CODEWITHSUNDEEP
r-markdownparameter-passingknitr
Rob Creel
Rob Creel

Reputation: 573

How do I knit child documents with parameters into a main RMarkdown document?

I have a parameterized RMarkdown file, parameterized.Rmd, with the following contents.

---
title: "Parameterized report"
output: html_document
params:
  input_df: NULL
---
```{r sec1}
head(params$input_df[, 1:2])
```

I can knit it from the console using rmarkdown::render, generating distinct documents for distinct dataframe inputs. This works as expected.

rmarkdown::render("parameterized.Rmd", 
                params = list(input_df = mtcars), 
                output_file = "cars.html")

rmarkdown::render("parameterized.Rmd", 
                  params = list(input_df = iris), 
                  output_file = "iris.html")

I would like to have each of these results be child documents to a common document. My first attempt is with knitr::knit_child, but it does not take params as an argument. So this failed.

---
title: "Main report"
output: html_document
---


```{r test-cars}
knitr::knit_child("parameterized.Rmd", envir = environment(), quiet = T,
                  params = list(input_df = mtcars))
```

How may I knit together child documents which require parameters?

Upvotes: 10

Views: 3113

Answers (2)

Monkj_
Monkj_

Reputation: 11

You can actually also "overwrite" parameters in the child document if you run it in a child environment where you manually add the params variable. It looks like the params set in YAML in the child document won't be used if you knitr::knit_child(). So this should work:

parameterized.Rmd

---
title: "Parameterized report"
output: html_document
params:
  input_df: NULL
---

```{r sec1}
head(input_df[, 1:2])
```

main.Rmd

---
title: "Main report"
output: html_document
---

# mtcars
```{r mtcars, echo=FALSE, results='asis'}
e1 <- new.env()
e1$params$input_df <- mtcars

cat(
  knitr::knit_child('parameterized.Rmd', envir = e1, quiet = TRUE)
  )
```

# iris
```{r iris, echo=FALSE, results='asis'}
e2 <- new.env()
e2$params$input_df <- iris 

cat(
  knitr::knit_child('parameterized.Rmd', envir = e2, quiet = TRUE)
  )
```

The outcome is quite similar. However, the child report will now be evaluated in a child environment. If you want to have variables from the child report to be available in the main report you have to set them via "<<-". Like this your global environment won't get flushed with variables that you might not need there. I use this for a child report that can also be knitted as a standalone version:

SeparateReport.Rmd

---
title: "Separate Report"
output: html_document
params:
  fullprint: TRUE
---

```{r, eval=fullprint}
# do this only if knitting the full version is intended
...

```

```{r}
# always include this chunk
...
some_var <<- some_local_result
```

Main.Rmd

---
title: "Main Report"
output: html_document
---

```{r child1}
e <- new.env()
e$params$fullprint <- FALSE

knitr::knit_child("SeparateReport.Rmd", envir = e)
print(some_var)
```

Or you can pass the parameter from the main file

Main.Rmd

---
title: "Main Report"
output: html_document
params:
  fullprint: FALSE
---

```{r child1}
knitr::knit_child("SeparateReport.Rmd")
print(some_var)
```

Upvotes: 0

Rob Creel
Rob Creel

Reputation: 573

What worked for me (derived from the documentation once I properly understood it.):

Instead of using the params field in the YAML header, set the values of the parameters inside the main document and call cat on the output of knitr::knit_child. The below files achieved the desired result.

parameterized.Rmd

---
title: "Parameterized report"
output: html_document
---

```{r}
head(df[, 1:2])
```

main.Rmd

---
title: "Main report"
output: html_document
---

# mtcars
```{r mtcars, echo=FALSE, results='asis'}
df <- mtcars 
cat(
  knitr::knit_child('parameterized.Rmd', envir = environment(), quiet = TRUE)
  )
```

# iris
```{r iris, echo=FALSE, results='asis'}
df <- iris 
cat(
  knitr::knit_child('parameterized.Rmd', envir = environment(), quiet = TRUE)
  )
```

Knitting main.Rmd applied the parameterized report to each dataframe.

Upvotes: 12

Related Questions