How do I get my accounts' positions at Interactive Brokers using Python API?

EDITED: I found a solution regarding the error messages - it was a bug on IB's API. The code I show as an answer below should be useful for those looking for a clean solution to read positions, and also NAVs, from their accounts at IB.

The original question [SEE SOLUTION BELOW; LEAVING ORIGINAL QUESTION HERE FOR CONTEXT]: I'm trying to get all my accounts positions at Interactive Brokers [IB] using the firm's API. Their documentation, although extensive, is extremely confusing. Sample codes are full of unnecessary commands - I want something very streamlined.

I need help with:

1. How to get the information "per account" [SOLVED]
2. How to bring the variables to a DataFrame [SOLVED]
3. How to avoid IB's API from printing a series of error messages [SOLVED]

The code so far:

from ibapi.client import EClient 
from ibapi.wrapper import EWrapper
from ibapi.common import * #for TickerId type
import pandas as pd

class ib_class(EWrapper, EClient): 
    def __init__(self): 
        EClient.__init__(self, self)
        self.all_positions = pd.DataFrame([], columns = ['Account','Symbol', 'Quantity', 'Average Cost'])

    def position(self, account, contract, pos, avgCost):
        index = str(account)+str(contract.symbol)
        self.all_positions.loc[index]=account,contract.symbol,pos,avgCost

    def error(self, reqId:TickerId, errorCode:int, errorString:str):
        if reqId > -1:
            print("Error. Id: " , reqId, " Code: " , errorCode , " Msg: " , errorString)

    def positionEnd(self):
        self.disconnect()

ib_api = ib_class() 
ib_api.connect("127.0.0.1", 7496, 0) 
ib_api.reqPositions()
current_positions = ib_api.all_positions
ib_api.run()

When I run the code above I get a series of (i) account numbers, (ii) the contract symbol, (iii) position and (iv) average cost. This, therefore, answers question #1. You might want to "print" these values to see how IB's API send you the information.

I was also able to define a DataFrame variable all_positions that receives the information I was looking for. See the result below. Note that I had to create an "index" variable that is a unique identifier for each row of the DataFrame (as a combination of the account number and symbol). I didn't find a way to 'append' information to the DataFrame without this 'index' (any idea on how to do it would be welcome):

As for the last issue (the error messages):

Brian's suggestion of the "error" function (see below) got rid of the "-1" errors. But I still get the following:

unhandled exception in EReader thread Traceback (most recent call last): File "C:\Users\danil\Anaconda3\lib\site-packages\ibapi-9.76.1-py3.7.egg\ibapi\reader.py", line 34, in run data = self.conn.recvMsg() File "C:\Users\danil\Anaconda3\lib\site-packages\ibapi-9.76.1-py3.7.egg\ibapi\connection.py", line 99, in recvMsg buf = self._recvAllMsg() File "C:\Users\danil\Anaconda3\lib\site-packages\ibapi-9.76.1-py3.7.egg\ibapi\connection.py", line 119, in _recvAllMsg buf = self.socket.recv(4096) OSError: [WinError 10038] An operation was attempted on something that is not a socket

After a lot of experimentation, here is the final code I wrote, which I placed on a ".py" file named AB_API (so I can "import" it as a library):

from ibapi.client import EClient
from ibapi.wrapper import EWrapper
from ibapi.common import TickerId
from threading import Thread

import pandas as pd
import time

class ib_class(EWrapper, EClient):

    def __init__(self):
        EClient.__init__(self, self)

        self.all_positions = pd.DataFrame([], columns = ['Account','Symbol', 'Quantity', 'Average Cost', 'Sec Type'])
        self.all_accounts = pd.DataFrame([], columns = ['reqId','Account', 'Tag', 'Value' , 'Currency'])

    def error(self, reqId:TickerId, errorCode:int, errorString:str, advancedOrderRejectJson = ""):
        if reqId > -1:
            print("Error. Id: " , reqId, " Code: " , errorCode , " Msg: " , errorString)

    def position(self, account, contract, pos, avgCost):
        index = str(account)+str(contract.symbol)
        self.all_positions.loc[index]= account, contract.symbol, pos, avgCost, contract.secType
        
    def accountSummary(self, reqId, account, tag, value, currency):
        index = str(account)
        self.all_accounts.loc[index]=reqId, account, tag, value, currency

def connect_to_server(app):

    def run_loop():
        app.run()
    app.connect('127.0.0.1', 7496, 0)
    #Start the socket in a thread
    api_thread = Thread(target=run_loop, daemon=True)
    api_thread.start()
    time.sleep(0.5) #Sleep interval to allow time for connection to server


def read_positions(): #read all accounts positions and return DataFrame with information

    app = ib_class()    
    connect_to_server(app)

    app.reqPositions() # associated callback: position
    print("Waiting for IB's API response for accounts positions requests...\n")
    time.sleep(3)
    current_positions = app.all_positions # associated callback: position
    current_positions.set_index('Account',inplace=True,drop=True) #set all_positions DataFrame index to "Account"
    
    app.disconnect()

    return(current_positions)


def read_navs(): #read all accounts NAVs
    
    app = ib_class()  
    connect_to_server(app)

    app.reqAccountSummary(0,"All","NetLiquidation")  # associated callback: accountSummary / Can use "All" up to 50 accounts; after that might need to use specific group name(s) created on TWS workstation
    print("Waiting for IB's API response for NAVs requests...\n")
    time.sleep(3)
    current_nav = app.all_accounts
    
    app.disconnect()

    return(current_nav)

I can now read all accounts positions and NAVs with two single-line functions:

import IB_API

print("Testing IB's API as an imported library:")

all_positions = IB_API.read_positions()
all_navs = IB_API.read_navs()

If you want to run a very quick test (without importing/calling a function), just below the read_positions() and read_navs(), do the following:

print("Testing IB's API as an imported library:")

all_positions = read_positions()
print(all_positions, '\n')
all_navs = read_navs()
print(all_navs)

I realized some users were getting confused with the concept of importing functions, hence the quick suggestion above.

I wish Interactive Brokers had some simple examples available for clients that would make it easier for someone to test a few ideas and decide if they want to use their API or not.