Unclear error when storing data.frame as edge attributes

[dgkf]

What happens, and what did you expect instead?

Not sure if these are bugs, or just user experiences that result in unclear behaviors. Reporting primarily for visibility.

When a user accidentally uses a data.frame for edge attributes, many operations continue to work fine but errors will arise when subsetting when indices can't be recycled. Arises from this line, which iterates over (an assumed list) of attributes, subsetting with the indices.

rigraph/R/iterators.R

Line 964 in 2a84bc2

for (i in seq_along(attrs)) attrs[[i]] <- attrs[[i]][xvec]

When the attributes are instead a data.frame, assignment with a value that can't be recycled to the data frame row count will throw an error.

When trying to create edge attributes with a data.frame, it would be very helpful if it was run through as.list automatically to avoid these pitfalls or if an error was raised to warn that this is unsupported.

To reproduce

library(igraph)  # v2.1.4

g <- make_tree(10, 3)
n_edges <- length(E(g))

# whoops! igraph does not support data.frames
igraph::edge.attributes(g) <- head(mtcars, n_edges)

# but they seem to work fine in many cases, easy to miss wrong behavior
E(g)[cyl == 4]
#> + 3/9 edges from 2eeb173:
#> [1] 1-> 4 3-> 9 3->10

E(g)[1:3]
#> + 3/9 edges from 2eeb173:
#> [1] 1->2 1->3 1->4

E(g)[c(1, 1)]
#> + 2/9 edges from 2eeb173:
#> Error in `[[<-.data.frame`(`*tmp*`, i, value = c(21, 21)) :
#>  replacement has 2 rows, data has 9

System information

R version 4.4.1 (2024-06-14)
Platform: aarch64-unknown-linux-gnu
Running under: Ubuntu 22.04.4 LTS (via rocker/r-ver image on podman)

Matrix products: default
BLAS:   /usr/lib/aarch64-linux-gnu/openblas-pthread/libblas.so.3 
LAPACK: /usr/lib/aarch64-linux-gnu/openblas-pthread/libopenblasp-r0.3.20.so;  LAPACK version 3.10.0

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

time zone: Etc/UTC
tzcode source: system (glibc)

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] igraph_2.1.4

loaded via a namespace (and not attached):
[1] compiler_4.4.1  magrittr_2.0.3  cli_3.6.3       lifecycle_1.0.4
[5] pkgconfig_2.0.3 rlang_1.1.5    

[schochastics]

Reprex with working workaround

library(igraph)
g <- h <- make_tree(10, 3)
n_edges <- ecount(g)
igraph::edge.attributes(g) <- head(mtcars, n_edges)
igraph::edge.attributes(h) <- as.list(head(mtcars, n_edges))
E(g)[c(1,2)]
#> + 2/9 edges from 9e2c041:
#> Error in `[[<-.data.frame`(`*tmp*`, i, value = c(21, 21)): replacement has 2 rows, data has 9
E(h)[c(1,2)]
#> + 2/9 edges from 9e2c041:
#> [1] 1->2 1->3

The issue is that data frames slip through this test

rigraph/R/attributes.R

Lines 1187 to 1201 in 22cd83e

	assert_named_list <- function(value) {
	error_msg <- "{.arg value} must be a named list with unique names"
	if (!is.list(value)) {
	rlang::abort(error_msg)
	}
	if (length(value) == 0) {
	return()
	}
	if (!rlang::is_named(value) || anyDuplicated(names(value)) > 0) {
	rlang::abort(error_msg)
	}
	}

(This test also happens for graph.attributes and vertex.attributes)

the data frame is then handled here

rigraph/R/attributes.R

Line 768 in 22cd83e

.Call(R_igraph_mybracket2_set, graph, igraph_t_idx_attr, igraph_attr_idx_edge, value)

It is easy to patch this behaviour (either forbid data.frames or convert them to lists) but I also would like to understand what happens internally.
@szhorvat can you immediately tell what internally goes wrong? I can dig a bit more on the R side