emacs-orgmode
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Orgmode] LaTeX export of images


From: Nick Dokos
Subject: Re: [Orgmode] LaTeX export of images
Date: Tue, 29 Sep 2009 15:32:42 -0400

Carsten Dominik <address@hidden> wrote:

> > 1) Images in the middle of the text
> >
> > If I have the following example:
> >
> >    ,----
> >    | For this example, I will use this image [[myimage.png]] which  
> > is a very good one.
> >    `----
> >
> > Then, when exported, it produces the following LaTeX code:
> >
> >    ,----
> >    | For this example, I will the use this image
> >    | \centerline{\includegraphics{myimage}} which is a very good one.
> >    `----
> >
> > Which produces weird layout when compiled to PDF.
> >
> > In this case, we should just generate \includegraphics{myimage}  
> > without the
> > centerline command.
> >

Agree.

> >
> > 2) Images in floats
> >
> > In the case of floats, the correct generated code should be  
> > something like:
> >
> > \begin{figure}[htb]
> >        \centering
> >        \includegraphics{myimage}
> >        \label{LABEL}
> >        \caption{CAPTION}
> > \end{figure}
> >
> > and not
> >
> > \begin{figure}[htb]
> >        \centerline{\includegraphics{myimage}}
> >        \label{LABEL}
> >        \caption{CAPTION}
> > \end{figure}
> >
> > This can also be fixed quickly I think.
> 
> Yes sure, but what is wrong with \centerline?  does the \centering  
> lead to a different result?
> 

I've been running with the following patch for a little while and have
seen no problems (it does \centering rather than \centerline but I don't
think it makes a difference for an image - it would make a difference for a
floating centered paragraph with multiple lines however.)

There is another problem as well: there is a \n added after the
\end{figure} which leads to spurious paragraphs. The patch fixes
that too.


diff --git a/lisp/org-latex.el b/lisp/org-latex.el
index 4058891..248d653 100644
--- a/lisp/org-latex.el
+++ b/lisp/org-latex.el
@@ -1553,8 +1585,8 @@ The conversion is made depending of STRING-BEFORE and 
STRING-AFTER."
        (cond ((and imgp (plist-get org-export-latex-options-plist 
:inline-images))
              (insert
               (concat
-               (if floatp "\\begin{figure}[htb]\n")
-               (format "\\centerline{\\includegraphics[%s]{%s}}\n"
+               (if floatp "\\begin{figure}[htb]\n\\centering\n")
+               (format "\\includegraphics[%s]{%s}\n"
                        attr
                        (if (file-name-absolute-p raw-path)
                            (expand-file-name raw-path)
@@ -1563,7 +1595,7 @@ The conversion is made depending of STRING-BEFORE and 
STRING-AFTER."
                    (format "\\caption{%s%s}\n"
                            (if label (concat "\\label{" label "}") "")
                            (or caption "")))
-               (if floatp "\\end{figure}\n"))))
+               (if floatp "\\end{figure}"))))
             (coderefp
              (insert (format
                       (org-export-get-coderef-format path desc)


But there is the additional question that's raised in the thread:

Carsten Dominik <address@hidden> wrote:

> 
> On Sep 23, 2009, at 11:09 PM, S=E9bastien Vauban wrote:
> 
> > Jeff Kowalczyk wrote:
> >> S=E9bastien Vauban <address@hidden> writes:
> >>> Francesco Pizzolante wrote:
> >>>> In this case, we should just generate \includegraphics{myimage} =20
> >>>> without
> >>>> the centerline command.
> >>>
> >>> Given the output of the `\centerline' command, I guess that the =20
> >>> inlining of
> >>> an image inside a line of text is not supported as is by Org. I =20
> >>> can second
> >>> such a choice.
> >>
> >> +1 for default to inline graphics. When using Org for How-To and =20
> >> project
> >> documentation delivered as PDF, floating screenshot-size images =20
> >> typically
> >> situate far from the descriptive text.
> >
> > My point was mainly that, IMHO, floating or not should not be tied =20
> > to the
> > presence or absence of caption/label.
> >
> > Your problem with screenshots far away should be solved by putting =20
> > the H
> > specifier to the float environment and loading the `float' package.
> >
> >
> >> I would prefer inline images by default (or option), and a file/=20
> >> subtree
> >> level option to use LaTeX floating layout.
> >
> > That'd be a great way to be able to specify which type of "inlining" =20=
> 
> > we want.
> > On a per file basis seems sufficient for me.
> 
> Hi Sebastien,
> 
> how about working out a detailed proposal how this should work?
> 
> - Carsten
> 
> 

Currently, specifying either a #+CAPTION or a #+LABEL (or both) floats
the image.  Absent both of these the image is inlined (although without
the above patch, it's inlined badly).

LaTeX implicitly assumes that a floating figure has a caption: that's
where the figure number is incremented for example. Given that, it does
not seem productive to have a separate indicator for floats.

Of course, if one is willing to forego the incrementing of the counter,
then a float without a caption is possible in LaTeX - but is it useful?
If it is, then using the #+LABEL specifier alone is enough to float the
image (but currently, there is a problem in org-latex in that a \caption
is always output - but that is a minor problem, easily fixed, if so
desired).

To summarize: org-latex currently always produces a \caption which
causes LaTeX to increment the figure counter, but also to produce
a "Figure <N>:" caption, even if the caption is empty. We could
change org-latex not to produce a \caption if #+CAPTION is absent,
but that would mean that the figure counter does not get incremented.

I'm not sure that a saparate specifier is needed (it might be
desirable only on readability grounds: saying #+FLOAT, rather than
#+CAPTION or #+LABEL makes it a bit more obvious). These specifiers are
per-image, afaict, not per-file/subtree. For my part, I don't
see the need for the coarser kind of control: per-image is enough.
But if desired, then maybe

#+FLOAT: t (default nil)

or perhaps better

#+OPTIONS: float:t

can be used for the coarser control.

I hope this is clear but if there are questions, let me know. What do
you think?

Thanks,
Nick















reply via email to

[Prev in Thread] Current Thread [Next in Thread]