New `rd_section_minidesc` error with roxygen2 7.2.1

[mjsteinbaugh]

With the roxygen2 7.2.1 update, I'm now seeing this error in one of my packages, and I'm having trouble locating the source:

Error in rd_section_minidesc(name = object_name(block$object), desc = tag$val$description,  :
  is_string(name) is not TRUE
Calls: <Anonymous> ... <Anonymous> -> inherits -> rd_section_minidesc -> stopifnot
Backtrace:
     ▆
  1. └─devtools::document()
  2.   └─roxygen2::roxygenise(pkg$path, roclets)
  3.     └─base::lapply(...)
  4.       ├─roxygen2 (local) FUN(X[[i]], ...)
  5.       └─roxygen2:::roclet_process.roclet_rd(X[[i]], ...)
  6.         ├─roxygen2:::block_to_rd(block, base_path, env)
  7.         └─roxygen2:::block_to_rd.roxy_block(block, base_path, env)
  8.           └─roxygen2:::topic_add_describe_in(rd, block, env)
  9.             ├─topic$add(...)
 10.             │ └─base::inherits(x, "RoxyTopic")
 11.             └─roxygen2:::rd_section_minidesc(...)
 12.               └─base::stopifnot(is_string(name))

Any ideas on how to improve the stack trace here? Would it be possible to modify the error handling to return which file is problematic?

Best,
Mike

[gaborcsardi]

Use options(error = recover) and then try to find more info interactively in the roxygen2 frames.

[mjsteinbaugh]

Cool thanks I'll try that!

[VLucet]

I'm running into the same issue.

I think it is related to a @describeIn tag. Using @gaborcsardi's suggestion above I think I tracked it down to that line:

roxygen2/R/rd-describe-in.R

Line 29 in 07e4d1d

name = object_name(block$object),

This returns NULL instead of a string, and roxygen2:::rd_section_minidesc(...) chokes expecting is_string(name) to be true. In this case I would assume that object_name() should return the equivalent of block$object$topic, the topic in which the block should be described.

For context, below is the code that lead to the issue. For that block the output of block$object$topic is "sspm_discrete_boundary-class".

See code

# -------------------------------------------------------------------------

#' sspm boundary structure
#'
#' One of the first step in the `sspm` workflow is to create one or more
#' object(s) of class `sspm_boundary` from an `sf` object.
#'
#' @slot boundaries **\[sf\]** Spatial boundaries (polygons).
#' @slot boundary **\[character\]** The column of `data` that represents the
#'     spatial boundaries.
#' @slot boundary_area **\[character\]** The column of `data` that represents the
#'     area of spatial boundaries.
#' @slot method **\[[discretization_method][discretization_method-class]\]**
#'     *(if discrete)* discretization method used.
#' @slot patches **\[sf\]** *(if discrete)* Patches resulting from
#'     discretization.
#' @slot points **\[sf or NULL\]** *(if discrete)* Sample points used for
#'     discretization.
#' @slot patches_area **\[character\]** The column of `data` that represents the
#'     area of patches.
#'
#' @name sspm_boundary-class
#' @rdname sspm_boundary-class
#'
setClass("sspm_boundary",
         slots = list(boundaries = "sf",
                      boundary = "character",
                      boundary_area = "character")
)

#' @describeIn sspm_boundary-class sspm_discrete_boundary
setClass("sspm_discrete_boundary",
         slots = list(method = "discretization_method",
                      patches_area = "character",
                      patches = "sf",
                      points = "ANY"),
         contains = "sspm_boundary"
)

[ralmond]

I'm running into this same issue. It is really hard to track, especially as I'm not even given a file name, so I don't know where to start. There needs to be better handling of context information when running into even an unexpected error.

[hadley]

This should be better in the dev version, but no one provided a reprex, so I can't be sure.