Continue with quotation cut even if html cut throws an exception

Add __init__.py into data folder, add data files into MANIFEST.in

.gitignore

@@ -39,6 +39,8 @@ nosetests.xml
 /.emacs.desktop
 /.emacs.desktop.lock
 .elc
+.idea
+.cache
 auto-save-list
 tramp
 .\#*
@@ -51,4 +53,4 @@ tramp
 _trial_temp
 
 # OSX
-.DS_Store
+.DS_Store

MANIFEST.in

@@ -5,3 +5,10 @@ include classifier
 include LICENSE
 include MANIFEST.in
 include README.rst
+include talon/signature/data/train.data
+include talon/signature/data/classifier
+include talon/signature/data/classifier_01.npy
+include talon/signature/data/classifier_02.npy
+include talon/signature/data/classifier_03.npy
+include talon/signature/data/classifier_04.npy
+include talon/signature/data/classifier_05.npy

README.rst

@@ -129,6 +129,22 @@ start using it for talon.
 .. _EDRM: http://www.edrm.net/resources/data-sets/edrm-enron-email-data-set
 .. _forge: https://github.com/mailgun/forge
 
+Training on your dataset
+------------------------
+
+talon comes with a pre-processed dataset and a pre-trained classifier. To retrain the classifier on your own dataset of raw emails, structure and annotate them in the same way the `forge`_ project does. Then do:
+
+.. code:: python
+
+    from talon.signature.learning.dataset import build_extraction_dataset
+    from talon.signature.learning import classifier as c 
+    
+    build_extraction_dataset("/path/to/your/P/folder", "/path/to/talon/signature/data/train.data")
+    c.train(c.init(), "/path/to/talon/signature/data/train.data", "/path/to/talon/signature/data/classifier")
+
+Note that for signature extraction you need just the folder with the positive samples with annotated signature lines (P folder).
+
+.. _forge: https://github.com/mailgun/forge
 
 Research
 --------

setup.py

@@ -29,7 +29,7 @@ class InstallCommand(install):
 
 
 setup(name='talon',
-      version='1.3.7',
+      version='1.4.8',
       description=("Mailgun library "
                    "to extract message quotations and signatures."),
       long_description=open("README.rst").read(),
@@ -48,7 +48,7 @@ setup(name='talon',
           "regex>=1",
           "numpy",
           "scipy",
-          "scikit-learn>=0.16.1", # pickled versions of classifier, else rebuild
+          "scikit-learn==0.16.1", # pickled versions of classifier, else rebuild
           'chardet>=1.0.1',
           'cchardet>=0.3.5',
           'cssselect',

talon/html_quotations.py

@@ -87,17 +87,24 @@ def cut_gmail_quote(html_message):
 
 def cut_microsoft_quote(html_message):
     ''' Cuts splitter block and all following blocks. '''
+    #use EXSLT extensions to have a regex match() function with lxml
+    ns = {"re": "http://exslt.org/regular-expressions"}
+
+    #general pattern: @style='border:none;border-top:solid <color> 1.0pt;padding:3.0pt 0<unit> 0<unit> 0<unit>'
+    #outlook 2007, 2010 (international) <color=#B5C4DF> <unit=cm>
+    #outlook 2007, 2010 (american)      <color=#B5C4DF> <unit=pt>
+    #outlook 2013       (international) <color=#E1E1E1> <unit=cm>
+    #outlook 2013       (american)      <color=#E1E1E1> <unit=pt>
+    #also handles a variant with a space after the semicolon
     splitter = html_message.xpath(
-        #outlook 2007, 2010 (international)
-        "//div[@style='border:none;border-top:solid #B5C4DF 1.0pt;"
-        "padding:3.0pt 0cm 0cm 0cm']|"
-        #outlook 2007, 2010 (american)
-        "//div[@style='border:none;border-top:solid #B5C4DF 1.0pt;"
-        "padding:3.0pt 0in 0in 0in']|"
+        #outlook 2007, 2010, 2013 (international, american)
+        "//div[@style[re:match(., 'border:none; ?border-top:solid #(E1E1E1|B5C4DF) 1.0pt; ?"
+        "padding:3.0pt 0(in|cm) 0(in|cm) 0(in|cm)')]]|"
         #windows mail
         "//div[@style='padding-top: 5px; "
         "border-top-color: rgb(229, 229, 229); "
         "border-top-width: 1px; border-top-style: solid;']"
+        , namespaces=ns
     )
 
     if splitter:

talon/quotations.py

@@ -22,7 +22,7 @@ import six
 log = logging.getLogger(__name__)
 
 
-RE_FWD = re.compile("^[-]+[ ]*Forwarded message[ ]*[-]+$", re.I | re.M)
+RE_FWD = re.compile("^[-]+[ ]*Forwarded message[ ]*[-]+\s*$", re.I | re.M)
 
 RE_ON_DATE_SMB_WROTE = re.compile(
     u'(-*[>]?[ ]?({0})[ ].*({1})(.*\n){{0,2}}.*({2}):?-*)'.format(
@@ -38,10 +38,14 @@ RE_ON_DATE_SMB_WROTE = re.compile(
             'Op',
             # German
             'Am',
+            # Portuguese
+            'Em',
             # Norwegian
             u'På',
             # Swedish, Danish
             'Den',
+            # Vietnamese
+            u'Vào',
         )),
         # Date and sender separator
         u'|'.join((
@@ -62,8 +66,12 @@ RE_ON_DATE_SMB_WROTE = re.compile(
             'schreef','verzond','geschreven',
             # German
             'schrieb',
+            # Portuguese
+            'escreveu',
             # Norwegian, Swedish
             'skrev',
+            # Vietnamese
+            u'đã viết',
         ))
     ))
 # Special case for languages where text is translated like this: 'on {date} wrote {somebody}:'
@@ -131,19 +139,23 @@ RE_ORIGINAL_MESSAGE = re.compile(u'[\s]*[-]+[ ]*({})[ ]*[-]+'.format(
         'Oprindelig meddelelse',
     ))), re.I)
 
-RE_FROM_COLON_OR_DATE_COLON = re.compile(u'(_+\r?\n)?[\s]*(:?[*]?{})[\s]?:[*]?.*'.format(
+RE_FROM_COLON_OR_DATE_COLON = re.compile(u'((_+\r?\n)?[\s]*:?[*]?({})[\s]?:([^\n$]+\n){{1,2}}){{2,}}'.format(
     u'|'.join((
         # "From" in different languages.
         'From', 'Van', 'De', 'Von', 'Fra', u'Från',
         # "Date" in different languages.
-        'Date', 'Datum', u'Envoyé', 'Skickat', 'Sendt',
-    ))), re.I)
+        'Date', '[S]ent', 'Datum', u'Envoyé', 'Skickat', 'Sendt', 'Gesendet',
+        # "Subject" in different languages.
+        'Subject', 'Betreff', 'Objet', 'Emne', u'Ämne',
+        # "To" in different languages.
+        'To', 'An', 'Til', u'À', 'Till'
+    ))), re.I | re.M)
 
 # ---- John Smith wrote ----
 RE_ANDROID_WROTE = re.compile(u'[\s]*[-]+.*({})[ ]*[-]+'.format(
     u'|'.join((
         # English
-        'wrote'
+        'wrote',
     ))), re.I)
 
 # Support polymail.io reply format
@@ -161,15 +173,15 @@ SPLITTER_PATTERNS = [
     RE_FROM_COLON_OR_DATE_COLON,
     # 02.04.2012 14:20 пользователь "bob@example.com" <
     # bob@xxx.mailgun.org> написал:
-    re.compile("(\d+/\d+/\d+|\d+\.\d+\.\d+).*@", re.S),
+    re.compile("(\d+/\d+/\d+|\d+\.\d+\.\d+).*\s\S+@\S+", re.S),
     # 2014-10-17 11:28 GMT+03:00 Bob <
     # bob@example.com>:
-    re.compile("\d{4}-\d{2}-\d{2}\s+\d{2}:\d{2}\s+GMT.*@", re.S),
+    re.compile("\d{4}-\d{2}-\d{2}\s+\d{2}:\d{2}\s+GMT.*\s\S+@\S+", re.S),
     # Thu, 26 Jun 2014 14:00:51 +0400 Bob <bob@example.com>:
     re.compile('\S{3,10}, \d\d? \S{3,10} 20\d\d,? \d\d?:\d\d(:\d\d)?'
                '( \S+){3,6}@\S+:'),
     # Sent from Samsung MobileName <address@example.com> wrote:
-    re.compile('Sent from Samsung .*@.*> wrote'),
+    re.compile('Sent from Samsung.* \S+@\S+> wrote'),
     RE_ANDROID_WROTE,
     RE_POLYMAIL
     ]
@@ -282,7 +294,7 @@ def process_marked_lines(lines, markers, return_flags=[False, -1, -1]):
     # inlined reply
     # use lookbehind assertions to find overlapping entries e.g. for 'mtmtm'
     # both 't' entries should be found
-    for inline_reply in re.finditer('(?<=m)e*((?:t+e*)+)m', markers):
+    for inline_reply in re.finditer('(?<=m)e*(t[te]*)m', markers):
         # long links could break sequence of quotation lines but they shouldn't
         # be considered an inline reply
         links = (
@@ -426,6 +438,9 @@ def _extract_from_html(msg_body):
     Extract not quoted message from provided html message body
     using tags and plain text algorithm.
 
+    Cut out first some encoding html tags such as xml and doctype
+    for avoiding conflict with unicode decoding
+
     Cut out the 'blockquote', 'gmail_quote' tags.
     Cut Microsoft quotations.
 
@@ -441,18 +456,27 @@ def _extract_from_html(msg_body):
         return msg_body
 
     msg_body = msg_body.replace(b'\r\n', b'\n')
+
+    msg_body = re.sub(r"\<\?xml.+\?\>|\<\!DOCTYPE.+]\>", "", msg_body)
+
     html_tree = html_document_fromstring(msg_body)
 
     if html_tree is None:
         return msg_body
 
-    cut_quotations = (html_quotations.cut_gmail_quote(html_tree) or
-                      html_quotations.cut_zimbra_quote(html_tree) or
-                      html_quotations.cut_blockquote(html_tree) or
-                      html_quotations.cut_microsoft_quote(html_tree) or
-                      html_quotations.cut_by_id(html_tree) or
-                      html_quotations.cut_from_block(html_tree)
-                      )
+    cut_quotations = False
+    try:
+        cut_quotations = (html_quotations.cut_gmail_quote(html_tree) or
+                          html_quotations.cut_zimbra_quote(html_tree) or
+                          html_quotations.cut_blockquote(html_tree) or
+                          html_quotations.cut_microsoft_quote(html_tree) or
+                          html_quotations.cut_by_id(html_tree) or
+                          html_quotations.cut_from_block(html_tree)
+                          )
+    except Exception as e:
+        log.exception('during html quotations cut')
+        pass
+
     html_tree_copy = deepcopy(html_tree)
 
     number_of_checkpoints = html_quotations.add_checkpoint(html_tree, 0)
@@ -498,9 +522,69 @@ def _extract_from_html(msg_body):
     if _readable_text_empty(html_tree_copy):
         return msg_body
 
+    # NOTE: We remove_namespaces() because we are using an HTML5 Parser, HTML
+    # parsers do not recognize namespaces in HTML tags. As such the rendered
+    # HTML tags are no longer recognizable HTML tags. Example: <o:p> becomes
+    # <oU0003Ap>. When we port this to golang we should look into using an
+    # XML Parser NOT and HTML5 Parser since we do not know what input a
+    # customer will send us. Switching to a common XML parser in python
+    # opens us up to a host of vulnerabilities.
+    # See https://docs.python.org/3/library/xml.html#xml-vulnerabilities
+    #
+    # The down sides to removing the namespaces is that customers might
+    # judge the XML namespaces important. If that is the case then support
+    # should encourage customers to preform XML parsing of the un-stripped
+    # body to get the full unmodified XML payload.
+    #
+    # Alternatives to this approach are
+    # 1. Ignore the U0003A in tag names and let the customer deal with it.
+    #    This is not ideal, as most customers use stripped-html for viewing
+    #    emails sent from a recipient, as such they cannot control the HTML
+    #    provided by a recipient.
+    # 2. Preform a string replace of 'U0003A' to ':' on the rendered HTML
+    #    string. While this would solve the issue simply, it runs the risk
+    #    of replacing data outside the <tag> which might be essential to
+    #    the customer.
+    remove_namespaces(html_tree_copy)
     return html.tostring(html_tree_copy)
 
 
+def remove_namespaces(root):
+    """
+    Given the root of an HTML document iterate through all the elements
+    and remove any namespaces that might have been provided and remove
+    any attributes that contain a namespace
+
+    <html xmlns:o="urn:schemas-microsoft-com:office:office">
+    becomes
+    <html>
+
+    <o:p>Hi</o:p>
+    becomes
+    <p>Hi</p>
+
+    Start tags do NOT have a namespace; COLON characters have no special meaning.
+    if we don't remove the namespace the parser translates the tag name into a
+    unicode representation. For example <o:p> becomes <oU0003Ap>
+
+    See https://www.w3.org/TR/2011/WD-html5-20110525/syntax.html#start-tags
+
+
+    """
+    for child in root.iter():
+        for key, value in child.attrib.items():
+            # If the attribute includes a colon
+            if key.rfind("U0003A") != -1:
+                child.attrib.pop(key)
+
+        # If the tag includes a colon
+        idx = child.tag.rfind("U0003A")
+        if idx != -1:
+            child.tag = child.tag[idx+6:]
+
+    return root
+
+
 def split_emails(msg):
     """
     Given a message (which may consist of an email conversation thread with
@@ -553,7 +637,6 @@ def _correct_splitlines_in_headers(markers, lines):
     updated_markers = ""
     i = 0
     in_header_block = False
-
     for m in markers:
         # Only set in_header_block flag when we hit an 's' and line is a header
         if m == 's':

talon/signature/bruteforce.py

@@ -1,15 +1,15 @@
 from __future__ import absolute_import
+
 import logging
 
 import regex as re
 
-from talon.utils import get_delimiter
 from talon.signature.constants import (SIGNATURE_MAX_LINES,
                                        TOO_LONG_SIGNATURE_LINE)
+from talon.utils import get_delimiter
 
 log = logging.getLogger(__name__)
 
-
 # regex to fetch signature based on common signature words
 RE_SIGNATURE = re.compile(r'''
                (
@@ -28,7 +28,6 @@ RE_SIGNATURE = re.compile(r'''
                )
                ''', re.I | re.X | re.M | re.S)
 
-
 # signatures appended by phone email clients
 RE_PHONE_SIGNATURE = re.compile(r'''
                (
@@ -45,7 +44,6 @@ RE_PHONE_SIGNATURE = re.compile(r'''
                )
                ''', re.I | re.X | re.M | re.S)
 
-
 # see _mark_candidate_indexes() for details
 # c - could be signature line
 # d - line starts with dashes (could be signature or list item)
@@ -64,7 +62,7 @@ RE_SIGNATURE_CANDIDATE = re.compile(r'''
 
 
 def extract_signature(msg_body):
-    '''
+    """
     Analyzes message for a presence of signature block (by common patterns)
     and returns tuple with two elements: message text without signature block
     and the signature itself.
@@ -74,7 +72,7 @@ def extract_signature(msg_body):
 
     >>> extract_signature('Hey man!')
     ('Hey man!', None)
-    '''
+    """
     try:
         # identify line delimiter first
         delimiter = get_delimiter(msg_body)
@@ -112,7 +110,7 @@ def extract_signature(msg_body):
 
             return (stripped_body.strip(),
                     signature.strip())
-    except Exception as e:
+    except Exception:
         log.exception('ERROR extracting signature')
         return (msg_body, None)
 
@@ -163,7 +161,7 @@ def _mark_candidate_indexes(lines, candidate):
     'cdc'
     """
     # at first consider everything to be potential signature lines
-    markers = bytearray('c'*len(candidate))
+    markers = list('c' * len(candidate))
 
     # mark lines starting from bottom up
     for i, line_idx in reversed(list(enumerate(candidate))):
@@ -174,7 +172,7 @@ def _mark_candidate_indexes(lines, candidate):
             if line.startswith('-') and line.strip("-"):
                 markers[i] = 'd'
 
-    return markers
+    return "".join(markers)
 
 
 def _process_marked_candidate_indexes(candidate, markers):

talon/signature/data/__init__.py

@@ -0,0 +1 @@
+

talon/signature/extraction.py

@@ -1,16 +1,15 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
+
 import logging
 
-import regex as re
 import numpy
-
-from talon.signature.learning.featurespace import features, build_pattern
-from talon.utils import get_delimiter
+import regex as re
 from talon.signature.bruteforce import get_signature_candidate
+from talon.signature.learning.featurespace import features, build_pattern
 from talon.signature.learning.helpers import has_signature
-
+from talon.utils import get_delimiter
 
 log = logging.getLogger(__name__)
 
@@ -33,7 +32,7 @@ RE_REVERSE_SIGNATURE = re.compile(r'''
 
 def is_signature_line(line, sender, classifier):
     '''Checks if the line belongs to signature. Returns True or False.'''
-    data = numpy.array(build_pattern(line, features(sender)))
+    data = numpy.array(build_pattern(line, features(sender))).reshape(1, -1)
     return classifier.predict(data) > 0
 
 
@@ -58,7 +57,7 @@ def extract(body, sender):
                 text = delimiter.join(text)
                 if text.strip():
                     return (text, delimiter.join(signature))
-    except Exception:
+    except Exception as e:
         log.exception('ERROR when extracting signature with classifiers')
 
     return (body, None)
@@ -81,7 +80,7 @@ def _mark_lines(lines, sender):
     candidate = get_signature_candidate(lines)
 
     # at first consider everything to be text no signature
-    markers = bytearray('t'*len(lines))
+    markers = list('t' * len(lines))
 
     # mark lines starting from bottom up
     # mark only lines that belong to candidate
@@ -96,7 +95,7 @@ def _mark_lines(lines, sender):
         elif is_signature_line(line, sender, EXTRACTOR):
             markers[j] = 's'
 
-    return markers
+    return "".join(markers)
 
 
 def _process_marked_lines(lines, markers):
@@ -111,3 +110,4 @@ def _process_marked_lines(lines, markers):
         return (lines[:-signature.end()], lines[-signature.end():])
 
     return (lines, None)
+

talon/signature/learning/classifier.py

@@ -6,9 +6,10 @@ body belongs to the signature.
 """
 
 from __future__ import absolute_import
+
 from numpy import genfromtxt
-from sklearn.svm import LinearSVC
 from sklearn.externals import joblib
+from sklearn.svm import LinearSVC
 
 
 def init():
@@ -29,4 +30,40 @@ def train(classifier, train_data_filename, save_classifier_filename=None):
 
 def load(saved_classifier_filename, train_data_filename):
     """Loads saved classifier. """
-    return joblib.load(saved_classifier_filename)
+    try:
+        return joblib.load(saved_classifier_filename)
+    except Exception:
+        import sys
+        if sys.version_info > (3, 0):
+            return load_compat(saved_classifier_filename)
+
+        raise
+
+
+def load_compat(saved_classifier_filename):
+    import os
+    import pickle
+    import tempfile
+
+    # we need to switch to the data path to properly load the related _xx.npy files
+    cwd = os.getcwd()
+    os.chdir(os.path.dirname(saved_classifier_filename))
+
+    # convert encoding using pick.load and write to temp file which we'll tell joblib to use
+    pickle_file = open(saved_classifier_filename, 'rb')
+    classifier = pickle.load(pickle_file, encoding='latin1')
+
+    try:
+        # save our conversion if permissions allow
+        joblib.dump(classifier, saved_classifier_filename)
+    except Exception:
+        # can't write to classifier, use a temp file
+        tmp = tempfile.SpooledTemporaryFile()
+        joblib.dump(classifier, tmp)
+        saved_classifier_filename = tmp
+
+    # important, use joblib.load before switching back to original cwd
+    jb_classifier = joblib.load(saved_classifier_filename)
+    os.chdir(cwd)
+
+    return jb_classifier

talon/signature/learning/dataset.py

@@ -17,13 +17,14 @@ suffix which should be `_sender`.
 """
 
 from __future__ import absolute_import
+
 import os
+
 import regex as re
+from six.moves import range
 
 from talon.signature.constants import SIGNATURE_MAX_LINES
 from talon.signature.learning.featurespace import build_pattern, features
-from six.moves import range
-
 
 SENDER_SUFFIX = '_sender'
 BODY_SUFFIX = '_body'
@@ -57,9 +58,14 @@ def parse_msg_sender(filename, sender_known=True):
     algorithm:
     >>> parse_msg_sender(filename, False)
     """
+    import sys
+    kwargs = {}
+    if sys.version_info > (3, 0):
+        kwargs["encoding"] = "utf8"
+
     sender, msg = None, None
     if os.path.isfile(filename) and not is_sender_filename(filename):
-        with open(filename) as f:
+        with open(filename, **kwargs) as f:
             msg = f.read()
             sender = u''
             if sender_known:
@@ -147,7 +153,7 @@ def build_extraction_dataset(folder, dataset_filename,
                 continue
             lines = msg.splitlines()
             for i in range(1, min(SIGNATURE_MAX_LINES,
-                                   len(lines)) + 1):
+                                  len(lines)) + 1):
                 line = lines[-i]
                 label = -1
                 if line[:len(SIGNATURE_ANNOTATION)] == \

talon/utils.py

@@ -1,19 +1,18 @@
 # coding:utf-8
 
 from __future__ import absolute_import
-import logging
+
 from random import shuffle
-import chardet
+
 import cchardet
-import regex as re
-
-from lxml.html import html5parser
-from lxml.cssselect import CSSSelector
-
+import chardet
 import html5lib
+import regex as re
+import six
+from lxml.cssselect import CSSSelector
+from lxml.html import html5parser
 
 from talon.constants import RE_DELIMITER
-import six
 
 
 def safe_format(format_string, *args, **kwargs):
@@ -128,11 +127,11 @@ def html_tree_to_text(tree):
 
         parent.remove(c)
 
-    text   = ""
+    text = ""
     for el in tree.iter():
         el_text = (el.text or '') + (el.tail or '')
         if len(el_text) > 1:
-            if el.tag in _BLOCKTAGS:
+            if el.tag in _BLOCKTAGS + _HARDBREAKS:
                 text += "\n"
             if el.tag == 'li':
                 text += "  * "
@@ -143,7 +142,8 @@ def html_tree_to_text(tree):
             if href:
                 text += "(%s) " % href
 
-        if el.tag in _HARDBREAKS and text and not text.endswith("\n"):
+        if (el.tag in _HARDBREAKS and text and
+            not text.endswith("\n") and not el_text):
             text += "\n"
 
     retval = _rm_excessive_newlines(text)
@@ -177,6 +177,8 @@ def html_to_text(string):
 def html_fromstring(s):
     """Parse html tree from string. Return None if the string can't be parsed.
     """
+    if isinstance(s, six.text_type):
+        s = s.encode('utf8')
     try:
         if html_too_big(s):
             return None
@@ -189,6 +191,8 @@ def html_fromstring(s):
 def html_document_fromstring(s):
     """Parse html tree from string. Return None if the string can't be parsed.
     """
+    if isinstance(s, six.text_type):
+        s = s.encode('utf8')
     try:
         if html_too_big(s):
             return None
@@ -203,7 +207,9 @@ def cssselect(expr, tree):
 
 
 def html_too_big(s):
-    return s.count('<') > _MAX_TAGS_COUNT
+    if isinstance(s, six.text_type):
+        s = s.encode('utf8')
+    return s.count(b'<') > _MAX_TAGS_COUNT
 
 
 def _contains_charset_spec(s):
@@ -248,8 +254,7 @@ def _html5lib_parser():
 _UTF8_DECLARATION = (b'<meta http-equiv="Content-Type" content="text/html;'
                      b'charset=utf-8">')
 
-
-_BLOCKTAGS  = ['div', 'p', 'ul', 'li', 'h1', 'h2', 'h3']
+_BLOCKTAGS = ['div', 'p', 'ul', 'li', 'h1', 'h2', 'h3']
 _HARDBREAKS = ['br', 'hr', 'tr']
 
 _RE_EXCESSIVE_NEWLINES = re.compile("\n{2,10}")

tests/__init__.py

@@ -1,6 +1,4 @@
 from __future__ import absolute_import
-from nose.tools import *
-from mock import *
 
 import talon
 

tests/html_quotations_test.py

@@ -1,13 +1,12 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from . import *
-from . fixtures import *
-
-import regex as re
 
+from tests.fixtures import REPLY_QUOTATIONS_SHARE_BLOCK, OLK_SRC_BODY_SECTION, REPLY_SEPARATED_BY_HR
+from nose.tools import eq_, ok_, assert_false, assert_true
 from talon import quotations, utils as u
-
+from mock import Mock, patch
+import re
 
 RE_WHITESPACE = re.compile("\s")
 RE_DOUBLE_WHITESPACE = re.compile("\s")
@@ -303,7 +302,12 @@ Reply
 
 
 def extract_reply_and_check(filename):
-    f = open(filename)
+    import sys
+    kwargs = {}
+    if sys.version_info > (3, 0):
+        kwargs["encoding"] = "utf8"
+
+    f = open(filename, **kwargs)
 
     msg_body = f.read()
     reply = quotations.extract_from_html(msg_body)
@@ -373,7 +377,7 @@ reply
 </blockquote>"""
     msg_body = msg_body.replace('\n', '\r\n')
     extracted = quotations.extract_from_html(msg_body)
-    assert_false(symbol in extracted)    
+    assert_false(symbol in extracted)
     # Keep new lines otherwise "My reply" becomes one word - "Myreply" 
     eq_("<html><head></head><body>My\nreply\n</body></html>", extracted)
 
@@ -419,3 +423,23 @@ def test_readable_html_empty():
 def test_bad_html():
     bad_html = "<html></html>"
     eq_(bad_html, quotations.extract_from_html(bad_html))
+
+
+def test_remove_namespaces():
+    msg_body = """
+    <html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns="http://www.w3.org/TR/REC-html40">
+        <body>
+            <o:p>Dear Sir,</o:p>
+            <o:p>Thank you for the email.</o:p>
+            <blockquote>thing</blockquote>
+        </body>
+    </html>
+    """
+
+    rendered = quotations.extract_from_html(msg_body)
+
+    assert_true("<p>" in rendered)
+    assert_true("xmlns" in rendered)
+
+    assert_true("<o:p>" not in rendered)
+    assert_true("<xmlns:o>" not in rendered)

tests/quotations_test.py

@@ -1,10 +1,10 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from . import *
-from . fixtures import *
 
+from mock import Mock, patch
 from talon import quotations
+from nose.tools import eq_
 
 
 @patch.object(quotations, 'extract_from_html')

tests/signature/bruteforce_test.py

@@ -1,9 +1,10 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from .. import *
+from nose.tools import eq_
 
 from talon.signature import bruteforce
+from mock import patch, Mock
 
 
 def test_empty_body():

tests/signature/extraction_test.py

@@ -1,15 +1,15 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from .. import *
 
-import os
-
-from talon.signature.learning import dataset
-from talon import signature
+from talon.signature import bruteforce, extraction, extract
 from talon.signature import extraction as e
-from talon.signature import bruteforce
+from talon.signature.learning import dataset
+from nose.tools import eq_
+from .. import STRIPPED, UNICODE_MSG
 from six.moves import range
+from mock import patch
+import os
 
 
 def test_message_shorter_SIGNATURE_MAX_LINES():
@@ -18,23 +18,28 @@ def test_message_shorter_SIGNATURE_MAX_LINES():
 
 Thanks in advance,
 Bob"""
-    text, extracted_signature = signature.extract(body, sender)
+    text, extracted_signature = extract(body, sender)
     eq_('\n'.join(body.splitlines()[:2]), text)
     eq_('\n'.join(body.splitlines()[-2:]), extracted_signature)
 
 
 def test_messages_longer_SIGNATURE_MAX_LINES():
+    import sys
+    kwargs = {}
+    if sys.version_info > (3, 0):
+        kwargs["encoding"] = "utf8"
+
     for filename in os.listdir(STRIPPED):
         filename = os.path.join(STRIPPED, filename)
         if not filename.endswith('_body'):
             continue
         sender, body = dataset.parse_msg_sender(filename)
-        text, extracted_signature = signature.extract(body, sender)
+        text, extracted_signature = extract(body, sender)
         extracted_signature = extracted_signature or ''
-        with open(filename[:-len('body')] + 'signature') as ms:
+        with open(filename[:-len('body')] + 'signature', **kwargs) as ms:
             msg_signature = ms.read()
             eq_(msg_signature.strip(), extracted_signature.strip())
-            stripped_msg = body.strip()[:len(body.strip())-len(msg_signature)]
+            stripped_msg = body.strip()[:len(body.strip()) - len(msg_signature)]
             eq_(stripped_msg.strip(), text.strip())
 
 
@@ -47,7 +52,7 @@ Thanks in advance,
 some text which doesn't seem to be a signature at all
 Bob"""
 
-    text, extracted_signature = signature.extract(body, sender)
+    text, extracted_signature = extract(body, sender)
     eq_('\n'.join(body.splitlines()[:2]), text)
     eq_('\n'.join(body.splitlines()[-3:]), extracted_signature)
 
@@ -60,7 +65,7 @@ Thanks in advance,
 some long text here which doesn't seem to be a signature at all
 Bob"""
 
-    text, extracted_signature = signature.extract(body, sender)
+    text, extracted_signature = extract(body, sender)
     eq_('\n'.join(body.splitlines()[:-1]), text)
     eq_('Bob', extracted_signature)
 
@@ -68,13 +73,13 @@ Bob"""
 
     some *long* text here which doesn't seem to be a signature at all
     """
-    ((body, None), signature.extract(body, "david@example.com"))
+    ((body, None), extract(body, "david@example.com"))
 
 
 def test_basic():
     msg_body = 'Blah\r\n--\r\n\r\nSergey Obukhov'
     eq_(('Blah', '--\r\n\r\nSergey Obukhov'),
-        signature.extract(msg_body, 'Sergey'))
+        extract(msg_body, 'Sergey'))
 
 
 def test_capitalized():
@@ -99,7 +104,7 @@ Doe Inc
 Doe Inc
 555-531-7967"""
 
-    eq_(sig, signature.extract(msg_body, 'Doe')[1])
+    eq_(sig, extract(msg_body, 'Doe')[1])
 
 
 def test_over_2_text_lines_after_signature():
@@ -110,25 +115,25 @@ def test_over_2_text_lines_after_signature():
     2 non signature lines in the end
     It's not signature
     """
-    text, extracted_signature = signature.extract(body, "Bob")
+    text, extracted_signature = extract(body, "Bob")
     eq_(extracted_signature, None)
 
 
 def test_no_signature():
     sender, body = "bob@foo.bar", "Hello"
-    eq_((body, None), signature.extract(body, sender))
+    eq_((body, None), extract(body, sender))
 
 
 def test_handles_unicode():
     sender, body = dataset.parse_msg_sender(UNICODE_MSG)
-    text, extracted_signature = signature.extract(body, sender)
+    text, extracted_signature = extract(body, sender)
 
 
-@patch.object(signature.extraction, 'has_signature')
+@patch.object(extraction, 'has_signature')
 def test_signature_extract_crash(has_signature):
     has_signature.side_effect = Exception('Bam!')
     msg_body = u'Blah\r\n--\r\n\r\nСергей'
-    eq_((msg_body, None), signature.extract(msg_body, 'Сергей'))
+    eq_((msg_body, None), extract(msg_body, 'Сергей'))
 
 
 def test_mark_lines():
@@ -137,19 +142,19 @@ def test_mark_lines():
         # (starting from the bottom) because we don't count empty line
         eq_('ttset',
             e._mark_lines(['Bob Smith',
-                          'Bob Smith',
-                          'Bob Smith',
-                          '',
-                          'some text'], 'Bob Smith'))
+                           'Bob Smith',
+                           'Bob Smith',
+                           '',
+                           'some text'], 'Bob Smith'))
 
     with patch.object(bruteforce, 'SIGNATURE_MAX_LINES', 3):
         # we don't analyse the 1st line because
         # signature cant start from the 1st line
         eq_('tset',
             e._mark_lines(['Bob Smith',
-                          'Bob Smith',
-                          '',
-                          'some text'], 'Bob Smith'))
+                           'Bob Smith',
+                           '',
+                           'some text'], 'Bob Smith'))
 
 
 def test_process_marked_lines():

tests/signature/learning/dataset_test.py

@@ -1,14 +1,13 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from ... import *
-import os
-
-from numpy import genfromtxt
-
-from talon.signature.learning import dataset as d
 
+from ... import EML_MSG_FILENAME, MSG_FILENAME_WITH_BODY_SUFFIX, TMP_DIR, EMAILS_DIR
 from talon.signature.learning.featurespace import features
+from talon.signature.learning import dataset as d
+from nose.tools import eq_, assert_false, ok_
+from numpy import genfromtxt
+import os
 
 
 def test_is_sender_filename():

tests/signature/learning/featurespace_test.py

@@ -1,9 +1,10 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from ... import *
 
 from talon.signature.learning import featurespace as fs
+from nose.tools import eq_, assert_false, ok_
+from mock import patch
 
 
 def test_apply_features():

tests/signature/learning/helpers_test.py

@@ -1,13 +1,13 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from ... import *
-
-import regex as re
 
 from talon.signature.learning import helpers as h
-from talon.signature.learning.helpers import *
+from talon.signature.learning.helpers import RE_RELAX_PHONE, RE_NAME
+from nose.tools import eq_, ok_, assert_false, assert_in
+from mock import patch, Mock
 from six.moves import range
+import re
 
 # First testing regex constants.
 VALID = '''

tests/text_quotations_test.py

@@ -1,16 +1,15 @@
 # -*- coding: utf-8 -*-
 
 from __future__ import absolute_import
-from . import *
-from . fixtures import *
 
-import os
-
-import email.iterators
+from tests.fixtures import STANDARD_REPLIES
 from talon import quotations
-import six
 from six.moves import range
-from six import StringIO
+from nose.tools import eq_
+from mock import patch
+import email.iterators
+import six
+import os
 
 
 @patch.object(quotations, 'MAX_LINES_COUNT', 1)
@@ -35,6 +34,7 @@ On 11-Apr-2011, at 6:54 PM, Roman Tkachenko <romant@example.com> wrote:
 
     eq_("Test reply", quotations.extract_from_plain(msg_body))
 
+
 def test_pattern_on_date_polymail():
     msg_body = """Test reply
 
@@ -119,6 +119,38 @@ On 11-Apr-2011, at 6:54 PM, Roman Tkachenko <romant@example.com> sent:
     eq_("Test reply", quotations.extract_from_plain(msg_body))
 
 
+def test_appointment():
+    msg_body = """Response
+
+10/19/2017 @ 9:30 am for physical therapy
+Bla
+1517 4th Avenue Ste 300
+London CA 19129, 555-421-6780
+
+John Doe, FCLS
+Mailgun Inc
+555-941-0697
+
+From: from@example.com [mailto:from@example.com]
+Sent: Wednesday, October 18, 2017 2:05 PM
+To: John Doer - SIU <jd@example.com>
+Subject: RE: Claim # 5551188-1
+
+Text"""
+
+    expected = """Response
+
+10/19/2017 @ 9:30 am for physical therapy
+Bla
+1517 4th Avenue Ste 300
+London CA 19129, 555-421-6780
+
+John Doe, FCLS
+Mailgun Inc
+555-941-0697"""
+    eq_(expected, quotations.extract_from_plain(msg_body))
+
+
 def test_line_starts_with_on():
     msg_body = """Blah-blah-blah
 On blah-blah-blah"""
@@ -158,14 +190,17 @@ Test"""
     eq_('Test reply', quotations.extract_from_plain(
         msg_body.format(six.text_type(original_message_indicator))))
 
+
 def test_english_original_message():
     _check_pattern_original_message('Original Message')
     _check_pattern_original_message('Reply Message')
 
+
 def test_german_original_message():
     _check_pattern_original_message(u'Ursprüngliche Nachricht')
     _check_pattern_original_message('Antwort Nachricht')
 
+
 def test_danish_original_message():
     _check_pattern_original_message('Oprindelig meddelelse')
 
@@ -264,6 +299,7 @@ On 04/19/2011 07:10 AM, Roman Tkachenko wrote:
 > Hello"""
     eq_("Hi", quotations.extract_from_plain(msg_body))
 
+
 def test_with_indent():
     msg_body = """YOLO salvia cillum kogi typewriter mumblecore cardigan skateboard Austin.
 
@@ -271,7 +307,8 @@ def test_with_indent():
 
 Brunch mumblecore pug Marfa tofu, irure taxidermy hoodie readymade pariatur.
     """
-    eq_("YOLO salvia cillum kogi typewriter mumblecore cardigan skateboard Austin.", quotations.extract_from_plain(msg_body))
+    eq_("YOLO salvia cillum kogi typewriter mumblecore cardigan skateboard Austin.",
+        quotations.extract_from_plain(msg_body))
 
 
 def test_short_quotation_with_newline():
@@ -311,6 +348,7 @@ Subject: The manager has commented on your Loop
 Blah-blah-blah
 """))
 
+
 def test_german_from_block():
     eq_('Allo! Follow up MIME!', quotations.extract_from_plain(
     """Allo! Follow up MIME!
@@ -323,6 +361,7 @@ Betreff: The manager has commented on your Loop
 Blah-blah-blah
 """))
 
+
 def test_french_multiline_from_block():
     eq_('Lorem ipsum', quotations.extract_from_plain(
     u"""Lorem ipsum
@@ -335,6 +374,7 @@ Objet : Follow Up
 Blah-blah-blah
 """))
 
+
 def test_french_from_block():
     eq_('Lorem ipsum', quotations.extract_from_plain(
     u"""Lorem ipsum
@@ -343,6 +383,7 @@ Le 23 janv. 2015 à 22:03, Brendan xxx <brendan.xxx@xxx.com<mailto:brendan.xxx@x
 
 Bonjour!"""))
 
+
 def test_polish_from_block():
     eq_('Lorem ipsum', quotations.extract_from_plain(
     u"""Lorem ipsum
@@ -353,6 +394,7 @@ napisał:
 Blah!
 """))
 
+
 def test_danish_from_block():
     eq_('Allo! Follow up MIME!', quotations.extract_from_plain(
     """Allo! Follow up MIME!
@@ -365,6 +407,7 @@ Emne: The manager has commented on your Loop
 Blah-blah-blah
 """))
 
+
 def test_swedish_from_block():
     eq_('Allo! Follow up MIME!', quotations.extract_from_plain(
     u"""Allo! Follow up MIME!
@@ -376,6 +419,7 @@ Till: Isacson Leiff
 Blah-blah-blah
 """))
 
+
 def test_swedish_from_line():
     eq_('Lorem', quotations.extract_from_plain(
     """Lorem
@@ -384,6 +428,7 @@ Den 14 september, 2015 02:23:18, Valentino Rudy (valentino@rudy.be) skrev:
 Veniam laborum mlkshk kale chips authentic. Normcore mumblecore laboris, fanny pack readymade eu blog chia pop-up freegan enim master cleanse.
 """))
 
+
 def test_norwegian_from_line():
     eq_('Lorem', quotations.extract_from_plain(
     u"""Lorem
@@ -392,6 +437,7 @@ På 14 september 2015 på 02:23:18, Valentino Rudy (valentino@rudy.be) skrev:
 Veniam laborum mlkshk kale chips authentic. Normcore mumblecore laboris, fanny pack readymade eu blog chia pop-up freegan enim master cleanse.
 """))
 
+
 def test_dutch_from_block():
     eq_('Gluten-free culpa lo-fi et nesciunt nostrud.', quotations.extract_from_plain(
     """Gluten-free culpa lo-fi et nesciunt nostrud.
@@ -402,6 +448,16 @@ Small batch beard laboris tempor, non listicle hella Tumblr heirloom.
 """))
 
 
+def test_vietnamese_from_block():
+    eq_('Hello', quotations.extract_from_plain(
+    u"""Hello
+
+Vào 14:24 8 tháng 6, 2017, Hùng Nguyễn <hungnguyen@xxx.com> đã viết:
+
+> Xin chào
+"""))
+
+
 def test_quotation_marker_false_positive():
     msg_body = """Visit us now for assistance...
 >>> >>>  http://www.domain.com <<<
@@ -413,7 +469,8 @@ def test_link_closed_with_quotation_marker_on_new_line():
     msg_body = '''8.45am-1pm
 
 From: somebody@example.com
-
+Date: Wed, 16 May 2012 00:15:02 -0600
+ 
 <http://email.example.com/c/dHJhY2tpbmdfY29kZT1mMDdjYzBmNzM1ZjYzMGIxNT
 >  <bob@example.com <mailto:bob@example.com> >
 
@@ -454,7 +511,9 @@ def test_from_block_starts_with_date():
     msg_body = """Blah
 
 Date: Wed, 16 May 2012 00:15:02 -0600
-To: klizhentas@example.com"""
+To: klizhentas@example.com
+
+"""
     eq_('Blah', quotations.extract_from_plain(msg_body))
 
 
@@ -524,11 +583,12 @@ def test_mark_message_lines():
              # next line should be marked as splitter
              '_____________',
              'From: foo@bar.com',
+             'Date: Wed, 16 May 2012 00:15:02 -0600',
              '',
              '> Hi',
              '',
              'Signature']
-    eq_('tessemet', quotations.mark_message_lines(lines))
+    eq_('tesssemet', quotations.mark_message_lines(lines))
 
     lines = ['Just testing the email reply',
              '',
@@ -767,6 +827,31 @@ def test_split_email():
         >
         >
 """
-    expected_markers = "stttttsttttetesetesmmmmmmssmmmmmmsmmmmmmmm"
+    expected_markers = "stttttsttttetesetesmmmmmmsmmmmmmmmmmmmmmmm"
     markers = quotations.split_emails(msg)
     eq_(markers, expected_markers)
+
+
+
+def test_feedback_below_left_unparsed():
+    msg_body = """Please enter your feedback below. Thank you.
+
+------------------------------------- Enter Feedback Below -------------------------------------
+
+The user experience was unparallelled. Please continue production. I'm sending payment to ensure
+that this line is intact."""
+
+    parsed = quotations.extract_from_plain(msg_body)
+    eq_(msg_body, parsed.decode('utf8'))
+
+
+def test_appointment():
+    msg_body = """Invitation for an interview:
+
+Date: Wednesday 3, October 2011 
+Time: 7 : 00am 
+Address: 130 Fox St
+
+Please bring in your ID."""
+    parsed = quotations.extract_from_plain(msg_body)
+    eq_(msg_body, parsed.decode('utf8'))

tests/utils_test.py

@@ -1,12 +1,13 @@
 # coding:utf-8
 
 from __future__ import absolute_import
-from . import *
 
+
+from nose.tools import eq_, ok_, assert_false
 from talon import utils as u
+from mock import patch, Mock
 import cchardet
 import six
-from lxml import html
 
 
 def test_get_delimiter():
@@ -16,35 +17,35 @@ def test_get_delimiter():
 
 
 def test_unicode():
-    eq_ (u'hi', u.to_unicode('hi'))
-    eq_ (type(u.to_unicode('hi')), six.text_type )
-    eq_ (type(u.to_unicode(u'hi')), six.text_type )
-    eq_ (type(u.to_unicode('привет')), six.text_type )
-    eq_ (type(u.to_unicode(u'привет')), six.text_type )
-    eq_ (u"привет", u.to_unicode('привет'))
-    eq_ (u"привет", u.to_unicode(u'привет'))
+    eq_(u'hi', u.to_unicode('hi'))
+    eq_(type(u.to_unicode('hi')), six.text_type)
+    eq_(type(u.to_unicode(u'hi')), six.text_type)
+    eq_(type(u.to_unicode('привет')), six.text_type)
+    eq_(type(u.to_unicode(u'привет')), six.text_type)
+    eq_(u"привет", u.to_unicode('привет'))
+    eq_(u"привет", u.to_unicode(u'привет'))
     # some latin1 stuff
-    eq_ (u"Versión", u.to_unicode(u'Versi\xf3n'.encode('iso-8859-2'), precise=True))
+    eq_(u"Versión", u.to_unicode(u'Versi\xf3n'.encode('iso-8859-2'), precise=True))
 
 
 def test_detect_encoding():
-    eq_ ('ascii', u.detect_encoding(b'qwe').lower())
-    ok_ (u.detect_encoding(
+    eq_('ascii', u.detect_encoding(b'qwe').lower())
+    ok_(u.detect_encoding(
         u'Versi\xf3n'.encode('iso-8859-2')).lower() in [
             'iso-8859-1', 'iso-8859-2'])
-    eq_ ('utf-8', u.detect_encoding(u'привет'.encode('utf8')).lower())
+    eq_('utf-8', u.detect_encoding(u'привет'.encode('utf8')).lower())
     # fallback to utf-8
     with patch.object(u.chardet, 'detect') as detect:
         detect.side_effect = Exception
-        eq_ ('utf-8', u.detect_encoding('qwe'.encode('utf8')).lower())
+        eq_('utf-8', u.detect_encoding('qwe'.encode('utf8')).lower())
 
 
 def test_quick_detect_encoding():
-    eq_ ('ascii', u.quick_detect_encoding(b'qwe').lower())
-    ok_ (u.quick_detect_encoding(
+    eq_('ascii', u.quick_detect_encoding(b'qwe').lower())
+    ok_(u.quick_detect_encoding(
         u'Versi\xf3n'.encode('windows-1252')).lower() in [
             'windows-1252', 'windows-1250'])
-    eq_ ('utf-8', u.quick_detect_encoding(u'привет'.encode('utf8')).lower())
+    eq_('utf-8', u.quick_detect_encoding(u'привет'.encode('utf8')).lower())
 
 
 @patch.object(cchardet, 'detect')
@@ -84,7 +85,7 @@ Haha
     eq_(u"привет!", u.html_to_text("<b>привет!</b>").decode('utf8'))
 
     html = '<body><br/><br/>Hi</body>'
-    eq_ (b'Hi', u.html_to_text(html))
+    eq_(b'Hi', u.html_to_text(html))
 
     html = """Hi
 <style type="text/css">
@@ -104,7 +105,7 @@ font: 13px 'Lucida Grande', Arial, sans-serif;
 
 }
 </style>"""
-    eq_ (b'Hi', u.html_to_text(html))
+    eq_(b'Hi', u.html_to_text(html))
 
     html = """<div>
 <!-- COMMENT 1 -->
@@ -115,15 +116,16 @@ font: 13px 'Lucida Grande', Arial, sans-serif;
 
 
 def test_comment_no_parent():
-    s = "<!-- COMMENT 1 --> no comment"
+    s = b'<!-- COMMENT 1 --> no comment'
     d = u.html_document_fromstring(s)
-    eq_("no comment", u.html_tree_to_text(d))
+    eq_(b"no comment", u.html_tree_to_text(d))
 
 
 @patch.object(u.html5parser, 'fromstring', Mock(side_effect=Exception()))
 def test_html_fromstring_exception():
     eq_(None, u.html_fromstring("<html></html>"))
 
+
 @patch.object(u, 'html_too_big', Mock())
 @patch.object(u.html5parser, 'fromstring')
 def test_html_fromstring_too_big(fromstring):
@@ -158,5 +160,5 @@ def test_html_too_big():
 
 @patch.object(u, '_MAX_TAGS_COUNT', 3)
 def test_html_to_text():
-    eq_("Hello", u.html_to_text("<div>Hello</div>"))
+    eq_(b"Hello", u.html_to_text("<div>Hello</div>"))
     eq_(None, u.html_to_text("<div><span>Hi</span></div>"))