Revdep issues for 0.7.0

[DavisVaughan]

obj_is_list() now rejects list matrices and list arrays

Packages using dplyr::bind_rows() with list matrices or list arrays now fail because obj_check_list() (called via list_flatten()) is stricter. These packages need to convert their list matrices/arrays to plain lists before passing to bind_rows().

• feisr Remove dim before calling bind_rows() ruettenauer/feisr#9
  • Uses rbind() to create a list matrix, then passes to dplyr::bind_rows().
• ggstats Remove dim attribute before calling bind_rows() larmarange/ggstats#109
  • Creates a list 1D array that is passed to dplyr::bind_rows().
• merTools Drop dim before calling bind_rows() jknowles/merTools#133
  • Uses rbind() to create a list matrix, then passes to dplyr::bind_rows().
• rainette Remove dim before putting the list-col in a tibble juba/rainette#39
  • .data$clusters is now a list, causing "invalid subscript type 'list'" errors.

purrr's list_flatten() function now rejects array inputs

• purrr Clear dim before calling list_flatten() tidyverse/purrr#1239
  • array_branch() fails because apply() returns an array which list_flatten() now rejects.
• diceR
  • Uses purrr's list_flatten() on an array-like structure.
  • Fixed by fixing purrr

new_list_of() now calls vec_ptype() on ptype

• riskmetric Issue with names() on empty pkg_ref pharmaR/riskmetric#392
  • Custom names.pkg_ref() method is called unexpectedly during vec_ptype() on an empty ptype.

list_unchop() stricter type coercion

list_unchop() is now stricter about type coercion. Code that previously allowed mixing types may now fail with incompatible type errors.

• orderly Revdep failure with dev vctrs mrc-ide/orderly#250
  • Test fails: Can't convert <double> to <character> in list_unchop().
  • Overly strict test

List-of assignment with NULL behavior changed

• blob Expand on NULL subassignment tests related to vctrs changes tidyverse/blob#65

Error message text changed

Error messages now include additional information (FAQ references). Snapshot tests that match exact error text will fail.

• testthat Use expect_snapshot() for vctrs error testthat#2301
  • Snapshot test for expect_vector() error message now includes FAQ reference.

vec_cast() and type coercion changes

• wk vctrs revdep failure: vec_cast.wk_rct.default() and vec_cast.wk_crc.default() seem incorrect? paleolimbot/wk#230

No more partial_factor

• tibble vctrs 0.7.0 is coming Jan 15 tidyverse/tibble#1686
  • Vignette rebuilding failed (no detailed error output provided).

False alarms

• rlang
  • Tests for expr_deparse() and as_label() expect matrix/array dimensions in output.
  • False alarm, circular issue with vctrs requiring dev rlang to work, but CRAN rlang is being run here. This all prevents pillar from loading because it can't meet vctrs's dependency requirements for rlang, and tests fail. Should resolve when rlang is on CRAN.
• klassR
  • External API error: HTTP 404 from data.ssb.no service.
• crmPack Upcoming vctrs 0.7.0 release openpharma/crmPack#920
  • Matrix dimension error: "non-conformable arguments" in simulation code.
  • Unsure if false positive
• myTAI Upcoming vctrs 0.7.0 release drostlab/myTAI#78
  • Vignette rebuilding failed (no detailed error output provided).
  • There is a stack imbalance issue, and they have C code, so I'm wondering if it may be a random failure issue on their end
• svytest
  • Vignette building process was killed (likely timeout or memory issue).
  • Can't reproduce locally, it passes with dev and cran vctrs