[Enhancement] RAW format

[sebi5361]

Enhancement: Adding the raw format

Sometimes one might have to preprocess a file to make it pandoc-compatible at first.
Adding the raw format would allow to use the pandoc CLI syntax to make preprocessing like standard pandoc filtering.

Illustration

My Markdown file contains @sinx@ which is common for AsciiMath notation.
I need to parse it to render AsciiMath notation to $:a sinx$ to use yuwash/asciimathml-pandocfilter afterwards.
As @ is a special element in pandoc-markdown, this substitution needs to be done beforehand.
Currently I am using a pre-pandoc-parser, but with the added raw format doing so might become useless.
The raw format would prevent pandoc from parsing its input to its internal dialect, thus allowing filters not based on pandoc grammar necessarily.

Commands

Typing
cat file.md | pandoc -f raw -F my_pre_pandoc_parser_becoming_a_pandoc_standard_filter_now | pandoc...
rather than
cat file.md | ./my_pre_pandoc_parser | pandoc...

---

• Is this a good idea?
• Can it be done easily?

[mb21]

I think this already exists and it's called -f native ..?

Or maybe I misunderstand.. from your example seems that the proposed raw format wouldn't do anything? What's wrong with this?

cat file.md | ./my_pre_pandoc_parser | pandoc...

a filter is not a preprocessor...

[alerque]

The raw format would prevent pandoc from parsing its input to its internal dialect, thus allowing filters not based on pandoc grammar necessarily.

What you are asking for is simply nonsensical. The filter system operates on the native internal AST format. What you want is a preprocessor not a filter. So just pre-process it. If you want syntax sugar try something like pandoc <(my_ preprocessor file.md) ....

If your preprocessor converts to Pandoc's AST format then you can read it in as JSON or native syntax.

[sebi5361]

Indeed what I was asking is nonsensical once understanding the philosophy around pandoc.
I have nothing against the syntax of using a preprocessor. It was more about syntax sugar indeed.
Thanks so much for your enlightenments 😃
Am closing this thread.

[tarleb]

I think @sebi5361's proposal this is not too far from some ideas mentioned in #6393.

[tarleb]

BTW, on Mac and Linux you can simulate a raw reader like so:

pandoc <(printf '```{=raw}') input.file <(printf '```')

This will produce a single RawBlock element which can then be processed further.

[alerque]

@tarleb I think @sebi5361 is expecting the input to still be processed as Markdown by Pandoc.

[tarleb]

Thanks @alerque; I should have read more closely. 👍

[sebi5361]

Very close indeed to what @tarleb mentioned.
Good to know about the {=raw} / RawBlock field encapsulation possibility.
Thanks to all of you for your reactivity and explanations 😀