Question about knitting vs. running chunks ("Run All")

[pat-s]

When running all chunks of my .Rmd file ("Run All"), here() is set to the R-project root dir. All paths are relative to this directory. Let's call it "main-dir".

The .Rmd file is placed one dir below the project root, let's say in "main-dir/sub-dir". When I knit the .Rmd file, I get .local object errors because knitr uses "sub-dir" as root dir when knitting.

Currently my workaround is to use a second R-project in "sub-dir" to ensure that both knitting and running all chunks works with the same relativ paths (and that my folder structure does not get messy).

I wonder what I'm doing wrong here so that here() is not adjusting paths correctly in my case. I initially expected that here() is setting the working directory to the project root when knitting. Did I get this wrong? 😅

[krlmlr]

Thanks. Could you please point me to your project, or attach an archive?

[pat-s]

In this example "Run All" chunks works. Knitting does not.

When changing here("spring-school-hsdar/data") to here("data"), i.e. making path relatives to the .Rmd file, knitting works but "Run All" does not work anymore (because paths to .Rproj files are wrong).

I also played around with knitr::opts_knit$set(root.dir = here()) but did not succeed. Even setting knitr::opts_knit$set(root.dir = "/Users/pjs/Desktop/test") (i.e. to the fixed path of the .Rproj dir) does not work.

You will need devtools::install_github("humburg/reportmd") for the example.

(Maybe its also a simple knitr problem and does not apply to here)

Archive.zip

[krlmlr]

Thanks. I'm seeing correct behavior with a simple document that is rendered with rmarkdown::html_document, but I can't get your example to run. I've installed the reportMD package, but knitting complains that a package "prompt" is missing, which I can't find.

Here's the minimal file I'm using in my tests:

---
title: "Part 1: Data investigation & preprocessing"
author: "Patrick Schratz"
date: "`r format(Sys.time(), '%a %d %b %Y')`"
output:
  reportMD::multi_document:
    toc: true
    fig_caption: true
    highlight: dracula 
    css: R.css
    use_namespace: false
params:
  version: !r 
---

```{r loadPackages, include=FALSE, cache=FALSE}
## load additional packages in this chunk
library(here)
```

```{r}
here()
```

Can you confirm that it works correctly with the following file:

---
title: "Part 1: Data investigation & preprocessing"
author: "Patrick Schratz"
date: "`r format(Sys.time(), '%a %d %b %Y')`"
output:
  rmarkdown::html_document
params:
  version: !r 
---

```{r loadPackages, include=FALSE, cache=FALSE}
## load additional packages in this chunk
library(here)
```

```{r}
here()
```

[pat-s]

Thanks for looking into it.
It should work if you delete 01_preprocessing_cache- there are some leftovers from my run in it. prompt can be found on Github.

I get the following error with here("spring-school-hsdar/data") using html_document:

Quitting from lines 76-78 (01_preprocessing.Rmd)
Error in if (x == "" | x == ".") { :
missing value where TRUE/FALSE needed
Calls: ... -> -> .rasterObjectFromFile

Using here("data") works. So same problem with html_document as with reportmd->

• root dir of here() when knitting = .Rmd dir
• root dir of here() when running chunks = .Rproj dir

Setting knitr::opts_knit$set(root.dir = foo) does not work.

[krlmlr]

Does my example file work for you? After deleting the cache I can run the first example, it works as expected both for "Run all" and "knit".

[pat-s]

It works and it also works using reportMD::multi_document. I refer to the example I sent you.

However, I'm kind of lost right now. What does not work is the original source from which the example, which I sent to you, was extracted. It has the same files, the same code and the same directory structure.

Knitting there, I get

Quitting from lines 75-77 (01_preprocessing.Rmd)
Error in if (x == "" | x == ".") { :
missing value where TRUE/FALSE needed
Calls: ... -> -> .rasterObjectFromFile
In addition: Warning messages:
1: code chunks must not depend on the uncached chunk "listfiles"
2: code chunks must not depend on the uncached chunk "makeraster"
Execution halted

for both reportMD::multi_document or html_document. It again refers to

# list files
file <- here("spring-school-hsdar/data") %>% list.files(pattern = ".tif$", full.names = TRUE)

and knitting only works when using here("data").

You proofed that it works correctly and I can reproduce it on my local copy. So I assume that there is some nasty stuff going on in my source...(which I do not want you to bother with).

Thanks for your time checking the files etc! Much appreciated! I'll let you know if I'll ever find out what's going on here.

[krlmlr]

I'm getting consistent results with the following example, too:

---
title: "Part 1: Data investigation & preprocessing"
author: "Patrick Schratz"
date: "`r format(Sys.time(), '%a %d %b %Y')`"
output:
  reportMD::multi_document:
    toc: true
    fig_caption: true
    highlight: dracula 
    css: R.css
    use_namespace: false
params:
  version: !r 
---

```{r loadPackages, include=FALSE, cache=FALSE}
## load additional packages in this chunk
library(here)
library(magrittr)
```

```{r listfiles}
# list files
file <- here("spring-school-hsdar/data") %>% list.files(pattern = ".tif$", full.names = TRUE)
```

Next, we will read in these files as 'raster bricks' into R. 

```{r makeraster, dependson='listfiles'}
raster <- raster::brick(file[1])
raster
```

I'd suggest you try to remove code from your document step by step until the problem disappears. And I'd definitely be interested in the cause of this weird problem!

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue and link to this old issue if necessary.