@jangorecki can you think of any other types of validations we'd want to do upfront here? if so maybe best to have a validate_dt function we can build up starting from here.

looking for NULL dim failed test 1980, which has an array with length(dim(x)) == 1L that ends up being setkey'd (confirmed that in this case, setDT wipes out the dim). Played around with changing as.data.table.list to route this case through as.data.table but we'd have to route through as.data.table.array, which fails because there's a check that length(dim(x)) >= 3 there.

If we change this to route this case to as.data.table.matrix instead of erroring, the solution I had wound up with the wrong name:

y = as.array(1:5)
names(data.table(y))
# [1] 'x'

So I gave up on that route; the one here does fine. Just recording my thoughts for posterity.

Codecov Report

Merging #3770 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #3770      +/-   ##
==========================================
+ Coverage   99.41%   99.41%   +<.01%     
==========================================
  Files          71       71              
  Lines       13222    13225       +3     
==========================================
+ Hits        13145    13148       +3     
  Misses         77       77

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d75265f...94b0cc6. Read the comment docs.