Plot several pictures with different size in a "for" loop

Question

I am new to knitr and markdown and this is my first question asked. Maybe this question has a simple answer that I just can't find.

I have a for-loop, which creates 3 ggplots. The loop runs 300 to 400 times depending on the data input. I want to define the size of these 3 pictures as:

1st picture: width = 7, height = 3

2nd picture: width = 7, height = 3

3nd picture: width = 7, height = 12

So far I am using the following code:

```{r calc, echo=FALSE, warning=FALSE, message=FALSE, results='asis', fig.show='asis',fig.height=3}
for(x.PS in 1:length(trace.input)) 
{
# some data transformation by self-written functions

# create and save plot for the normalised version
ggp.PS.norm <- ggplot(print.PS.norm, aes(x = Time, y = Voltage, col = Pulse))
fig.PS.norm <- ggp.PS.norm + geom_line()

# create and save plot for the modified version
ggp.PS.smooth <- ggplot(print.PS.smooth, aes(x = Time, y = Voltage, col = Pulse))
fig.PS.smooth <- ggp.PS.smooth + geom_line()

# create and save plot for the modified version as facet grid
fig.PS.smooth.single <- ggp.PS.smooth + geom_line() + facet_grid(FigCol ~ FigRow)

print(fig.PS.norm)
print(fig.PS.smooth)
print(fig.PS.smooth.single)
}
```

In the end I hope to get one big PDF-file with 3 x 300 to 400 pictures

I hope this code is understandable even without any hard data.

best paj

score 4 · Accepted Answer · answered Dec 01 '14 at 18:06

I don't know if this is kosher but here is one approach:

adapted from here

I'm using rmarkdown since that is what your question is in, but this could be adapted to latex

---
output: 
  html_document:
    css: ~/knitr.css
---

```{r, include=FALSE}
library(knitr)
opts_knit$set(progress = FALSE, verbose = FALSE)
opts_chunk$set(warning=FALSE, message=FALSE, echo=FALSE)

## this function is basically creating chunks within chunks, and then
## I use results='asis' so that the html image code is rendered 
kexpand <- function(ht, cap) {
  cat(knit(text = knit_expand(text = 
     sprintf("```{r %s, fig.height=%s, fig.cap='%s'}\n.pl\n```", cap, ht, cap)
)))}

library(ggplot2)
```


```{r, results='asis', fig.width=7}
for (ii in 1:2) {
  ## do some stuff

  cat('<br />')
  cat(paste0('Page', ii))

  ## all plots should be saved as .p1 and then use kexpand
  .pl <- qplot(mpg, wt, data=mtcars)
  kexpand(3, 'fig1')

  .pl <- qplot(mpg, wt, data=mtcars, colour=cyl)
  kexpand(3, 'fig2')

  .pl <- qplot(mpg, wt, data=mtcars, size=cyl)
  kexpand(7, 'fig3')

  cat('<br /><br />')
}
```

And this is my output

enter image description here

Thanks a lot. That function seem to be very helpful on other topics too. However, I was so frustrated that I reprogrammed my whole script using knit_child(). Exact code shown in the comment below. — alias_paj, Dec 02 '14 at 10:21
Thank you so much, I thought this solution did not exist after a long google session. — Helgi Guðmundsson, Mar 02 '17 at 21:47

score 1 · Answer 2 · answered Dec 01 '14 at 20:35

1

you can print in a viewport of a specific size,

library(grid)
print_size = function(p, width=7, height=3) 
    print(p, vp=viewport(width=unit(width,"inch"), height=unit(height, "in")))

for(ii in 1:3)
   print_size(qplot(1,1), height = c(3,3,12)[ii])

answered Dec 01 '14 at 20:35

baptiste

75,767
19
198
294

score 0 · Answer 3 · edited May 23 '17 at 12:00

My own solution in the end:

At some point I call this chunk below in my parent script

<<test_child,results='asis'>>=
    knit.out <- NULL
    for(x.PS in 1:length(trace.input)) {
      knit.out <- c(knit.out,
                    knit_child("rawCAP_child_0-2.Rnw",
                           quiet = TRUE,
                           envir = globalenv()
                           )
      )
    }
    cat(paste(knit.out, collapse = '\n'))
@

The child script does some calculations and creates the figures like this

    <<,echo=FALSE>>=
# some functions creating the figures, for example the one below
ggp.PS.smooth <- ggplot(print.PS.smooth, aes(x = Time, y = Voltage, col = Pulse))
fig.PS.smooth <- ggp.PS.smooth + 
  geom_line() +
  scale_x_continuous(
    limits = c(len.1,len.total)
    ) + 
  scale_y_continuous(
    limits = c(-4,4)
    ) +
  labs(
    #title = "Data smoothed by a value of XX",
    x = "Time [sec.]",
    y = "Voltage [mV]")
@

These chunks print the figures with different heights, there is one table too.

<<,fig.height=2,fig.width=7,fig.align='center'>>=
print(fig.PS.smooth)
@

<<,echo=FALSE,results='asis',fig.align='center'>>=

xtable(table.PS.CAP.stat[c(2:5,13:16)],
       caption = "Statistics about all Pulses")
@

<<,fig.height=3,fig.width=7,fig.align='center'>>=
print(fig.PS.CAP)
@


<<,fig.height=12,fig.width=7,fig.align='center'>>=
print(fig.PS.smooth.single)
@

I am very happy with this solution, cause you can set almost everything without a workaround

For further help please look at Yihui's page, especially look here and maybe also here (this is stackoverflow). Yihui seems to be on stackoverflow too.

Best paj

Plot several pictures with different size in a "for" loop

3 Answers3

Linked