with rate, split by region (where rate is cases/population for a given region). To calculate an overall rate for my summary, I need (sum cases)/(sum population). Not possible to do it with just the vector of the individual column.

Thank you for your work on this.

[nicomunting]

I ran into the same limitation. It would be nice if this capability was added in v0.4.0.

As a workaround (or dirty hack) I just passed a vector with the weights directly to the weighted.mean function. This would not work for tables with grouped rows though.

df <- tibble::tribble(
  ~price, ~value,
  5, 4,
  2, 6,
  3, 1
)

df %>% 
  dplyr::mutate(value_pct = value / price) %>% 
  gt() %>% 
  summary_rows(fns = list(total = ~ sum(.)), columns = c(value, price)) %>% 
  summary_rows(
    fns = list(total = ~ weighted.mean(., w = df$price)), 
    columns = value_pct,
    formatter = fmt_percent, decimals = 0
  ) %>% 
  fmt_percent(columns = contains("pct"), decimals = 0)

By the way, is this not a duplicate of #383?