From 95da927828f30a68fa4dbe9bd6bd53b3b5fe858e Mon Sep 17 00:00:00 2001
From: Vasily Zubarev <vzubarev@anaconda.com>
Date: Tue, 18 Feb 2020 15:32:35 +0100
Subject: [PATCH] Add telegram message cleanup

---
 parsing/telegram/cleanup.py | 16 ++++++++++++++
 parsing/telegram/models.py  | 42 ++++++++++---------------------------
 parsing/telegram/parsers.py | 34 +++++++++---------------------
 3 files changed, 37 insertions(+), 55 deletions(-)
 create mode 100644 parsing/telegram/cleanup.py

diff --git a/parsing/telegram/cleanup.py b/parsing/telegram/cleanup.py
new file mode 100644
index 0000000..b3b68cd
--- /dev/null
+++ b/parsing/telegram/cleanup.py
@@ -0,0 +1,16 @@
+import re
+
+
+def cleanup_telegram_message_text(text):
+    text = str(text)
+    cleanup_expressions = [
+        r"\[.*?\]\(.*?\)",  # attached files and images
+        r"#[\S]*",  # hashtags
+        r"[\*\~\`]+",  # stars and tildas
+        r"<[^>]*>",  # html tags
+        r"[\n]{3,}",  # triple newlines
+    ]
+    for pattern in cleanup_expressions:
+        text = re.sub(pattern, "", text)
+
+    return text.strip()
diff --git a/parsing/telegram/models.py b/parsing/telegram/models.py
index 7e3c7ce..1bffe97 100644
--- a/parsing/telegram/models.py
+++ b/parsing/telegram/models.py
@@ -3,28 +3,18 @@ from enum import Enum
 
 class TelegramChannel:
     def __init__(
-        self, channel_id=None, title=None, link=None, description=None, messages=None
+        self,
+        channel_id=None,
+        title=None,
+        link=None,
+        description=None,
+        messages=None
     ):
         self.channel_id = channel_id
         self.title = title
         self.link = link
         self.description = description
-        self.messages = messages if messages is not None else []
-
-    def add_message(self, message):
-        self.messages.append(message)
-
-    def remove_message(self, message):
-        self.messages.remove(message)
-
-    def to_dict(self):
-        return {
-            "channel_id": self.channel_id,
-            "title": self.title,
-            "link": self.link,
-            "description": self.description,
-            "messages": list(map(lambda message: message.to_dict(), self.messages)),
-        }
+        self.messages = messages or []
 
 
 class TelegramChannelMessage:
@@ -32,7 +22,8 @@ class TelegramChannelMessage:
         self,
         telegram_id=None,
         title=None,
-        description=None,
+        text=None,
+        clean_text=None,
         link=None,
         channel=None,
         grouped_id=None,
@@ -41,25 +32,14 @@ class TelegramChannelMessage:
     ):
         self.telegram_id = telegram_id
         self.title = title
-        self.description = description
+        self.text = text
+        self.clean_text = clean_text
         self.link = link
         self.channel = channel
         self.grouped_id = grouped_id
         self.type = type
         self.timestamp = timestamp
 
-    def to_dict(self):
-        return {
-            "telegram_id": self.telegram_id,
-            "title": self.title,
-            "description": self.description,
-            "link": self.link,
-            "channel": self.channel.channel_id,
-            "grouped_id": self.grouped_id,
-            "type": self.type.value,
-            "timestamp": self.timestamp.strftime("%Y-%m-%d %H:%M:%S"),
-        }
-
 
 class MessageType(Enum):
     TEXT = "text"
diff --git a/parsing/telegram/parsers.py b/parsing/telegram/parsers.py
index 80793bc..a120b3d 100644
--- a/parsing/telegram/parsers.py
+++ b/parsing/telegram/parsers.py
@@ -1,6 +1,7 @@
 from abc import ABC, abstractmethod
+
+from parsing.telegram.cleanup import cleanup_telegram_message_text
 from parsing.telegram.models import TelegramChannelMessage, TelegramChannel, MessageType
-import re
 
 
 class Parser(ABC):
@@ -26,34 +27,19 @@ class Parser(ABC):
 
 
 class SimpleTextParser(Parser):
-    MARKDOWN_BOLD = ".+\*.+\*.+"
-    SIMPLE_TEXT_POST = "(.+)\n+(.+)"
-
     def parse(self, channel, message):
         parsed_message = super().parse(channel, message)
-        (title, description) = self.parse_text(message)
-        parsed_message.title = self.enhance_title(title)
-        parsed_message.description = description
+        parsed_message.text = message
+        parsed_message.clean_text = cleanup_telegram_message_text(message)
+        parsed_message.title = self.parse_message_title(parsed_message.clean_text)
         parsed_message.type = MessageType.TEXT
         return parsed_message
 
-    def parse_text(self, message):
-        matcher = re.match(self.SIMPLE_TEXT_POST, message.text)
-        if matcher is not None:
-            if len(matcher.groups()) > 1:
-                title = matcher.group(1)
-                description = matcher.group(2)
-            else:
-                title = matcher.group(1)
-                description = message.text
-        else:
-            title = message.text
-            description = message.text
-        return title, description
-
-    def enhance_title(self, title):
-        matcher = re.match(self.MARKDOWN_BOLD, title)
-        return title.replace("*", "") if matcher is not None else title
+    @classmethod
+    def parse_message_title(cls, text):
+        if text:
+            return text.split("\n", 1)[0]
+        return ""
 
     def matches(self, channel, message):
         return message.text is not None and len(message.text) > 0