How to include output from the end of an Rmd in its introduction or TL;DR

[joepowers16]

It's always great to lead off a report with the key takeaways, for instance a TL:DR header. But it's also great to have a reproducible report where key takeaways are directly linked to your analysis via inline code (E.g., Mean performance was r MY_MEAN).

I'm wondering if there's a smart way to include stats in the TL;DR of an RMarkdown report when the TL;DR necessarily comes before the bulk of the code chunks that would generate its values.

[technocrat]

I prefer putting almost everything in the opening chunk with assignment to named objects, which are them availability inline or for subsequent chunks.

[olyerickson]

I believe you can use the dependson code chunk option for this. Put your tl;dr code in a chunk with dependson set to a character vector of the (labelled) chunks providing values to be listed.

See https://yihui.org/knitr/options/ for further info...

[cderv]

Hi @joepowers16!

I believe you can do that several way. here are some ressources that will help find the solution that suits you.

• Best solution is with caching mechanism - your own or the one provided with rmarkdown.

  • See about using an object before it is created (https://bookdown.org/yihui/rmarkdown-cookbook/load-cache.html).
  • and new cache mechanism (https://bookdown.org/yihui/rmarkdown-cookbook/cache-rds.html) that complement the existing one (https://bookdown.org/yihui/rmarkdown-cookbook/cache.html)

  The trick is that it will work with two rendering - first one will not found the object be the second one will.

There is also other feature that could help:

• You can reuse some chunk easily - meaning defining a chunk once but use it in several place https://bookdown.org/yihui/rmarkdown-cookbook/reuse-chunks.html
• You can make some content of the Rmd file conditional (https://bookdown.org/yihui/rmarkdown-cookbook/eng-asis.html)
• Regroup all code in the same place - https://bookdown.org/yihui/rmarkdown-cookbook/code-appendix.html

Hope it helps for your own needs.

[joepowers16]

Using cache with two renderings worked great!

[joepowers16]

[cderv]

@joepowers16 you can manage to not have to removed then paste back.

The function knitr::load_cache() handles for you the fact that if the variable is not found it will not error.
Example:

---
title: An important report
output: html_document
---

# TL;DR

* The mean mpg is `r knitr::load_cache("mpg", "my_mean")`
* The median mpg is `r knitr::load_cache("mpg", "my_median")`

# Details

We calculate some stats

```{r mpg, cache=TRUE}
my_mean <- mean(mtcars$mpg)
my_median <- median(mtcars$mpg)
```

First run will show NOT AVAILABLE.

You can recreate this by following the first example in the book chapter using the if clause.

The trick is to print something if the object is not found to not error and render anyway the first time. knitr::load_cache to that for you.