#!/usr/bin/python3

import email.header
import email.parser
import hashlib
import html
import html.parser
import mailbox
import os
import re
import subprocess
import sys
import tempfile
import urllib.parse

import jinja2
import tinycss

basedir = "."

jenv = jinja2.Environment(
                loader=jinja2.FileSystemLoader(["templates"]),
                autoescape=True,
              )

def get_message_id(msg):
    """
    Extract the message id from a message

    Note that this assumes that there is (at least) one message id. If
    this is not the case, it will raise an exception (currently an
    IndexError, but we may use something more suitable in the future).
    """
    match = re.search(r'<(.*?)>', msg["Message-ID"])
    return match.group(1)


def encode_message_id(msgid):
    encmsgid = re.sub('[^!"$(-.0-9:=@-z|~]', lambda x: "{%02x}" % (ord(x.group(0))), msgid)
    return encmsgid


def decode_rfc2047(s):
    if s is None:
        return None
    r = ""
    for chunk in email.header.decode_header(s):
        if chunk[1]:
            try:
                r += chunk[0].decode(chunk[1])
            except LookupError:
                r += chunk[0].decode("windows-1252")
            except UnicodeDecodeError:
                r += chunk[0].decode("windows-1252")
        elif type(chunk[0]) == bytes:
            r += chunk[0].decode('us-ascii')
        else:
            r += chunk[0]
    return r


def render_message(msg):
    msgtmpl = jenv.get_template("message2.html")
    bodyhtml = render_body(msg)
    context = {
        "msg": msg,
        "message_id": msg["Message-Id"],
        "subject": decode_rfc2047(msg["Subject"]),
        "from": decode_rfc2047(msg["From"]),
        "date": msg["Date"],
        "bodyhtml": bodyhtml,
    }
    msghtml = msgtmpl.render(context)
    return jinja2.Markup(msghtml)


def save_part(msg, disposition):
    content_type = msg.get_content_type()
    extension = {
        "application/octet-stream":                ".bin",
        "text/html":                               ".html",
        "text/x-vcard":                            ".vcf",
        "text/plain":                              ".txt",
        "application/x-gzip":                      ".gz", # more likely tar.gz, but we can't know without looking into it which we ain't
        "image/gif":                               ".gif",
        "text/x-c":                                ".c",
        "application/x-perl":                      ".pl",
        "application/msword":                      ".doc",
        "application/ms-tnef":                     ".ms-tnef",
        "application/x-bzip2":                     ".bz2", # more likely tar.bz2, but we can't know without looking into it which we ain't
        "application/x-shellscript":               ".sh",
        "application/x-java-vm":                   ".bin", # The only instances are mis-labelled
        "image/png":                               ".png",
        "application/pgp-keys":                    ".pgp",
        "application/x-gunzip":                    ".gz", # that sort of makes sense, but not really
        "image/jpeg":                              ".jpg",
        "text/x-python":                           ".py",
        "text/x-java":                             ".java",
        "application/x-sh":                        ".sh",
        "text/x-patch":                            ".patch",
        "text/x-c++src":                           ".c++",
        "application/x-compressed-tar":            ".tar.gz",
        "application/vnd.oasis.opendocument.text": ".odt",
        "text/x-perl":                             ".pl",
        "application/pgp-signature":               ".pgp",
        "image/svg+xml":                           ".svg",
    }[content_type]

    name = msg.get_param("name") or "(data)"
    m = hashlib.sha256()
    payload = msg.get_payload(decode=True)
    m.update(payload)
    filename = m.hexdigest() + extension
    os.makedirs("parts", exist_ok=True)
    with open("parts/" + filename, "wb") as fh:
        fh.write(payload)
    url = "../../parts/" + filename
    if disposition == "_url":
        return url
    else:
        template_name = disposition + "_" + content_type.replace("/", "_") + ".html"
        bodytmpl = jenv.get_template(template_name)
        context = {
            "name": name,
            "url": url,
        }
        bodyhtml = bodytmpl.render(context)
        return bodyhtml


partial_message_cache = {}

def render_body(msg, extra=None):
    def render_text_plain(msg, extra=None):
        # msg.get_charset() doesn't work
        ct_params = dict(msg.get_params() or [])
        charset = ct_params.get("charset", "iso-8859-1")
        format = ct_params.get("format", "fixed")
        if format == "fixed":
            bodytmpl = jenv.get_template("body_text_plain.html")
            partbytes = msg.get_payload(decode=True)
            try:
                parttext = partbytes.decode(charset, errors="replace")
            except LookupError as e:
                # Unknown encoding? Probably win-1252
                print(e, file=sys.stderr)
                parttext = partbytes.decode("windows-1252", errors="replace")

            context = {
                "body": parttext
            }
            return bodytmpl.render(context)
        elif format == "flowed":
            bodytmpl = jenv.get_template("body_text_plain_flowed.html")
            parthtml = TextFlowedPart(msg).as_string()
            context = {
                "body": jinja2.Markup(parthtml),
            }
            return bodytmpl.render(context)
        else:
            raise NotImplementedError()
    def render_multipart_mixed(msg, extra=None):
        parts = msg.get_payload()
        if type(parts) == str:
            # mislabelled, assume text/plain
            return render_text_plain(msg)
        # First, scan for parts with a content-id. A multipart/mixed shouldn't
        # have them, but I've seen them in the wild and it should be harmless
        # to support at least images. We don't want all content types, though,
        # because save_part doesn't support nested parts and I don't want to
        # fully implement what is really just a workaround for buggy software.
        for i, part in enumerate(msg.get_payload()):
            content_id = part.get("Content-Id")
            content_type = part.get_content_type()
            if content_id and content_type.startswith("image/"):
                if extra is None:
                    extra = {}
                extra[content_id] = {
                    "i": i,
                    "part": part,
                    "url": save_part(part, "_url"),
                }
        partshtml = []
        for part in msg.get_payload():
            partshtml.append(render_body(part, extra))
        bodytmpl = jenv.get_template("body_multipart_mixed.html")
        context = {
            "parts": partshtml
        }
        return bodytmpl.render(context)
    def render_multipart_digest(msg, extra=None):
        partshtml = []
        for part in msg.get_payload():
            partshtml.append(render_message(part))
        bodytmpl = jenv.get_template("body_multipart_digest.html")
        context = {
            "parts": partshtml
        }
        return bodytmpl.render(context)
    def render_message_rfc822(msg, extra=None):
        partshtml = []
        for part in msg.get_payload():
            partshtml.append(render_message(part))
        bodytmpl = jenv.get_template("body_message_rfc822.html")
        context = {
            "parts": partshtml
        }
        return bodytmpl.render(context)
    def render_text_html(msg, extra=None):
        htmlpart = HTMLPart(extra)
        ct_params = dict(msg.get_params())
        charset = ct_params.get("charset", "iso-8859-1")
        raw_text = msg.get_payload(decode=True).decode(charset, errors="replace")
        htmlpart.feed(raw_text)
        bodytmpl = jenv.get_template("body_text_html.html")
        context = {
            "body": jinja2.Markup(htmlpart.as_string())
        }
        return bodytmpl.render(context)
    def render_text_enriched(msg, extra=None):
        payload = msg.get_payload(decode=True).decode(msg.get_charset() or "iso-8859-1")
        tepart = TextEnrichedPart(payload)
        bodytmpl = jenv.get_template("body_text_enriched.html")
        context = {
            "body": jinja2.Markup(tepart.as_string())
        }
        return bodytmpl.render(context)
    def render_message_partial(msg, extra=None):
        # Default header for get_param is Content-Type
        whole_msg_id = msg.get_param("id")
        if not whole_msg_id in partial_message_cache:
            # For now we assume that total is present on all parts. This
            # isn't guarantueed, however, and we may need to handle the 
            # case where total is only present on the last part.
            partial_message_cache[whole_msg_id] = [None] * int(msg.get_param("total"))
        payload = msg.get_payload()
        s = payload[0].as_string() # Only one part
        partial_message_cache[whole_msg_id][int(msg.get_param("number"))-1] = s
        if not None in partial_message_cache[whole_msg_id]:
            p = email.parser.Parser()
            whole_msg = p.parsestr("".join(partial_message_cache[whole_msg_id]))
            whole_msg_embedded_id = whole_msg["Message-Id"]
            if not whole_msg_embedded_id:
                whole_msg.add_header("Message-Id", "<" + whole_msg_id + ">")
                whole_msg_embedded_id = whole_msg_id
            archive(whole_msg)
            del partial_message_cache[whole_msg_id]
        return "<p>This is part %d of %d of <a href='../%s/'>%s</a></p>" % (
                        int(msg.get_param("number")),
                        int(msg.get_param("total")),
                        encode_message_id(whole_msg_id),
                        html.escape(whole_msg_id))

    def render_application_octet_stream(msg, extra=None):
        return save_part(msg, "attachment")

    def render_multipart_signed(msg, extra=None):
        content, signature = msg.get_payload()
        with tempfile.NamedTemporaryFile(buffering=0) as content_fh:
            content_fh.write(content.as_bytes())
            with tempfile.NamedTemporaryFile(buffering=0, suffix=".asc") as signature_fh:
                signature_fh.write(signature.get_payload(decode=True))

                r = subprocess.run(["gpg", "--verify", signature_fh.name, content_fh.name],
                                   stdout=subprocess.PIPE,
                                   stderr=subprocess.PIPE,
                                   universal_newlines=True)
                gpgresult = r.stderr
                # XXX - Analyze gpgresult or just use r,returncode?
                gpgstatus = "dubious"

        contenthtml = render_message(content)
        bodytmpl = jenv.get_template("body_multipart_signed.html")
        context = {
            "content": contenthtml,
            "gpgresult": gpgresult,
            "gpgstatus": gpgstatus,
        }
        return bodytmpl.render(context)

    def render_application_pgp(msg, extra=None):
        with tempfile.NamedTemporaryFile(buffering=0) as content_fh:
            content_fh.write(msg.get_payload(decode=True))
            r = subprocess.run(["gpg", "--decrypt", content_fh.name],
                               stdout=subprocess.PIPE,
                               stderr=subprocess.PIPE,
                              )
            gpgresult = r.stderr.decode()
            # XXX - Analyze gpgresult or just use r,returncode?
            gpgstatus = "dubious"
            decrypted_content = r.stdout

        p = email.parser.BytesParser()
        embedded_message = p.parsebytes(decrypted_content)
        contenthtml = render_message(embedded_message)
        bodytmpl = jenv.get_template("body_application_pgp.html")
        context = {
            "content": contenthtml,
            "gpgresult": gpgresult,
            "gpgstatus": gpgstatus,
        }
        return bodytmpl.render(context)

    def render_multipart_alternative(msg, extra=None):
        partshtml = []
        partstypes = []
        for part in msg.get_payload():
            partstypes.append(part.get_content_type())
            partshtml.append(render_body(part, extra))
        bodytmpl = jenv.get_template("body_multipart_alternative.html")
        context = {
            "types": partstypes,
            "parts": partshtml,
        }
        return bodytmpl.render(context)

    def render_application_x_unknown_content_type_scpfile(msg, extra=None):
        bodytmpl = jenv.get_template("body_application_x-unknown-content-type-scpfile.html")
        context = {
            "body": msg.get_payload(decode=True).decode(msg.get_charset() or "iso-8859-1")
        }
        return bodytmpl.render(context)

    def render_application_pgp_signature(msg, extra=None):
        # A PGP signature outside of a multipart/signed - useless
        bodytmpl = jenv.get_template("body_application_pgp-signature.html")
        context = {
        }
        return bodytmpl.render(context)

    def render_application_x_gzip(msg, extra=None):
        return save_part(msg, "attachment")

    def render_message_news(msg, extra=None):
        partshtml = []
        for part in msg.get_payload():
            partshtml.append(render_message(part))
        bodytmpl = jenv.get_template("body_message_news.html")
        context = {
            "msg": msg,
            "parts": partshtml,
        }
        return bodytmpl.render(context)

    def render_image_gif(msg, extra=None):
        return save_part(msg, "inline")

    def render_multipart_related(msg, extra=None):
        start = msg.get_param("start")
        start_part = None
        # collect content-ids
        content = {}
        for i, part in enumerate(msg.get_payload()):
            content_id = part.get("Content-Id")
            if start_part is None and (start is None or content_id == start):
                start_part = part
                continue
            if content_id:
                content[content_id] = {
                    "i": i,
                    "part": part,
                    "url": save_part(part, "_url"),
                }



        parthtml = render_body(start_part, content)
        bodytmpl = jenv.get_template("body_multipart_related.html")
        context = {
            "msg": msg,
            "parts": [parthtml],
        }
        return bodytmpl.render(context)

    def render_image_jpeg(msg, extra=None):
        return save_part(msg, "inline")

    def render_message_delivery_status(msg, extra=None):
        bodytmpl = jenv.get_template("body_message_delivery_status.html")
        # A message/delivery status consists of one per-message block
        # followed by one or more per-recipient blocks.
        # Pythons message parser apparently parses each block as a message
        # consisting only of headers. So we just stringify and concatenate them
        parts = msg.get_payload()
        parttext = "".join([str(p) for p in parts])
        context = {
            "body": parttext
        }
        return bodytmpl.render(context)

    renderers = {
        "text/plain": render_text_plain,
        "multipart/mixed": render_multipart_mixed,
        "multipart/digest": render_multipart_digest,
        "message/rfc822": render_message_rfc822,
        "text/html": render_text_html,
        "text/enriched": render_text_enriched,
        "message/partial": render_message_partial,
        "application/octet-stream": render_application_octet_stream,
        "multipart/signed": render_multipart_signed,
        "application/pgp": render_application_pgp,
        "multipart/alternative": render_multipart_alternative,
        "application/x-unknown-content-type-scpfile": render_application_x_unknown_content_type_scpfile,
        "application/pgp-signature": render_application_pgp_signature,
        "application/x-gzip": render_application_x_gzip,
        "message/news": render_message_news,
        "image/gif": render_image_gif,
        "multipart/related": render_multipart_related,
        "application/x-java-vm": render_application_octet_stream,
        "image/jpeg": render_image_jpeg,
        "application/x-compressed-tar": render_application_octet_stream,
        "message/delivery-status": render_message_delivery_status,
        "application/pgp-keys": render_application_octet_stream,
    }
    content_type = msg.get_content_type()
    content_disposition = msg.get_content_disposition()
    if content_disposition == "attachment":
        # XXX - not sure, if we should just store all content-types. 
        # We probably should clean up html. Alternatively we could just store
        # all of them application/octet-stream, which browsers should download
        # and not try to display.
        bodyhtml = save_part(msg, content_disposition)
    else:
        bodyhtml = renderers[content_type](msg, extra)

    return jinja2.Markup(bodyhtml)


def archive(msg):
    mid = get_message_id(msg)
    print("M", mid, file=sys.stderr)
    encmid = encode_message_id(mid)
    msgdir = basedir + "/msg/" + encmid
    os.makedirs(msgdir, exist_ok=True)
    with open(msgdir + "/index.html", "w") as hfd:
        msgtmpl = jenv.get_template("message.html")
        bodyhtml = render_body(msg)
        context = {
            "list": "LUGA",
            "message_id": mid,
            "subject": decode_rfc2047(msg["Subject"]),
            "from": decode_rfc2047(msg["From"]),
            "date": msg["Date"],
            "bodyhtml": bodyhtml,
        }
        msghtml = msgtmpl.render(context)
        hfd.write(msghtml)


class HTMLPart(html.parser.HTMLParser):
    """
    A text/html part

    This is a subclass of HTMLParser, so the handle_* methods will be invoked
    as appropriate during parsing. There are a few additional attributes to
    keep track of the state:

    ... attribute:: content

        Accumulates parts fragments of the final, cleaned up, html message as
        strings

    ... attribute:: base

        The base URL

    ... attribute:: extra

        Context information. This includes info about cids or references to
        other messages

    ... attribute:: hide

        If true, the content of the current tag is omitted from the output.
        This is set when encountering a start tag in hide_tags, and reset at
        each end tag (so it works only for leaves).

    ... attribute:: current_tag

        The current tag. Similar to hide, this is set and reset when
        encountering start end end tags, so it is only correct while processing
        a leaf element. But since we use it only for style elements, that's
        acceptable.
    """
    allowed_tags = [
        "h2", "a", "wbr", "hr", "pre", "img", "font", "i", "br", "table", "tr",
        "th", "td", "b", "select", "option", "input", "sup", "address",
        "center", "p", "h1", "dl", "h3", "ul", "li", "ol", "u", "blockquote",
        "h4", "div", "span", "style",
    ]
    hide_tags = [ "title" ]
    ignore_tags = [ "html", "head", "body", "marquee", "meta", "form", ]

    def __init__(self, extra):
        super().__init__()
        self.hide = False
        self.content = []
        self.base = None
        self.extra = extra or {}
        self.current_tag = None

    def handle_starttag(self, tag, attrs):
        self.current_tag = tag
        if tag == "base":
            href = [x[1] for x in attrs if x[0] == "href"]
            if href:
                self.base = href[0]
        elif tag in self.allowed_tags:
            cleaned_attrs, extra = self.clean_attrs(tag, attrs)
            attrstr = "".join(
                            [' %s="%s"' % (a[0],  html.escape(a[1])) if a[1] else ' %s' % (a[0])
                                for a in cleaned_attrs
                            ]
                         )
            self.content.append("<%s%s>" % ( tag, attrstr ))
            if extra:
                self.content.append(extra)
        elif tag in self.hide_tags:
            self.hide = True
        elif tag in self.ignore_tags:
            pass
        else:
            print("Encountered unknown start tag", tag, attrs, file=sys.stderr)

    def handle_endtag(self, tag):
        if tag in self.allowed_tags:
            self.content.append("</%s>" % tag)
        elif tag in self.hide_tags:
            self.hide = False # XXX - Need stack?
        elif tag in self.ignore_tags:
            pass
        else:
            print("Encountered unknown end tag", tag, file=sys.stderr)
        self.current_tag = None

    def handle_data(self, data):
        if self.current_tag == "style":
            data = self.clean_style(data)
        if not self.hide:
            self.content.append(data)

    def as_string(self):
        return "".join(self.content)

    def clean_attrs(self, tag, attrs):
        safe_attrs = [
            "border", "alt", "size", "face", "width", "height", "hspace",
            "cellpadding", "cellspacing", "bgcolor", "valign", "nowrap",
            "color", "colspan", "name", "value", "type", "align", "clear",
            "noshade", "type",
        ]
        clean_attrs = []
        extra = None
        for a in attrs:
            if a[0] in safe_attrs:
                clean_attrs.append(a)
            elif a[0] == "href":
                url = a[1]
                url = urllib.parse.urljoin(self.base, url)
                u = urllib.parse.urlparse(url)
                if u[0] in ['https', 'http', 'ftp']:
                    clean_attrs.append((a[0], url))
            elif a[0] == "src":
                url = a[1]
                url = urllib.parse.urljoin(self.base, url)
                u = urllib.parse.urlparse(url)
                if u[0] == "cid":
                    print("Encountered src cid attribute", a, file=sys.stderr)
                    clean_attrs.append((a[0], self.extra["<" + u.path + ">"]["url"]))
                else:
                    print("Ignored src attribute", a, file=sys.stderr)
            elif a[0] == "target":
                pass
            elif a[0] == "cite":
                if a[1].startswith("mid:"):
                    mid = a[1][4:]
                    encmid = encode_message_id(mid)
                    extra = "<a class='citesource' href='../%s'>\u2397</a>" % encmid
            elif a[0] == "class":
                clean_attrs.append((a[0], "msg-" + a[1],))
            else:
                print("Encountered unknown attribute", a, "in", tag, file=sys.stderr)
        return clean_attrs, extra

    def clean_style(self, stylesheet):
        cssparser = tinycss.make_parser()
        stylesheet = cssparser.parse_stylesheet(stylesheet)
        clean_stylesheet = ""

        for rule in stylesheet.rules:

            # first clean up selectors: Prepend "msg-" to every class or id
            next_is_local_id = False
            new_selector = []
            for token in rule.selector:
                if next_is_local_id and token.type == "IDENT":
                    new_id = "msg-" + token.value
                    new_selector.append(tinycss.token_data.Token(token.type, new_id, new_id, token.unit, token.line, token.column))
                else:
                    new_selector.append(token)
                next_is_local_id = token.type == "DELIM" and (token.value == "." or token.value == "#")
            rule.selector = tinycss.token_data.TokenList(new_selector)
            clean_stylesheet += rule.selector.as_css()

            # Then clean up declarations.
            # We keep only declarations we recognize
            safe_declarations = {
                "background-color",
                "border-bottom-style",
                "border-color",
                "border-left",
                "border-left-style",
                "border-right-style",
                "border-style",
                "border-top-style",
                "color",
                "font-family",
                "font-size",
                "font-style",
                "font-weight",
                "height",
                "list-style",
                "margin",
                "margin-bottom",
                "margin-left",
                "margin-right",
                "margin-top",
                "padding",
                "padding-bottom",
                "padding-left",
                "padding-right",
                "padding-top",
                "page-break-after",
                "text-align",
                "text-decoration",
                "white-space",
                "width",
            }

            # Ignore these silently to avoid cluttering logs
            ignore_declarations = {
                "mso-ansi-font-size",
                "mso-ansi-language",
                "mso-ascii-font-family",
                "mso-bidi-font-family",
                "mso-bidi-font-size",
                "mso-bidi-font-weight",
                "mso-fareast-font-family",
                "mso-fareast-language",
                "mso-gram-e",
                "mso-hansi-font-family",
                "mso-margin-bottom-alt",
                "mso-margin-top-alt",
                "mso-outline-level",
                "mso-pagination",
                "mso-spl-e",
                "mso-style-link",
                "mso-style-name",
                "mso-style-next",
                "mso-style-noshow",
                "mso-style-parent",
                "mso-style-priority",
                "mso-style-type",
                "page", # doesn't exist in CSS 2.2
                "panose-1", # doesn't exist in CSS 2.2
                "text-underline", # doesn't exist in CSS 2.2
            }
            new_declarations = []
            for declaration in rule.declarations:
                if declaration.name in safe_declarations:
                    new_declarations.append(declaration)
                elif declaration.name == "background-image":
                    # check if URL is cid, discard if not
                    ok = False
                    if len(declaration.value) == 1:
                        if declaration.value[0].type == "URI":
                            if declaration.value[0].value.startswith("cid:"):
                                print("accepting url", declaration.value[0].value)
                                # Get the real converted url here
                                new_declarations.append(
                                    parser.parse_style_attr("background-image: url(/whatever.png)")[0][0]
                                )
                                ok = True
                    if not ok:
                        print("ignoring unsafe CSS property", declaration)
                    pass
                elif declaration.name in ignore_declarations:
                    pass
                else:
                    print("ignoring unknown CSS property", declaration.name)
            clean_stylesheet += " {\n"
            for declaration in new_declarations:
                clean_stylesheet += "\t" +  declaration.name + ":" + declaration.value.as_css() + ";\n"
            clean_stylesheet += "}\n\n"
        return clean_stylesheet


class TextEnrichedPart:
    class TEElement:
        def __init__(self, t, parent):
            self.type = t.lower()
            self.content = []
            if self.type == "nofill":
                self.filled = False
            elif parent:
                self.filled = parent.filled
            else:
                self.filled = True
        
        def append_text(self, s):
            s = s.replace("<<", "<")
            if self.filled:
                s = re.sub(r'\n+',
                           lambda m: m.group(0)[1:] if len(m.group(0)) > 1 else " ",
                           s)
            self.content.append(s)

        def as_string(self):
            if self.type == "":
                pre = "<div class='text-enriched'>"
                post = "</div>"
            elif self.type == "bold":
                pre = "<b>"
                post = "</b>"
            elif self.type == "param":
                # We shouldn't ever get here since the param should be consumed
                # by the parent, but there are broken messages ...
                return ""
            elif self.type.startswith("x-"):
                # Just ignore all experimental elements and render their
                # contents.
                pre = ""
                post = ""
            elif self.type == "flushleft":
                pre = "<div class='flushleft'>"
                post = "</div>"
            elif self.type == "smaller":
                # HTML has a "small" element, but that is meant for "side
                # comments such as small print", while t/e "smaller" is purely
                # typographical
                pre = "<span style='font-size: 0.9em'>"
                post = "</span>"
            elif self.type == "color":
                param = self.content.pop(0)
                if param.type != "param":
                    raise RuntimeError("Expected 'param', got '%s'" % param.type)
                colorstring = param.content[0]
                if re.match(r'^\w+$', colorstring):
                    # a single word, i.e. a colorname like "red" or cyan". 
                    # The 8 colors in the spec aren't a subset of the 17 colors in CSS2,
                    # but recognized by most/all browsers. And if we encounter a non-standard
                    # color the best we can do is let the browser handle it.
                    pass
                else:
                    m = re.match(r'([0-9a-f]{4}),([0-9a-f]{4}),([0-9a-f]{4})', colorstring, re.IGNORECASE)
                    if m:
                        # an RGB triple. Use only the top 8 bits of each component:
                        colorstring = "#%s%s%s" % (m.group(1)[:2], m.group(2)[:2], m.group(3)[:2])
                    else:
                        # syntax error. Replace with "black"
                        colorstring = "#000"
                pre = "<span style='color: %s'>" % colorstring
                post = "</span>"
            elif self.type == "nofill":
                pre = "<div class='nofill'>"
                post = "</div>"
            elif self.type == "fontfamily":
                param = self.content.pop(0)
                if param.type != "param":
                    raise RuntimeError("Expected 'param', got '%s'" % param.type)
                fontfamily = param.content[0]
                if "'" in fontfamily or '"' in fontfamily:
                    raise RuntimeError("Can't handle quotes in font names (%s)" % fontfamily)
                pre = "<span style='font-family: \"%s\"'>" % fontfamily
                post = "</span>"
            elif self.type == "bigger":
                # HTML used to have a "big" element, but that has been removed from HTML5
                pre = "<span style='font-size: 1.1em'>"
                post = "</span>"
            elif self.type == "underline":
                # HTML5 redefined the meaning of "u", but I'm using it anyway
                pre = "<u>"
                post = "</u>"
            else:
                raise NotImplementedError("Unknown type " + self.type)

            s = pre
            for c in self.content:
                if isinstance(c, type(self)):
                    s += c.as_string()
                else:
                    s += html.escape(c)
            s += post
            return s

    def __init__(self, s):
        self.stack = [ self.TEElement("", None) ]
        while s:
            stack_top = self.stack[-1]
            m = re.match(r'(.*?)<(/?[A-Za-z0-9-]{,60})>(.*)', s, re.DOTALL)
            if m:
                text = m.group(1)
                tag = m.group(2).lower()
                if not (tag == "param" and re.match(r'\s*', text) or text == ""):
                    stack_top.append_text(text)
                if tag[0] != "/":
                    new = self.TEElement(tag, stack_top)
                    stack_top.content.append(new)
                    self.stack.append(new)
                else:
                    closed_tag = tag[1:]
                    if stack_top.type == closed_tag:
                        self.stack.pop()
                    elif closed_tag in [e.type for e in self.stack]:
                        # We close a tag which has been opened, but it
                        # wasn't the last one. This is clearly a nesting
                        # error, but there was broken software (e.g.
                        # http://www.fozztexx.com/Mynah/) which used
                        # non-closing tags, and by just popping them off
                        # the stack we can "re-synchronize".
                        while self.stack.pop().type != closed_tag:
                            pass
                    else:
                        raise RuntimeError("Nesting error: Expected %s, got %s near %s" % (self.stack[-1].type, closed_tag, s))
                s = m.group(3)
            else:
                stack_top.append_text(s)
                s = ""

    def as_string(self):
        return self.stack[0].as_string()


class TextFlowedPart:
    def __init__(self, msg):
        self.quote_depth = 0
        self.current_line = ""
        self.flowed = False
        self.lines = []
        self.buffer_filled = False

        ct_params = dict(msg.get_params())
        charset = ct_params.get("charset", "iso-8859-1")
        format = ct_params.get("format", "fixed")
        delsp = ct_params.get("delsp", "no") == "yes"
        charset_map = {
            "x-mac-roman": "mac_roman",
        }
        if charset in charset_map:
            charset = charset_map[charset]
        raw_text = msg.get_payload(decode=True).decode(charset, errors="replace")
        raw_lines = raw_text.split("\n")

        for rl in raw_lines:
            quote_depth = 0
            while rl[:1] == ">":
                quote_depth += 1
                rl = rl[1:]
            if rl[:1] == " ":
                rl = rl[1:]
            if rl == "-- ":
                flowed = None
            elif rl[-1:] == " ":
                flowed = True
                if delsp:
                    rl = rl[:-1]
            else:
                flowed = False
            self.add_buffer(rl, quote_depth, flowed)
        self.flush()

    def add_buffer(self, line, quote_depth, flowed):
        if flowed is None:
            self.flush()
            flowed = False
        if quote_depth != self.quote_depth:
            self.flush()
            self.quote_depth = quote_depth
        self.current_line += line
        self.flowed |= flowed
        self.buffer_filled = True
        if not flowed:
            self.flush()

    def flush(self):
        if self.buffer_filled:
            self.lines.append({
                            "quote_depth": self.quote_depth,
                            "flowed": self.flowed,
                            "content": self.current_line
                        })
        self.current_line = ""
        self.flowed = False
        self.buffer_filled = False

    def as_string(self):
        prev_quote_depth = 0
        s = ""
        for ln in self.lines:
            while ln["quote_depth"] > prev_quote_depth:
                s += "<blockquote>\n"
                prev_quote_depth += 1
            while ln["quote_depth"] < prev_quote_depth:
                s += "</blockquote>\n"
                prev_quote_depth -= 1
            if ln["flowed"]:
                s += "<p class='flowed'>" + html.escape(ln["content"]) + "</p>\n"
            else:
                s += "<p class='fixed'>" + html.escape(ln["content"]) + "</p>\n"
        while 0 < prev_quote_depth:
            s += "</blockquote>"
            prev_quote_depth -= 1
        return s






for f in sys.argv[1:]:
    print("F", f, file=sys.stderr)
    mb = mailbox.mbox(f)

    for m in mb:
        archive(m)

# vim: tw=79