Rmd on Rmd

Paul E. Johnson

YAML Header

YAML = Yet Another Markdown Language

A set of text declarations that govern a post-processing program
Bare minimum is
```
---
title: "Rmd on Rmd"
---
```
for presentations, the designation
```
output: slidy_presentation
```
slidy_presentation is convenient, but not strictly necessary. Selector could be included in a script that does post-processing instead.

Style following that

It is very flexible design (in practice, made less flexible by assumptions about “most” users)
Some seemingly obvious details are not yet attended to (must be patient)
markdown style & instructions vary grossly across authors because of different output destination.
user can directly write any HTML desired into a presentation intended for HTML output, but same not necessarily compatible with other output formats.

Style guides & tipsheets

write more in Pandoc markdown: http://rmarkdown.rstudio.com/authoring_pandoc_markdown.html

Daring Fireball: http://daringfireball.net/projects/markdown/syntax

RStudio Docs: https://support.rstudio.com/hc/en-us/articles/200486468-Authoring-R-Presentations

Understand the process

If there were no R involved, this is what would happen.

User writes document in markdown, calls file “mgp.md” (my great presentation)
A program, currently pandoc, is used to convert the document the user prepared into HTML, PDF, and so forth.
1. Make an html slideshow in S5 format
```
pandoc -t S5 mpg.md -o mgp.html
```
  2 . Make a PDF slideshow in beamer format
```
pandoc -t beamer mgp.md -V theme:Warsaw -o mgp.pdf
```

See Producing slide shows with Pandoc by John Macfarlane, who has been a key contributor in the markdown movement.

I tested many output formats in Ubuntu 14.10. Many formats available that are interesting, not all work well with my example document (more on that below)

Important take away point: Limitations

pandoc includes routines to convert just about anything to just about anything else.
That does not imply that anything can be turned into a nice version in any other format.
- Works very well when the programmer’s idea of what should go into which format matches the user’s idea.
- Expect trouble if you prepare something to go into HTML and then try to put same into PDF, bad outcomes.
Many parts of markdown syntax are specific to the document type.
- slide markdown different from article markdown. (That’s a problem.)

Now introduce R complication

User wants to insert code chunks in a file


 ` ` `{r option-option-option}
 rcode here
 ` ` `

That fills the same role as the Sweave chunk style.

<>=
rcode here
@

Honestly, I don’t understand why the markdown people didn’t keep the same style

render equals knit + pandoc

knit: convert Rmd to md The markdown file is run and the code chunks are–literally–replaced with R input and output.
- Can be done in various ways. Most direct
  1
```
R CMD knit mgp.Rmd
```
  2
```
R -e "library(knitr); knit(\"mgp.Rmd\")"
```
pandoc: convert the new md into HTML or PDF or whatnot

Do both steps at once

Rmarkdown package

 R -e “library(rmarkdown); render("mgp.Rmd")”

RStudio creates a little confusion in terminology by using word “knit” in their GUI. Should be “render”. Read help for Rmarkdown package and the render function. The RStudio documents on what is going on, they are putting their own spin on it.

Tidbits about render

Hides & erases the intermediate files (optionally)
The “all in one” HTML document format.
- Codes images and such into MIME format, so they appear as text in the html document itself
- Same used in http://pj.freefaculty.org/R/WorkingExamples

Can’t resize images that are imported with markdown

Stop using markdown to import the image, use true HTML img tags.

<img src="importfigs/Fumble-Chart-1.png" width="700" alt="Fumbles">

images are stored in a folder importfigs, here’s going to be an image that is 700 wide on next page.

Chart asserting Patriots are Cheaters

Fumbles

Warren Sharp, “The New England Patriots Prevention of Fumbles is Nearly Impossible”, http://www.sharpfootballanalysis.com/blog/2015/the-new-england-patriots-prevention-of-fumbles-is-nearly-impossible

slides too long for page

The slide argument {.allowframebreaks} is based on the LaTeX Beamer argument, it is intended to solve this problem
However the ioslides_presentation converter in pandoc does not respect .allowframebreaks, which is a crushing disappointment.
The slidy_presentation is not quite so polished as ioslides, but it DOES respect .allowframebrakes. I used one in the “Understand the Process Slide Above”.

code lines too long

can have knitr call formatR to reformat code with line wrapping.
Since formatR fails on about 25% of the code examples I run through it, this is not a perfect solution.
Manipulate the R output. In R, run this at the start of the session
```
options(width = 70)
```
to keep line output shorter. This solution works dependably.

Cannot have R code chunk in 2 column format

Apparently unsolveable problem.

I hate double pound signs in output

Can be removed. It is discussed here http://stackoverflow.com/questions/15081212/remove-hashes-in-r-output-from-rmarkdown-and-knitr

I confirm on next slide that does work one slide at a time
Have not found document-wide solution yet. Person there asserts this will erase the point signs, opts_chunk$set(comment = NA), but I don’t know where how to use that

Lets test that

dat <- data.frame(x1 = rnorm(100), y1 = rnorm(100))
print(dat)

           x1        y1
1   -0.024779 -0.532615
2   -0.595708 -0.065403
3   -0.679282 -0.175966
4    0.016064  1.250413
5   -0.063805 -0.508312
6   -1.608321 -0.618807
7   -0.841354 -1.166242
8   -0.343233  0.688238
9    0.969689  0.304409
10  -0.232601  0.298952
11  -0.326746  0.192090
12  -1.274201  0.470676
13   0.341392 -0.424203
14   0.402024 -0.568133
15   1.296188  0.488586
16  -0.050726 -1.180882
17   0.843548  0.286996
18   0.188009  0.958711
19  -1.230055  0.130812
20  -0.181701  0.108920
21   1.666027 -0.130299
22  -0.360731 -1.528802
23  -0.011685 -0.482568
24   2.405842 -0.463031
25   0.593297 -1.436603
26   0.605710 -1.723115
27  -0.530536 -2.544667
28  -0.047831 -0.384994
29  -0.179572 -0.837010
30  -0.246159  0.762408
31   0.273934  1.794685
32  -0.119660  0.110273
33  -0.595599 -0.247153
34   0.238085  0.768608
35   0.654706 -0.578918
36   1.134818 -0.652662
37   0.639108  0.733875
38  -0.692816  0.907084
39   0.486734 -1.450094
40   1.444495  0.836457
41   0.336546 -0.781412
42  -0.435300 -1.768317
43   1.881777 -0.381787
44   1.492625 -1.013744
45  -0.041094  0.126990
46  -2.897957 -0.262404
47   1.000007 -1.415912
48   0.850069  0.587383
49  -1.703858  0.337869
50  -0.651114 -2.093873
51   0.185027 -0.277187
52   2.250010 -1.325448
53   0.782664 -0.403607
54   0.024567  0.619302
55   1.197839 -1.828095
56   0.745664 -0.127138
57   0.734189 -0.051966
58   0.093099 -0.411510
59   0.530404 -1.059356
60   0.190963 -1.120711
61   0.227050  1.567924
62  -0.616940  0.141238
63   1.448590  0.895315
64  -2.031631  0.685063
65   1.264764 -0.538453
66   0.012714  2.189421
67   0.486739 -0.481019
68   2.371039 -1.239347
69   1.217416 -0.969725
70   0.785134 -0.763601
71   0.430800 -1.683312
72   1.530047  0.434939
73   0.003901  0.025294
74   0.209648 -0.565944
75  -1.425481 -0.012462
76  -0.420065  0.435284
77  -0.570460 -1.994284
78   0.024647 -0.097332
79  -0.203127 -2.010018
80  -0.095364 -0.153945
81  -0.557543  0.195889
82  -0.313085 -0.421158
83   0.344355  1.455487
84   0.339508  0.365507
85   0.373399 -0.348570
86   1.513025  0.458718
87  -0.244489 -1.520692
88  -1.850653 -0.290735
89  -1.288580  0.545778
90  -0.477307 -0.774239
91   1.522609 -0.001126
92   0.251029  1.831316
93  -0.636104  1.058433
94  -0.139130 -0.213064
95   0.450503  1.697548
96   0.262357  1.495269
97   0.282939  0.156782
98   0.252195  0.516725
99  -1.003409 -0.400118
100 -0.942661  0.013978

I want to save the fig files and insert them in my document

This is a LaTeX strongpoint that I find lacking in the markdown approach.
Sweave has a “split” feature where we can drop all the generated figures and LaTeX output into a folder, and then later insert those things when we want them.
Especially since 2 columns does not tolerate R code embedded, such a feature would be especially helpful
Also, I want to insert same figures into a writeup.
Very difficult to un-do the cleanup effort of render here.
- I will eventually wade through the layers of simplifications and user interface “enhancements”, but it is really frustrating