Notebook: missing markdown features: image size and colspan

[amueller]

So I'm trying to write a book in a notebook. I'd like to be able to have stuff render in the notebook and work with nbconvert. Currently nbconvert / pandoc strips all html when converting to latex. Which leads me to writing markdown.

However, markdown is fairly limited as far as I can tell. The two things I currently struggle with are image size and "complex" tables:

• There doesn't seem to be a way to specify the size of an image in markdown.
• There doesn't seem to be a way to do colspan or rowspan in markdown.

Please correct me if I'm wrong.

To be able to write "complex" documents like books, it would be great to either extend the markdown, or improve the handling of html when converting. I'm not sure which route is more promising.

See
https://github.com/ipython/ipython/issues/3503
and
https://github.com/jupyter/nbconvert/issues/241

[amueller]

multimarkdown looks interesting btw: http://fletcherpenney.net/multimarkdown/

[amueller]

Hm so it looks like image sizing was added to pandoc recently, so it will work with nbconvert, but it doesn't look like it works in the notebook itself: https://github.com/jgm/pandoc/commit/244cd5644b44f43722530379138bd7bb9cbace9b

[takluyver]

It appears that there's no mention of that syntax in the latest CommonMark spec.

[amueller]

We could add a custom renderer to marked. Not sure if the notebook already is messing with marked or if you're trying to stay away from that: https://github.com/chjj/marked/issues/339

[amueller]

does CommonMark include tables at all?

[TomAugspurger]

Sample notebook: https://gist.github.com/fb654d5e66aed5db2ed9 demonstrating the new pandoc image attributes syntax.

A PDF from jupyter nbconvert --to=pdf: Untitled.pdf, which is converted with the image size reduced.

The HTML version doesn't get converted with the size set according to the image attributes.

[takluyver]

I think the only place where we're messing with marked is to prevent it from parsing latex-y math expressions, so that MathJax can do its stuff with them after the markdown is rendered.

[amueller]

I'm asking pandoc for help here: https://github.com/jgm/pandoc/issues/2716

having html be interpreted by pandoc when using nbconvert would solve my problems.

[amueller]

@minrk proposed to do markdown -> html -> target to get around the pandoc restriction.

[amueller]

actually, that won't work. pandoc can't render colspan tables. If you try to convert a html table with colspan to latex, the table layout is discarded. If you don't have a colspan, the table is rendered correctly.

[randy3k]

I have exactly the problem of @amueller on the image size issue. I end up in writing a simple script with the following functions to convert between ![](){width=..} and <img></img> syntax.

re_mdimg = re.compile(r"(!\[([^\]]*)\]\(([^\)]*)\)\{(.+?)\})")
re_imgtag = re.compile(r"(<img ([^>]*?)(?:/>|>(.*?)</img>))")


def mdimg_to_imgtag(para):
    for (match, tag, link, opts) in re_mdimg.findall(para):
        w = re.search(r"width\s*=\s*(\S*)", opts)
        h = re.search(r"height\s*=\s*(\S*)", opts)
        width = "width: %s;" % w.group(1) if w else ""
        height = "height: %s;" % h.group(1) if h else ""
        style_opts = " style=\"%s%s\"" % (width, height) if w or h else ""

        if tag:
            para = para.replace(
                match,
                "<img src=\"%s\"%s>%s</img>" % (link, style_opts, tag))
        else:
            para = para.replace(
                match,
                "<img src=\"%s\"%s />" % (link, style_opts))
    return para


def imgtag_to_mdimg(para):
    for (match, opts, tag) in re_imgtag.findall(para):
        tag = tag if tag else ""
        l = re.search(r"""src\s*=\s*["'](\S*?)["']""", opts)
        w = re.search(r"""width\s*[=:]\s*["']?(\S+?)["'; ]""", opts)
        h = re.search(r"""height\s*[=:]\s*["']?(\S+?)["'; ]""", opts)
        style_opts = ""
        style_opts = "width=%s" % w.group(1) if w else ""
        if h:
            if style_opts:
                style_opts = style_opts + " "
            style_opts = style_opts + "height=%s" % h.group(1)
        para = para.replace(
            match,
            "![%s](%s){%s}" % (tag, l.group(1), style_opts))
    return para

[ianhbell]

I'm struggling with exactly this issue right now. I can use tag to resize image for markdown, but I would like to be able to print as PDF via latex, and all the tags get stripped completely :(

[soamaven]

@randy3k where/how do you implement this script when converting to pdf?

[randy3k]

@soamaven
You can export the notebook as .tex or .md, then apply the functions and then compile the latex via pandoc/latex.

[cekees]

Has there been any action on the basic issue of allowing image scaling directly in markdown?

[amueller]

I don't think so since that would require a different dialect of markdown.
Depending on what your target is, you might be able to use custom
nbconverter processors to achieve this, though.

Sent from phone. Please excuse spelling and brevity.

On Apr 6, 2017 11:11, "Chris Kees" [e-mail] wrote:

Has there been any action on the basic issue of allowing image scaling
directly in markdown?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/jupyter/notebook/issues/1095#issuecomment-292205337,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAbcFpIZCy29Tg7H9bHbVHVqgtE_j5VFks5rtQCagaJpZM4HalzH
.

[gandalfsaxe]

The problem with using HTML tag is that it doesn't support the new attachement type, where you can drag n' drop / paste from clipboard an image into a markdown cell. The HTML tag only works on local image files / links to images.