refactor puzzling loop in attributes.R

[maelle]

rigraph/R/attributes.R

Lines 557 to 564 in 36ad3c9

	for (i in seq_along(value)) {
	tmp <- value[[i]]
	length(tmp) <- 0
	length(tmp) <- length(vs)
	tmp[index] <- value[[i]]
	value[[i]] <- tmp
	}
	}

[maelle]

L559 can be deleted, the rest is more difficult.

[maelle]

value is a named list.

the code above is for the case where we don't have as many values as vertices. we create a list as long as the number of vertices. then the indices are used to fill the list elements. the other elements remain NULL.

surely there's an elegant way to do this?!

[maelle]

elegant and readable

[maelle]

first fill with NA, then make NA NULL? 🤔 no, they are NA in the current version.

[maelle]

a good thing is that the lines are covered by tests

[maelle]

first make the loop easier then use purrr::reduce?

[maelle]

      tmp <- rep(NA, length(vs))
      tmp[index] <- value[[i]]
      value[[i]] <- tmp

[maelle]

oooh for value a vector, we get NA as fillers but for value a named list, we get NULL as fillers.

[clpippel]

My attempt to isolate the essence of the coding:

vs           <- c(1,2,3,4)
index        <- c(1,4)
value        <- list(list(a= 10, d = 14))
names(value) <- "color"

i            <- 1
tmp          <- value[[i]]         # Copy structure of value.
length(tmp)  <- 0                  # Set list to empty, while removing names.
length(tmp)  <- length(vs)         # Extend to all vertices with NULL.
tmp[index]   <- value[[i]]         # Copy indexed elements,
value[[1]]   <- tmp                # into value.
names(value[[i]])

The effect of length(tmp) <- 0 is to remove the names initially present in value.
When leaving out the statement the names in value in this example are: [1] "a" "d" "" "" .

In this example using the old logic the expression
>value[[1]]["a"] will give

$<NA>
NULL

thereby blocking indexing by name. In the new logic the result will be:

$a
[1] 10

Removing the names could be a matter of defensive coding.
All the code need is some explanation.

[clpippel]

Create an empty named list.

L558, L559 are equivalent to:
tmp <- setNames(list(), character(0)) # Easy to read.
  or alternatively
tmp <- list(dummy = 1)[NULL] # Faster.
  or
tmp <- value[[i]][NULL] # As current coding.

To replace lines L558, L559, L560 by a single line:
tmp <- 'length<-'(list(dummy = 1)[NULL], length(vs))