pandoc.txt For Vim version 7.3 Last change: 2011 Sep 05
[This document is formatted as a markdown file and as a vim helpfile. If viewing from within vim, set ft=help to view it as a helpfile, and set ft=pandoc to view it as a (pandoc) markdown file.]
Vim Plugin for pandoc
- some settings to make vim a pleasant |pandoc-writing-environment|,
- |pandoc-syntax-highlighting|, with support for pandoc's extended markdown,
- |pandoc-snippets| for use with snipMate,
- some simple |pandoc-conversion-functions| and |pandoc-tidying-functions|,
- a few |pandoc-localleader-mappings|
Download and Install
For the latest version, see https://github.com/vim-pandoc/vim-pandoc.
If using Vundle, add
~/.vimrc file and run
If using Pathogen,
cd ~/.vim/bundle git clone email@example.com:vim-pandoc/vim-pandoc.git
If using Pathogen and managing plugins as submodules,
cd ~/.vim git submodule add https://github.com/vim-pandoc/vim-pandoc.git bundle/vim-pandoc
If you don't want to (or can't) use git, you can download a tarball from the github page and unzip it into ~/.vim/bundle.
To run this plugin, your Vim needs to be compiled with python support. To see if your Vim has python support,
and look for
Mac Users: MacVim is compiled with python support. But the
/usr/bin/vim may not be. For information about how to
override the system vim, and have
vim use MacVim's binary, see this
discussion on superuser.com.
Debian Wheezy Users: the default
vim package is not compiled with python
support. Try installing
Filetype Detection and File Extensions
If you have this plugin installed alongside vim-markdown, be aware that both plugins attempt to claim the common markdown extensions for their own. We match the following extensions:
- .markdown, .md, .mkd, .pd, .pdk, .pandoc, and .text
In our experience, vim-pandoc trumps vim-markdown.
We do not claim files with the
.txt extension, since that would seem
a bit presumptuous. If you want
.txt files to be treated as pandoc
au BufNewFile,BufRead *.txt set filetype=pandoc
The bundle sets the following options, in an effort to provide a pleasant writing environment:
setlocal linebreak setlocal breakat-=*
|linebreak| tells vim to break lines at word boundaries. The |breakat| setting is to account for asterisks used for emphasis.
If you are using soft-wrapping, this will tell vim to go ahead and show part of a long line that runs off the bottom of the screen.
|nojoinspaces| allows you to use SHIFT-J in normal mode to join the next line with the current line without adding extra unwanted spaces.
setlocal commentstring=<!--%s--> setlocal comments=s:<!--,m:\ \ \ \ ,e:-->
tells vim to use HTML style comments in your markdown files. For more, see |comments| and |commentstring|.
By default, the bundle assumes that you will be using soft wrapping, and applies
If you prefer hard wrapping, set g:pandoc_use_hard_wraps, i.e., put something like
let g:pandoc_use_hard_wraps = 1
in your vimrc. When that is set, the bundle instead sets
If you also want autoformatting, set g:pandoc_auto_format:
let g:pandoc_auto_format = 1
in your vimrc. For more information, see |formatoptions|. (Note that autoformatting can slow things down.)
The syntax highlighting should be fairly accurate and complete. By default, the bundle tries to be smart, and only highlight implicit links if there is a matching link definition. This can be expensive on large files and slow things down. To disable it, set g:pandoc_no_empty_implicits:
let g:pandoc_no_empty_implicits = 1
Due to the way they are handled, span elements (emphasis, bold, tt, subscript and superscript) can cause slowdowns. If you desire to disable highlighting of them altogether, set g:pandoc_no_spans:
let g:pandoc_no_spans = 1
Note that this will also disable highlighting of embedded html, because
otherwise stuff like
<code> (which is unformatted) is detected as containing
html that doesn't end.
If you use |snipMate|, take a look at snippets/pandoc.snippets to see what is available. Here are a couple of examples:
%%<TAB>to generate a title block, with the filename as title, the value of
g:snips_authoras author, and the current date.
[*<TAB>to generate an inline link with the contents of the clipboard as the url.
The bundle tells vim to fold documents by section. If you haven't used folding before, see |fold-commands|.
Folding can slow things down. To disable it, set g:pandoc_no_folding:
let g:pandoc_no_folding = 1
For more powerful folding and outline support, use VOoM. If you are using
VOoM, you can use
<localleader>o to open it in markdown mode when editing a
In order to complete citation keys, we need to know the location of your
bibliographical database. When opening a file, we try to find something
appropiate. We look in various places for files with the extension
.ris (RIS), or
JSON), in that order. We add them to a buffer-local variable called
If you want, you can set a default bibliography to add to this list by putting something like
let g:pandoc_bibfiles = ['/the/path/to/your/bibtex/file.bib']
in your vimrc.
You can set b:pandoc_bibfiles and g:pandoc_bibfiles to any format citeproc-hs supports.
Please note the following regarding the searches we perform in order to determine the value of b:pandoc_bibfiles: The first place we look is in the directory that contains the file in the current buffer. In that directory, we first look for a file that shares the same basename as the file in the current buffer, but with a bibfile extension. So, for example, if you are editing
we will look for
/some/long/path/to/your/file.bib /some/long/path/to/your/file.biblatex /some/long/path/to/your/file.xml
and so on. If no such file is in the directory, we then look for a file
default with an appropriate extension.
If we can't find anything in the directory of file in the current buffer,
we proceed to look for files named
default with an appropriate extension
in your pandoc data directory (~/.pandoc on POSIX systems). If you want to
use a single global bibliographic database for all your documents, we
recommend that you put it (or a symlink to it) here.
If we don't find anything in your data directory, we look in your local texmf tree, using the path specified by
kpsewhich -var-value TEXMFHOME
Again, we look for a file named
default with an appropriate extension.
You can always check to see what we've settled on for the buffer by
Using Citation Completion
To use completion, start typing a citekey, e.g.,
and then, while still in insert mode, hit CTRL-X CTRL-O (vim's shortcut for |omni-completion|), and you'll get a popup window with a list of matching keys, e.g.,
@geach1970 Geach, P. T. – Entailment @geach1972 Geach, P. T. – Logic Matters
Regular expressions work too:
should suggest both '@leftow1990' and '@lewis1990', assuming those are both keys in your bibliography.
The plugin also provides support for using this with SuperTab. You will want something like this in your vimrc:
let g:SuperTabDefaultCompletionType = "context"
Then you can just hit
<TAB> in the middle of typing a citation
to complete the citation.
By default, citation completion only matches entries by citekey. If you have bibtool installed, and your bibliography is a bibtex file, and you set
let g:pandoc_use_bibtool = 1
we will match by all keyword in all fields, so that
might return both
@armstrong1989 Armstrong, David M. - A Combinatorial Theory of Possibility @lewis1992 Lewis, David - Armstrong on Combinatorial Possibility
Omni-completion of citations is quick and powerful. But maybe you just want dictionary completion of citations? If you create a text file,
that contains a list of citation keys, one per line, like so
@adams1967a @adams1971 @adams1972a @adams1974 @adams1977 @adams1986a
we will add those keys to vim's |dictionary|, allowing for completion by typing part of a citekey, e.g.,
and then hitting CTRL-X CTRL-K (or via SuperTab).
Pandoc can be used to perform lots of different conversions. For one-off conversions, it is easy to do something like
:!pandoc -o example.rtf %
For conversions you use regularly, you can define a leader mapping in your vimrc, e.g.,
nmap <leader>p !pandoc -o %:r.pdf % --latex-engine xelatex --number-sections<CR>
Alternatively, you can define a command and a mapping using the
PandocRegisterExecutor command. It tells vim-pandoc to register some
local commands and mappings for a certain set of pandoc parameters. We
call this an executor. The syntax of
:PandocRegisterExecutor NAME MAP TYPE COMMAND
where NAME is the name of the Command, MAP is the mapping for it, TYPE is the extension of the created file, and COMMAND is a description of the command to create the document. For example:
:PandocRegisterExecutor PdcPdf <Leader>pdf pdf markdown2pdf %%
registers a command, named PdcPdf, that runs markdown2pdf on the current file, and it maps pdf to that command. It also creates a command, PdcPdfOpen, that opens the generated pdf file using your default pdf viewer, and maps pdf+ to that command.
Executors are registered before opening a file.
Executors depend on the
pandoc_execute function, which can be called
:py pandoc_execute(COMMAND, TYPE, OPEN)
where COMMAND is a string describing some pandoc arguments, TYPE is the extesion of the created file, and OPEN isn flag determining wether to open the created file or not. COMMAND can include some variables:
%%for the current filename
%:rfor the current filename without the extension
PANDOC#P_BIBSfor a sequence of --bibliography arguments for every value in
PANDOC#BIBSfor a list of the values in
Those will be evaluated before executing the command, so the user can define executors that can access those variables. For example:
:PandocRegisterExecutor PdcPdfB <Leader>pdfb pdf markdown2pdf PANDOC#P_BIBS %%
The value of command in PandocRegisterExecutor and pandoc_execute can describe an *NIX pipeline, like:
pandoc -t json %% | filter | pandoc -f json -t odt -o %:r.odt
By default, vim-pandoc provides some handy executors:
- PandocPdf (
<localleader>pdf), which runs pandoc, passing the value of
g:pandoc_bibfileas the argument of
- PandocPdfOpen (pdf+), same as above, but opens the pdf using your default pdf viewer.
- PandocHtml (
<localleader>html), which creates a standalone html.
- PandocHtmlOpen (html+), same as above, but opens the html file in your default browser.
- PandocOdt (
<localleader>odt), which creates an ODT file, passing the value of
g:pandoc_bibfileas the argument of
- PandocOdtOpen (
<localleader>odt+), same as above, but opens the ODT in your default ODT viewer.
For the mappings to work, you must have |maplocalleader| defined in your vimrc, e.g.,
let maplocalleader = ","
Then, for example, you can type something like
The plugin defines two functions designed for tidying up your markdown: MarkdownTidy and MarkdownTidyWrap. Running MarkdownTidyWrap is equivalent to
:%!pandoc -t markdown -s
Note that this can have some unexpected effects: it will replace all of your reference link ids with implicit reference links; it will replace all of your footnote ids with numbers; it will transform any setext style headers into atx style headers; it will process any latex macros you may have defined and delete the macro definitions; if your document lacks a title block, it will add one.
The plugin also sets
setlocal equalprg=pandoc\ -t\ markdown\ --reference-links
In practice, this value for |equalprg| makes it easy to hard-wrap
paragraphs nicely, using vim's
= command. If you prefer soft wrapping,
you might want to change this to
setlocal equalprg=pandoc\ -t\ markdown\ --reference-links\ --no-wrap
Note that this will remove any title block, as well as processing any
custom latex macros. So you might not want to do something like
If |maplocalleader| is defined, vim-pandoc defines a handful of localleader mappings in addition to the mappings for the default executors.
when invoked from within a url, like http://example.com, will open that url in your default browser.
Second, there are two commands for jumping back and forth between a reference-style link and its definition:
gr jumps from the link to its definition;
br jumps back. If you want
to try this out on this document, and you are reading this within vim,
:set maplocalleader = "," :set ft=pandoc
then move your cursor somewhere in here, and type
,gr. Wasn't that
fun?[^note] Now you probably want to go ahead and
to get back to viewing this as a help file.
[^note]: Also supports footnotes. Type
,br to jump back.
Finally, there is
,nr creates a new reference definition from within a reference link.
So if you have typed something like
This is a [link][id]. And here is another sentence.
,nr from somewhere within
[link][id], you will get
This is a [link][id]. And here is another sentence. [id]: <your cursor here>
Of course, when you are done filling in the link, you can hit
jump back to where you were.
See the github issues page for a current set of issues.
Thanks also to Wei Dai for bug fixes and improvements.
The snippets file is a slight extension (and contraction) of the markdown.snippets file that is part of many of the vim-markdown repositories on github.
Autocompletion was implemented by hacking away at LaTeX Box's implementation of bibtex citation completion, even if the results don't look much like the original.