#!/usr/bin/python
#
# Being a modest reimplementation of mhn -show in Python.
# We are invoked as '<prog> <messages> -show', more or less; the messages
# are always in the current folder. The usual case is a single message.
#
# BUGS: boy do we hardcode things.
#
import sys, os, commands, re
import getopt
import email, email.Message, email.Errors, email.Iterators, email.Parser

class mhnError(Exception):
	pass
progname = "mhnshow"

# specifying the absolute path makes cks happy in obscure situations,
# and since this all about cks he's happy with the potential lossage.
nukehdrsProg = "/u/cks/bin/mh/nukehdrs"
prefText = 1
demoronise = 1
doHtmlRender = 1
wrapLongText = 0

class MyMsg(email.Message.Message):
	def __init__(self):
		self.hdrs = None
		email.Message.Message.__init__(self)
	def sethdrs(self, hdrs):
		self.hdrs = hdrs

# Given a list of message numbers, return full paths.
# commands.getoutput helpfully kills the final newline for us, so we only
# care about interior ones.
def mhpath(msglst):
	return commands.getoutput("mhpath %s" % " ".join(msglst)).split('\n')

# This function loads an email message. Since mhn itself pukes on invalid
# mail messages, we are no worse off than we used to be if this explodes
# in our face (and better off in the case of Python 2.4).
def loadmsg(fn):
	hdrs = []
	try:
		fp = open(fn, "r")
		while 1:
			l = fp.readline()
			if not l or l == '\n':
				break
			# special bonus ZMailer hack; skip envelope headers
			# entirely.
			sl = l.split()
			if not sl or \
			   sl[0] in ("from", "to", "external", "rcvdfrom",
				     "with"):
				continue
			hdrs.append(l)
		# The REAL headers are now, unaltered, in hdrs. Glue the
		# rest in.
		body = fp.read()
		hdrs = "".join(hdrs)
		prs = email.Parser.Parser(MyMsg)
		m = prs.parsestr("%s\n%s" % (hdrs, body))
		m.sethdrs(hdrs)
		return m
		#return email.message_from_file(fp)
	except EnvironmentError, e:
		raise mhnError, "cannot read in %s: %s" % (fn, str(e))
	except email.Errors.MessageParseError, e:
		raise mhnError, "cannot parse %s: %s" (fn, str(e))

# Get the decoded contents of a given part.
def getpartcont(p):
	t = None
	try:
		t = p.get_payload(None, True)
	except:
		pass
	if t == None:
		t = p.as_string()
	if t == None:
		t = ""
	return t

# This goes into the output stream:
def complain(msg):
	sys.stdout.write("!! %s: %s\n" % (progname, msg))
def note(msg):
	sys.stdout.write("++ %s: %s\n" % (progname, msg))

# Demoronize content if desired.
charPairs = (('\x82', ','),
	     ('\x84', ',,'),
	     ('\x85', '...'),
	     ('\x88', '^'),
	     ('\x8b', '<'),
	     ('\x8c', 'Oe'),
	     ('\x91', "`"),
	     ('\x92', "'"),
	     ('\x93', '"'),
	     ('\x94', '"'),
	     ('\x95', '*'),
	     ('\x96', '-'),
	     ('\x97', '--'),
	     ('\x9b', '>'),
	     ('\x9c', 'oe'),
	     # These are overridden by HTML specific translations later.
	     # (or will be when we have that)
	     ('\x83', 'f'),
	     ('\x98', '~'),
	     ('\x99', '(tm)'),
	     )
def demoroniser(buf):
	if not demoronise:
		return buf
	for o, n in charPairs:
		buf = buf.replace(o, n)
	return buf

# Handle displaying various sorts of content types.
# We make gratuitously ugly assumptions, like 'the current terminal
# can display all of these character sets', which are perhaps a bit
# laughable.
def showhtml(p):
	c = demoroniser(getpartcont(p))
	if not doHtmlRender:
		sys.stdout.write(c)
		return
	try:
		sys.stdout.flush()
		fp = os.popen("lynx -dump -force_html /dev/stdin", "w")
		fp.write(c)
		fp.close()
	except EnvironmentError, e:
		raise mhnError, "error dumping HTML contents: %s" % str(e)

# Optionally, we wrap overly long lines in text blocks. We do this in the
# same way 'fmt' does, which is to say that we never reflow blocks.
LONGLINE = 100
LLRATIO = 0.3
def countbiglines(c):
	cl = c.split("\n")
	ll = 0; tl = 0
	for l in cl:
		if len(l) >= LONGLINE:
			ll += len(l)
		tl += len(l)
	return (tl, ll, cl)
# initial whitespace, non-whitespace, and trailing whitespace, all
# elements optional.
wsre = re.compile("^([ \t]*)([^ \t].*)\s*$")
def fmtline(ln, wlen):
	# The 'display length' of a string of characters, counting tabs.
	def displen(str):
		pos = 0
		for c in str:
			if c == '\t':
				pos = (pos // 8)*8 + 8
			else:
				pos += 1
		return pos
	# Find the span of initial whitespace, if any.
	mr = wsre.search(ln)
	# failure to match means that we have no non-whitespace, because
	# it's the only required element. In that case we return a blank
	# line.
	if not mr:
		return ['']
	ws = mr.group(1)
	nws = mr.group(2)
	# If the length of the initial whitespace is guaranteed to wrap
	# no matter what, we punt the line (minus trailing space) back
	# as is.
	dl = displen(ws)
	if dl >= wlen:
		return [ws+nws]
	# Otherwise, we split to words and then wrap the words.
	wds = nws.split()
	lines = []; curline = []; curlen = dl
	for wd in wds:
		if curline and (curlen + len(wd) + 1 > wlen):
			lines.append(ws + " ".join(curline))
			curline = []; curlen = dl
		curline.append(wd)
		curlen += len(wd) + 1
	if curline:
		lines.append(ws + " ".join(curline))
	return lines
def fmtwrap(larray, wlen):
	na = []
	for l in larray:
		if len(l) < wlen:
			na.append(l)
		else:
			na.extend(fmtline(l, wlen))
	return "\n".join(na)
def showplain(p):
	c = demoroniser(getpartcont(p))
	if wrapLongText:
		(tot, lng, splt) = countbiglines(c)
		# We base on a) ratio or b) absolute characters in long
		# lines, because I have a limited tolerance for the latter.
		if tot and (lng > (2*1024) or (float(lng) / tot) > LLRATIO) \
		   and lng > 150:
			c = fmtwrap(splt, 77)
	# trim surplus trailing newlines: we want exactly one blank line
	# at the end.
	while c and c.endswith("\n\n"):
		c = c[:-1]
	sys.stdout.write(c)
	sys.stdout.write("\n")

# this is nominally plaintext, except that it pisses us off in a clever
# way from email.*, because email.* manages to bodge it up and we have
# to correct their bodge. (Perhaps this correction should take place
# elsewhere. Perhaps.)
def showdelstat(p):
	c = getpartcont(p)
	l = c.split("\n", 2)
	if len(l) == 3 and l[1] == "" and \
	   l[0].lower() == "content-type: message/delivery-status":
		c = l[2]
	sys.stdout.write(c)
	sys.stdout.write("\n")

# Handle multipart/alternative in the approved manner: we pick the best
# (ie last) content that we can handle.
def showalternative(p):
	m = p.get_payload()
	# If we prefer text/plain to text/html and this is a text/plain +
	# text/html message, we pick the text/plain version.
	if prefText and len(m) == 2 \
	   and m[0].get_content_type() == "text/plain" \
	   and m[1].get_content_type() == "text/html":
		noteshowpart(m[0])
		return
	m.reverse()
	for alt in m:
		if alt.get_content_type() in knownParts:
			noteshowpart(alt)
			return
	# Nothing we know how to handle.
	complain("cannot display multipart/alternative composed of: %s" %
		 (", ".join([x.get_content_type() for x in m])))
def showmulti(p):
	# It can happen that this returns a string, not payload.
	# This seems to happen for broken MIME messages; the case I have
	# seen is a mail system that replaced the body of a multipart/mixed
	# message in a bounce with '(Body supressed)'.
	t = p.get_payload()
	if isinstance(t, str):
		sys.stdout.write(t)
		return
	for sp in p.get_payload():
		noteshowpart(sp)

# Our over-all decision maker about what to do about a particular part.
def prepost(str):
	l = str.strip()
	if not l:
		return
	if l == 'This is a multi-part message in MIME format.' or \
	   l == 'This is a MIME-encapsulated message':
		return
	sys.stdout.write(str)
def showpart(p):
	ct = p.get_content_type()
	cm = p.get_content_maintype()
	if p.preamble:
		prepost(p.preamble)
	if ct in knownParts:
		knownParts[ct](p)
	elif cm in knownParts:
		knownParts[cm](p)
	else:
		if ct[0] in ('a', 'e', 'i', 'o', 'u'):
			ana = "an"
		else:
			ana = "a"
		note("skipping %s %s part" % (ana, ct))
	if p.epilogue:
		prepost(p.epilogue)
# We don't show notes about parts that showpart will only produce a
# complaint about.
def noteshowpart(p):
	ct = p.get_content_type(); cm = p.get_content_maintype()
	if ct in knownParts or cm in knownParts or p.preamble or p.epilogue:
		sys.stdout.write("++ part %s:\n" % ct)
	showpart(p)

# Show an RFC822 message. We must deal specially with the headers, and
# then we dump the body.
def showmsg(p):
	if hasattr(p, "hdrs") and p.hdrs:
		htxt = p.hdrs
	else:
		# Unnatural intimacy, but we do what we can.
		# (we would actually like to do our own header parsing
		# instead of email.Parser.Parser()'s, so we can save the
		# raw information, but that's too big a project for version
		# 0.1)
		htxt = "\n".join(["%s:\t%s" % (x[0], x[1]) for x in p._headers])
		htxt += "\n"
	try:
		sys.stdout.flush()
		fp = os.popen(nukehdrsProg, "w")
		fp.write(htxt); fp.write("\n")
		fp.close()
	except EnvironmentError, e:
		raise mhnError, "error writing message headers: %s" % str(e)
	if p.is_multipart() and len(p.get_payload()) > 1:
		noteshowpart(p)
	else:
		showpart(p)

# A message/rfc822 object is merely a thin container; the actual *message*
# is inside it as its sole content. At least we hope it's the sole content!
def showrfc822(p):
	pl = p.get_payload()
	if len(pl) != 1:
		complain("message/rfc822 with multiple contents!")
		complain("types: %s" %
			 ", ".join([x.get_content_type() for x in pl]))
	for pe in pl:
		showmsg(pe)
	# and we're done.

# application/pgp is merely the signature portion, so we don't need to
# mention it.
knownParts = {
	'text/plain': showplain, 'text/html': showhtml,
	# this notice is actually plaintext.
	'message/delivery-status': showdelstat,
	# we show these plain instead of running them through nukehdrs
	# because they're usually in bounces of spam and we want to see the
	# full headers to see, eg, Received:.
	'text/rfc822-headers': showplain,
	# the parcels:
	'multipart/alternative': showalternative,
	'message/rfc822': showrfc822,
	# We think of these as generic things in which we spew out anything
	# we can recognize.
	'message/digest': showmulti, 'multipart': showmulti,
	}

# Actually do stuff.
def warn(msg):
	sys.stderr.write("%s: %s\n" % (progname, msg))
def die(msg):
	warn(msg)
	sys.exit(1)
# process handles one file (not message; we have turned message numbers
# into files before now).
def process(fn):
	try:
		showmsg(loadmsg(fn))
	except EnvironmentError, e:
		die("showing %s: %s" % (fn, str(e)))
	except mhnError, e:
		die("showing %s: %s" % (fn, str(e)))

# Handle the arguments. If our argument list ends in -show, it is a list
# of MH message numbers, and we mhpath it; if it doesn't, it is a list of
# filenames and we just use them directly.
def usage():
	sys.stderr.write("usage: %s [-N prog] [-HDsw] arg [... arg] [-show]\n")
	sys.exit(2)
def parseargs(sargs):
	global nukehdrsProg; global prefText; global demoronise
	global doHtmlRender; global wrapLongText
	try:
		opts, args = getopt.getopt(sargs, "N:HDsw", [])
	except getopt.error, cause:
		warn(cause)
		usage()
	for o, a in opts:
		if o == '-N':
			nukehdrsProg = a
		elif o == '-H':
			prefText = 0
		elif o == '-D':
			demoronise = 0
		elif o == '-s':
			doHtmlRender = 0
		elif o == '-w':
			wrapLongText = 1
		else:
			die("Chris failed to handle option '%s'" % o)
	return args
# We accept options from either $MHNSHOWOPTS or from the command line,
# because when we're in the MH context we can't pass command line arguments
# through to mhnshow any other way. Because -show shows up in a funky place
# we remove it from the arglist before we run the thing past getopt().
def main(args):
	mhnmode = 0
	if args[-1] == "-show":
		mhnmode = 1
		args = args[:-1]
	# Take options from the environment first.
	envopts = os.getenv("MHNSHOWOPTS")
	if envopts:
		parseargs(envopts.split())
	# Then take command-line arguments.
	files = parseargs(args)
	if not files:
		warn("No arguments supplied.")
		usage()
	# Now process.
	if mhnmode:
		files = mhpath(files)
	for fn in files:
		process(fn)

if __name__ == "__main__":
	main(sys.argv[1:])