patchwork plots cause internal data to be dumped into chunk outputs

[kevinushey]

System details

RStudio Edition : 2023.08.0-daily+60
RStudio Version : Desktop [Open Source]
OS Version      : macOS Ventura 13.5
R Version       : R version 4.3.1 (2023-06-16)

Steps to reproduce the problem

From thomasp85/patchwork#231.

library(ggplot2)
library(patchwork)
gg1 <- ggplot(mtcars, aes(mpg, cyl)) + 
  geom_point()

gg2 <- ggplot(mtcars, aes(hp, disp)) + 
  geom_point()

foo <- gg1 + gg2 + plot_layout(ncol = 1)

Describe the problem in detail

A data.frame is printed, which is not expected.

Describe the behavior you expected

There shouldn't be any chunk outputs after executing this code.

Session Info

Tracing print(rInfo) on entry 
[1] "hi"
R version 4.3.1 (2023-06-16)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Ventura 13.5

Matrix products: default
BLAS:   /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.11.0

locale:
Tracing print(strsplit(x$locale, ";", fixed = TRUE)[[1]], quote = FALSE,  .... on entry 
[1] "hi"
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

time zone: America/Los_Angeles
tzcode source: internal

attached base packages:
Tracing print(x$basePkgs, quote = FALSE, ...) on entry 
[1] "hi"
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
Tracing print(mkLabel(x, "otherPkgs"), quote = FALSE, ...) on entry 
[1] "hi"
[1] patchwork_1.1.2 ggplot2_3.4.2  

loaded via a namespace (and not attached):
Tracing print(mkLabel(x, "loadedOnly"), quote = FALSE, ...) on entry 
[1] "hi"
 [1] vctrs_0.6.3       cli_3.6.1         knitr_1.43        rlang_1.1.1.9000 
 [5] xfun_0.39         DBI_1.1.3         generics_0.1.3    textshaping_0.3.6
 [9] labeling_0.4.2    glue_1.6.2        colorspace_2.1-0  htmltools_0.5.5  
[13] ragg_1.2.5        scales_1.2.1      fansi_1.0.4       rmarkdown_2.23   
[17] grid_4.3.1        evaluate_0.21     munsell_0.5.0     tibble_3.2.1     
[21] fastmap_1.1.1     yaml_2.3.7        lifecycle_1.0.3   compiler_4.3.1   
[25] dplyr_1.1.2       pkgconfig_2.0.3   rstudioapi_0.15.0 farver_2.1.1     
[29] systemfonts_1.0.4 digest_0.6.33     R6_2.5.1          tidyselect_1.2.0 
[33] utf8_1.2.3        pillar_1.9.0      magrittr_2.0.3    withr_2.5.0      
[37] tools_4.3.1       gtable_0.3.3     

[kevinushey]

I've taken a look, and it looks like a bug in RStudio. To explain the behavior:

1. When executing a Notebook chunk, RStudio inserts its own print() overrides, so that it can detect and capture attempts to print R objects. This is what allows RStudio to capture e.g. data.frame print-outs in a clean way.
2. RStudio also tries to detect changes to objects in the Environment pane as code chunks are executed. When new objects are created, or changed, we try to introspect those new objects to learn a bit more about those objects.
3. To that end, when inspecting the foo object, we call str(foo) -- but this involves calling print() when examining some of the internal data in foo.
4. Unfortunately, because the print() overrides in (1) are still active, we end up inadvertently capturing that output into the Notebook chunk.

I think we can fix this on the RStudio side by making sure we deactivate our print overrides when updating the Environment pane.

[kevinushey]

rstudio/src/cpp/session/modules/NotebookData.R

Lines 16 to 60 in f369336

	.rs.addFunction("dataCaptureOverrides", function(outputFolder, libraryFolder)
	{
	defaultOverride <- function(x, options) list(x = x, options = options, className = class(x), nRow = .rs.scalar(nrow(x)), nCol = .rs.scalar(ncol(x)))
	c(
	"print.data.frame" = defaultOverride,
	"print.tbl_df" = defaultOverride,
	"print.paged_df" = defaultOverride,
	"print.grouped_df" = defaultOverride,
	"print.rowwise_df" = defaultOverride,
	"print.tbl_sql" = defaultOverride,
	"print.data.table" = function(x, options) {
	shouldPrintTable <- TRUE
	if ("data.table" %in% loadedNamespaces() &&
	exists("shouldPrint", envir = asNamespace("data.table")))
	{
	shouldPrint <- get("shouldPrint", envir = asNamespace("data.table"))
	shouldPrintTable <- tryCatch(
	shouldPrint(x) || !inherits(x, "data.table"),
	error = function(e) TRUE
	)
	}
	if (!shouldPrintTable) {
	return(NULL)
	}
	list(x = x, options = options, className = class(x), nRow = .rs.scalar(nrow(x)), nCol = .rs.scalar(ncol(x)))
	},
	"print.tbl_lazy" = function(x, options) {
	tblLazyData <- lapply(dplyr::tbl_vars(x), function(e) character(0))
	names(tblLazyData) <- dplyr::tbl_vars(x)
	lazyFrame <- do.call("data.frame", tblLazyData)
	list(
	x = lazyFrame,
	options = options,
	className = class(x),
	nRow = "?",
	nCol = "?"
	)
	}
	)
	})