However, a caveat. It is not mentioned in the documentation, but Gitit does not enable PDF export by default, when Pandoc is more than capable of doing so? Why? Because, in all fairness, it can add significant load. According to the conf file comments.
> pdf-export: yes
> # if yes, PDF will appear in export options. PDF will be created using
> # pdflatex, which must be installed and in the path. Note that PDF
> # exports create significant additional server load.
How much load, I am not sure. I would love to see a similar Happstack app like this program that runs gitit or the new fork gitit2, rewritten for Happstack updates now that the API has matured. This makes me think performance is an issue.
As an aside, I know ShareLatex opened up their code. I have not had time to check, but I was curious if they do use standard Latex for generating files and if they do, how they make it performant (I know they have a couple layers, I watched their old Node/JS London talk, I mean hear Latex in particular) so fast for potentially thousands of people at once.
[0] https://gitprint.com/jgm/gitit/blob/master/README.markdown
Is it surprising that processing inline TeX equations or Bibtex citations takes more resources? No.
Compare and Contrast:
Gitprint's pdf of README: https://gitprint.com/jgm/gitit/blob/master/README.markdown
Gitit's html export of README: http://gitit.net/README?printable
If I don't care about the quality of the text layout gitit's printable html page is equivalent to the gitprint pdf. Adding Merriweather to default.html is trivial. But as another commentator pointed out maybe using a new trendy font might not be the best choice if you want decent unicode coverage; just in case one of your users is among the "minority" of people who were unsatisfied by the basic ASCII charset.
Moreover comparing gitprint's markdown support to pandoc as if it is apples to apples is silly.
Is Latex requiring processing power to generate PDFs that become professional publishing house books surprising? Of course not. Latex is sick. I was a non-tech student in college, and when I discovered Latex in a CS class my mind was blow and loved it for a while, despite fucking up regularly with its learning curve. (Funny anecdote: I even had a terrible philosophy professor in my liberal arts college senior year who went to Stanford and noticed he used Latex for everything, unlike any professor at my tech-illiterate school; he used it because his friend was this guy Donald Knuth; I kissed that guy's ass for the rest of the semester after that tidbit just for stories about Knuth). My point was scaling Latex in a server environment would be difficult, and I was addressing a few people equating this whole thing to Pandoc, which I did not find fair. Running Latex concurrently for lots of people sounds rough to me, but then again I never implemeneted anything nearly as complicated as that.
Gitit does not just give you just a printable HTML page. If you notice, gitit uses the Pandoc backend to export many formats. It might not set them as printable, but to see gitit+pandoc in action, look at the export menu to the bottom left corner of the default them on the page you link. You do not see the PDF format export, because as I mentioned in my original comment, by default Pandoc Latex+PDF support is not enabled because of said performance load.
I was speculating writing an equivalent of this service with Pandoc and Haskell is doable, but solely generating PDFs would be a load that might make it difficult at scale. I do not know. ShareLatex, who arguably is way ahead beyond my speculation because they implemented a powerful web-based Latex editor, did some very unorthodox (as far as just reloading Latex in the shell every time) architecture to make it fast and resilient.
I am not sure how to respond to the comment re Gitit and Pandoc. Gitit is basically a wrapper around Pandoc, and it is merely a web interface to everything Pandoc does for different document formats, and hackage clearly lists it as a dependency with version 1.10+.[0] I know they are not apple to apple comparison. It was just a speculation. In the other post I was actually using Gitit and Pandoc interchangeably because the latter is the integral component of the former.
I feel there is a type checking joke lurking in my last line, but I am too tired to find it.
To be clear this project:
1. Converts a certain unknown basic markdown syntax to html
(iff the filename is something that it recognizes[^1])
2. adds css for Merriweather, Open Sans, Dejavu Mono and a silly iconfont
(I prefer pandoc's pygments syntax highlighting over goofy icon)
3. renders html->pdf using webkit by way of phantomjs.
Pandoc to pdf at ROFLSCALE? Speculate no more: $ pandoc -s -S -t html5 -o out.html README.md
$ wkhtmltopdf out.html out.pdf
or $ pandoc -s -S -t html5 -o out.html README.md
$ phantomjs rasterize.js ./out.html out.pdf Letter
To be honest using pdflatex is just as fast as phantomjs/wkhtmltopdf.The type checking joke is that I never compared gitit to pandoc. So don't worry about responding to any comment you think I made regarding gitit and pandoc.
[^1]: https://gitprint.com/jgm/pandoc/master/README = "Something went wrong! Couldn't process https://raw.github.com/jgm/pandoc/master/README/master/READM... = ROFLSATIRE