On further looking, this has been previously reported: https://groups.google.com/forum/#!topic/knitr/MJVLiVyGCro

If there were a way to make this fix automatic, that would be beneficial. If not, perhaps it could be documented in more detail.

As a followup thought: Perhaps a simple knit_print method doing similar to the following would fix the issue. Since it has caused confusion for about 4 years (given the Google groups thread), it will probably continue to cause problems in the future, too.

knit_print.ggplot <- function(x, ..., inline=FALSE) {
  if (!inline) {
    cat("\n\n")
  }
  print(x, ...)
  if (!inline) {
    cat("\n\n")
  }
}

And then doing similar for base graphics and lattice so that they all use the same method. The real method would probably need to be more complex, but hopefully this general idea could help.

@billdenney thanks for the solution

@yihui I am not sure how to classify this one.

Currently, for LaTeX output when there is caption, we add a space after the markdown syntax so that it is not seen by Pandoc as a inline image.

![cap](img/path) 

However, when there are several images in the same chunk and echo = FALSE, the chunk output will be wrap using hook plot with the syntax above, and results will be pasted on the same line. So we get

![cap1](path/img1) ![cap2](path/img2) 

This leads to no caption by Pandoc as seen as inline images.

This does not happen when echo = TRUE because the are some \n\n appended before and after a source code chunk.

This also happens with HTML output also.

Should we consider the addition of cat('\n\n') a solution for this case or could we add a new line in hook_plot_md() so that it does not happen ?

I am not sure of the potential effect of such change.

@cderv It seems the PR #1760 solves this problem.

Oh great thanks @atusy ! I am going though old issues to make some triage and know what is in the backlog before looking at the PR. Aim is to match when possible! And you've just help me. Thank you!

I'll link these 2 so that we remember it will close.