df <- data.frame(cals = clock::year_quarter_day(year = 2000:2001, quarter = 2))
DT::datatable(df)
#> Error:
#> ! Names must be a character vector.
Ultimately I think this is an htmlwidgets issue.
I followed it all the way down to htmlwidgets:::shouldEval() which recurses over the input if is.list() is TRUE on it. shouldEval() is first given something like list(data = df), so it seems like it is reasonable to do 1 layer of recursion on the elements of that list, but then it does another layer of recursion because df is a data frame and is.list() returns TRUE for data frames. It actually then does another layer of recursion because df$cals is clock's year-quarter-day class built on vctrs::new_rcrd(), which is also technically built on a list structure, even though it isn't a list.
One solution for this that we typically use is to replace is.list(x) with inherits(x, "list"). This ensures that you only recurse over things that are either:
Bare lists
Things that have explicitly declared that they are lists with structure(x, class = c("my_subclass", "list"))
I received this bug report https://github.com/r-lib/clock/issues/335
Ultimately I think this is an htmlwidgets issue.
I followed it all the way down to
htmlwidgets:::shouldEval()which recurses over the input ifis.list()isTRUEon it.shouldEval()is first given something likelist(data = df), so it seems like it is reasonable to do 1 layer of recursion on the elements of that list, but then it does another layer of recursion becausedfis a data frame andis.list()returnsTRUEfor data frames. It actually then does another layer of recursion becausedf$calsis clock's year-quarter-day class built onvctrs::new_rcrd(), which is also technically built on a list structure, even though it isn't a list.One solution for this that we typically use is to replace
is.list(x)withinherits(x, "list"). This ensures that you only recurse over things that are either:structure(x, class = c("my_subclass", "list"))I feel like that may work here.