To modify default HTML template on-the-fly

[jsx97]

Is it possible to modify the default HTML template on-the-fly, without using a file? Specifically, I need to strip out this line:

<title>$if(title-prefix)$$title-prefix$ – $endif$$pagetitle$</title>

Is it possible to do using something like the example below?

pandoc input.md -s --template=default | sed ... -o output.htm 

[jsx97]

Okay, to answer my own question. To store contents of the default template in a variable and strip the <title> line from there, one can use

template=$(grep -v '<title>' <<< $(pandoc -D html))

or, if using Zsh,

template=${(F)"${(f@)$(pandoc -D html)}":#*<title>*}

But then we need to pass the resulting string as a --template value. I will open a separate discussion about this.

[tarleb]

Alternative approach: use a custom Lua writer and remove the title line in Lua.

Extensions = pandoc.format.extensions 'html5'
Template = pandoc.template.default('html5'):gsub('\n<title>[^\n]*', '')
Writer = function (doc, opts)
  return pandoc.write(doc, {format='html5', extensions=opts.extensions}, opts)
end

[jsx97]

Many thanks, Albert!

[jsx97]

Because the <title> line is indented, the regex should also match spaces before it. I tried %s (and, of course, \s*), which should match zero or more spaces, but, for some reason, it doesn't work for me. Could you point what is wrong here? I never worked with Lua-flavored regular expressions before.

gsub('\n%s<title>[^\n]*', '')

Edit: Nevermind, it works.

[jsx97]

Albert, maybe you know, is it possible to use this code not as a custom writer, but as a regular Lua filter?

[tarleb]

I don't think it's possible¹ to do that in a filter, as only the writer can modify the template.

¹ or rather wise, in that a lot of complicated hacks could be thought up, but none that seem sensible.

[jsx97]

@tarleb Hello, Albert. I know this is an old thread, and this is not related to Pandoc directly, but could you explain why you have used Writer = function (doc, opts) instead of Writer (doc, opts)?

Another example, from #9697. Your code:

local hex_to_char = function(x)
  return string.char(tonumber(x, 16))
end

local unescape = function(url)
  return url:gsub("%%(%x%x)", hex_to_char)
end

function Image (img)
  if not (img.src:match('^https://') or img.src:match('%.%.')) then
    return nil
  end

  local mt, contents = pandoc.mediabag.fetch(img.src)
  img.src = pandoc.path.filename(unescape(img.src))
  pandoc.mediabag.insert(img.src, mt, contents)

  return img
end

Why you have used local hex_to_char = function(x) and local unescape = function(url) instead of local function hex_to_char(x) and local function unescape(url)?

One another example, from #9514. Here you use local function starts_with(start, str), and not local starts_with = function(start, str):

local function starts_with(start, str)
  return str:sub(1, #start) == start
end

function RawInline (el)
  return starts_with('<!--', el.text)
    and pandoc.RawInline('markdown', el.text)
    or {}
end

Asking such silly questions is my regular way to grok into something new to me, like a new programming language. :)

[tarleb]

The Writer = ... is just for symmetry with the other definitions. It just looks slightly more pleasant to my eyes. There's no deeper meaning.

It's similar for the other code snippets: I sometimes use the alternative syntax to mark some functions as being different than others. E.g., to make the distinction of helper functions and filter functions more obvious. But it's not a consistent convention, as you noticed.