Telegram userbot in python

Question

I have a telegram userbot, monitoring certain group topic and answering certain messages. I want it to be faster than human, now it's surprisingly not. But I need to speed up its reaction, any tips?

from pyrogram import Client, filters
from decouple import config
from pyrogram.types import Message
import re
api_id = config('API_ID')
api_hash = config('API_HASH')
phone = config('PHONE')
login = config('LOGIN')
pattern = r'^Request'
bot = Client(name=login, api_id=api_id, api_hash=api_hash, phone_number=phone)
@bot.on_message(filters.text)
async def command_handler(client: Client, message: Message):
 if re.match(pattern, message.text):
 await client.send_message(
 chat_id=message.chat.id,
 text="reply",
 message_thread_id=message.message_thread_id
 )
 await bot.stop()
bot.run()

My requirements.txt file:

python-decouple==3.8
tgcrypto
kurigram

• J_H: You're complaining the response time is "slow", but it's unclear what the magnitude of the delay is. Rather than having a human type in "Request foo", write a short program that types in "Request timestamp [hh:mm:ss.fff]" so we have millisecond resolution. Then instrument the OP code to log a timestamp immediately upon invocation, and also upon starting and completing the .send_message()
• STerliakov: This program essentially does nothing, all the time spent in user code is spent in pyrogram internals. Since that library is built with performance in mind, I suspect you're either expecting some unreasonable response times (you won't achieve tens of milliseconds, telegram API is the limiting factor) or running the bot on a slow network connection. Anyway, provide your desired metrics and current metrics, profile first, and exclude first message after initialization from your profiling data (because python and pyrogram need some time to import and prepare).
• voipp: @STerliakov how to measure correctly latency?

Answer 1

pre-compiled regex

Your complaint is that response latency is high. It's unclear how many milliseconds "high" is, and what the target value after refactoring would be.

need to speed up its reaction, any tips?

No, not really. This is simple vanilla code, not doing anything ambitious.

pattern = r'^Request'
...
async def command_handler( ... ):
 if re.match(pattern, message.text):

We're doing slightly more work at request time than absolutely necessary. There's an opportunity to assign pattern = re.compile( ... ), and then test if pattern.match(message.text):. But I expect that's going to be insignificant, not enough to get you to your (unstated) target latency. Pre-compiling tends to be more interesting for a regex that is

• complex, and
• within a hot loop.

But that doesn't describe the current situation.

BTW, kudos on ^ anchoring the regex, so even a very long message.text cannot cause the .match() to do a lot of work.

warm the cache

Some statements, like import pandas, take a "long" time to execute the first time out, and thereafter are instantaneous due to a cache hit.

It's conceivable that the first .send_message() call issues a new import or does similar work, and that subsequent message sends happen quicker. Write an instrumented test to measure the initial and subsequent latencies, so we can prove or disprove this hypothesis. Consider warming the cache by sending an unprompted "I'm alive!" message at startup, to current channel or perhaps to some unmonitored dev-null channel.

• STerliakov: Even using just r"Request" will be instant (less than 1 ms) on the Telegram message due to its length limit, so there's no point looking at user code here:)
• voipp: wow, warmin up cache helped me. Every first message in listener slower than subsequent ones
• voipp: I measured latencies by calculating messages from chat with client.get_chat_history. So i calculate diff between my message and bot reply. Every first message is slower for 1 second. Subsequent messages and replies somehow have equal message time