markdown2html.py

""" Notice how this file is completely independent of sublime text

I think it should be kept this way, just because it gives a bit more organisation,
and makes it a lot easier to think about, and for anyone who would want to, test since
markdown2html is just a pure function
"""

import io
import struct
import os.path
import concurrent.futures
import urllib.request
import base64
import bs4

from functools import partial

from .lib.markdown2 import Markdown

__all__ = ("markdown2html",)

markdowner = Markdown(extras=["fenced-code-blocks", "cuddled-lists"])

# FIXME: how do I choose how many workers I want? Does thread pool reuse threads or
#        does it stupidly throw them out? (we could implement something of our own)
executor = concurrent.futures.ThreadPoolExecutor(max_workers=5)

def markdown2html(markdown, basepath, re_render, resources, viewport_width):
    """ converts the markdown to html, loads the images and puts in base64 for sublime
    to understand them correctly. That means that we are responsible for loading the
    images from the internet. Hence, we take in re_render, which is just a function we 
    call when an image has finished loading to retrigger a render (see #90)
    """
    html = markdowner.convert(markdown)

    soup = bs4.BeautifulSoup(html, "html.parser")
    for img_element in soup.find_all("img"):
        src = img_element["src"]

        # already in base64, or something of the like
        # FIXME: what other types are possible? Are they handled by ST? If not, could we
        #        convert it into base64? is it worth the effort?
        if src.startswith("data:image/"):
            continue

        if src.startswith("http://") or src.startswith("https://"):
            path = src
        elif src.startswith("file://"):
            path = src[len("file://") :]
        else:
            # expanduser: ~ -> /home/math2001
            # realpath: simplify that paths so that we don't have duplicated caches
            path = os.path.realpath(os.path.expanduser(os.path.join(basepath, src)))

        base64, (width, height) = get_base64_image(path, re_render, resources)

        img_element["src"] = base64
        if width > viewport_width:
            img_element["width"] = viewport_width
            img_element["height"] = viewport_width * (height / width)

    # remove comments, because they pollute the console with error messages
    for comment_element in soup.find_all(
        text=lambda text: isinstance(text, bs4.Comment)
    ):
        comment_element.extract()

    # FIXME: how do tables look? should we use ascii tables?

    # pre aren't handled by ST3. The require manual adjustment
    for pre_element in soup.find_all("pre"):
        # select the first child, <code>
        code_element = next(pre_element.children)

        # FIXME: this method sucks, but can we do better?
        fixed_pre = (
            str(code_element)
            .replace(" ", '<i class="space">.</i>')
            .replace("\n", "<br />")
        )

        code_element.replace_with(bs4.BeautifulSoup(fixed_pre, "html.parser"))

    # FIXME: highlight the code using Sublime's syntax

    # FIXME: report that ST doesn't support <br/> but does work with <br />... WTF?
    return "<style>\n{}\n</style>\n\n{}".format(resources["stylesheet"], soup).replace(
        "<br/>", "<br />"
    )

images_cache = {}
images_loading = []

def get_base64_image(path, re_render, resources):
    """ Gets the base64 for the image (local and remote images). re_render is a
    callback which is called when we finish loading an image from the internet
    to trigger an update of the preview (the image will then be loaded from the cache)

    return base64_data, (width, height)
    """

    def callback(path, resources, future):
        # altering images_cache is "safe" to do because callback is called in the same
        # thread as add_done_callback:
        # > Added callables are called in the order that they were added and are always
        # > called in a thread belonging to the process that added them
        # > --- Python docs
        try:
            images_cache[path] = future.result()
        except urllib.error.HTTPError as e:
            images_cache[path] = resources['base64_404_image']
            print("Error loading {!r}: {!r}".format(path, e))

        images_loading.remove(path)

        # we render, which means this function will be called again, but this time, we
        # will read from the cache
        re_render()

    if path in images_cache:
        return images_cache[path]

    if path.startswith("http://") or path.startswith("https://"):
        # FIXME: submiting a load of loaders, we should only have one
        if path not in images_loading:
            executor.submit(load_image, path).add_done_callback(partial(callback, path, resources))
            images_loading.append(path)
        return resources['base64_loading_image']

    with open(path, "rb") as fhandle:
        image_content = fhandle.read()
        width, height = get_image_size(io.BytesIO(image_content), path)

        image = "data:image/png;base64," + base64.b64encode(image_content).decode(
            "utf-8"
        )
        images_cache[path] = image, (width, height)
        return images_cache[path]


def load_image(url):
    with urllib.request.urlopen(url, timeout=60) as conn:

        image_content = conn.read()
        width, height = get_image_size(io.BytesIO(image_content), url)

        content_type = conn.info().get_content_type()
        if "image" not in content_type:
            raise ValueError(
                "{!r} doesn't point to an image, but to a {!r}".format(
                    url, content_type
                )
            )
        return (
            "data:image/png;base64," + base64.b64encode(image_content).decode("utf-8"),
            (width, height),
        )


def get_image_size(fhandle, pathlike):
    """ Thanks to https://stackoverflow.com/a/20380514/6164984 for providing the basis
        of a working solution.

    fhandle should be a seekable stream. It's not the best for non-seekable streams,
    but in our case, we have to load the whole stream into memory anyway because base64
    library only accepts bytes-like objects, and not streams.

    pathlike is the filename/path/url of the image so that we can guess the file format
    """

    format_ = os.path.splitext(os.path.basename(pathlike))[1][1:]

    head = fhandle.read(24)
    if len(head) != 24:
        return "invalid head"
    if format_ == "png":
        check = struct.unpack(">i", head[4:8])[0]
        if check != 0x0D0A1A0A:
            return
        width, height = struct.unpack(">ii", head[16:24])
    elif format_ == "gif":
        width, height = struct.unpack("<HH", head[6:10])
    elif format_ == "jpeg":
        try:
            fhandle.seek(0)  # Read 0xff next

            size = 2
            ftype = 0
            while not 0xC0 <= ftype <= 0xCF:
                fhandle.seek(size, 1)
                byte = fhandle.read(1)
                if byte == b"":
                    fhandle = end
                    byte = fhandle.read(1)

                while ord(byte) == 0xFF:
                    byte = fhandle.read(1)
                ftype = ord(byte)
                size = struct.unpack(">H", fhandle.read(2))[0] - 2
            # We are at a SOFn block
            fhandle.seek(1, 1)  # Skip `precision' byte.
            height, width = struct.unpack(">HH", fhandle.read(4))
        except Exception as e:  # IGNORE:W0703
            raise e
    else:
        return "unknown format {!r}".format(format_)
    return width, height


def independent_markdown2html(markdown):
    return markdown2html(
        markdown,
        ".",
        lambda: None,
        {
            "base64_404_image": ("", (0, 0)),
            "base64_loading_image": ("", (0, 0)),
            "stylesheet": "",
        },
        960,
    )