Conflict with furrr?

[sammo3182]

Hi Vincent,

I just encountered an error when using both furrr and modelsummary together. I illustrated this in the following example. I guess the problem was caused by some batched function in furrr overwriting the original ones in plain purrr embedded in modelsummary? When inputting each model separately rather than packed as a list into modelsummary, the function works fine.

library(ordinal)
library(modelsummary)

fm1 <- clm(rating ~ temp * contact, data = wine)
fm2 <- update(fm1, ~ . - temp:contact)

modelsummary(list(fm1, fm2))

	(1)	(2)
1	2	-1.411
	(0.545)	(0.517)
2	3	1.144
	(0.510)	(0.438)
3	4	3.377
	(0.638)	(0.598)
4	5	4.942
	(0.751)	(0.731)
tempwarm	2.321	2.503
	(0.701)	(0.529)
contactyes	1.347	1.528
	(0.660)	(0.477)
tempwarm × contactyes	0.360
	(0.924)
:———————-	——–:	——–:
Num.Obs.	72	72
AIC	186.8	185.0
BIC	202.8	198.6
RMSE	2.80	2.80

library(furrr)
#> Loading required package: future

plan(multisession, workers = parallel::detectCores() - 2)
modelsummary(list(fm1, fm2))
#> Error: `modelsummary could not extract the required information from a model of
#>   class "clm". The package tried a sequence of 2 helper functions to
#>   extract estimates:
#>   
#>   parameters::parameters(model)
#>   broom::tidy(model)
#>   
#>   To draw a table, one of these commands must return a `data.frame` with a
#>   column named "term". The `modelsummary` website explains how to
#>   summarize unsupported models or add support for new models yourself:
#>   https://vincentarelbundock.github.io/modelsummary/articles/modelsummary.html
#>   
#>   These errors messages were generated during extraction:
#>   
#>   `parameters::parameters(model)` did not return a valid data.frame.
#> `broom::tidy(model)` did not return a valid data.frame.

Here's my sessionInfo:

sessionInfo()
#> R Under development (unstable) (2023-04-20 r84291 ucrt)
#> Platform: x86_64-w64-mingw32/x64 (64-bit)
#> Running under: Windows 11 x64 (build 22621)
#> 
#> Matrix products: default
#> 
#> 
#> locale:
#> [1] LC_COLLATE=English_United States.utf8 
#> [2] LC_CTYPE=English_United States.utf8   
#> [3] LC_MONETARY=English_United States.utf8
#> [4] LC_NUMERIC=C                          
#> [5] LC_TIME=English_United States.utf8    
#> 
#> time zone: Asia/Shanghai
#> tzcode source: internal
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> loaded via a namespace (and not attached):
#>  [1] styler_1.10.1     digest_0.6.31     fastmap_1.1.1     xfun_0.39        
#>  [5] magrittr_2.0.3    glue_1.6.2        R.utils_2.12.2    knitr_1.43       
#>  [9] htmltools_0.5.5   rmarkdown_2.22    lifecycle_1.0.3   cli_3.6.1        
#> [13] R.methodsS3_1.8.2 vctrs_0.6.3       reprex_2.0.2      withr_2.5.0      
#> [17] compiler_4.4.0    R.oo_1.25.0       R.cache_0.16.0    purrr_1.0.1      
#> [21] rstudioapi_0.14   tools_4.4.0       evaluate_0.21     yaml_2.3.7       
#> [25] rlang_1.1.1       fs_1.6.2

[vincentarelbundock]

Have you tried disabling future, as described in the Parallel Computation section of the ?modelsummary documentation?

[vincentarelbundock]

For what it's worth, I have found parallel problems to be extremely difficult to diagnose, because they depend so much on people's machines, and I can't test that myself. I'd say that if you want to use modelsummary in parallel, it's probably best to use the mc.cores approach if it works on your computer.

[vincentarelbundock]

Thanks for the report and sorry for the delay. Version 1.4.1.9010 provides a fallback in cases where this error arises, we revert to sequential extraction rather than parallel. This may be a big longer, but at least it won't error anymore.