带插座连接的断管

Question

I am using a socket connection to download data through a third party API. It works fine for a while but every now and then my script will crash giving the following error: BrokenPipeError: [Errno 32] Broken pipe

After some research it seems the suggestion ( link here ) is to do the following:

from signal import signal, SIGPIPE, SIG_DFL

signal(SIGPIPE,SIG_DFL)

However im firstly not sure what this actually does (im still confused after reading the python manual on signal). And I also don't know where to put the code.

If anyone is familiar with this error please could you advise if this is infact the correct solution and where the signal(SIGPIPE,SIG_DFL) would be placed. Should there be a try/except block inside which this is placed, or is it simply placed at the start of the program? Im confused.

Here's some of the relevant code. I basically have a dataframe consisting of several thousand items. I loop through each item passing it to the download method. The download method downloads the data via the api and then writes it to a database. I then move to the next item to download.

def recv_data(sock, recv_buffer=4096, delim='\n'):

    buffer = ''
    data = True
    while data:
        data = sock.recv(recv_buffer)
        buffer += str(data.decode('latin-1'))

        while buffer.find(delim) != -1:
            line, buffer = buffer.split('\n', 1)
            yield line

def update_existing_symbol_data(engine, sock, exchange, exchange_id, symbol, symbol_id, start_date):
    data = ''
    message = #request data message
    sock.sendall(message.encode())

    for line in recv_data(sock):
        if "!ENDMSG!" in line:
            break
        data += line[:-2] + '\n'

    df = pd.read_csv(io.StringIO(data))
    df.set_index('date', inplace=True)
    df.to_sql('daily', engine, if_exists='append')

def main():
    df = #dataframe all symbols that need to be downloaded
    for index, row in df.iterrows():
        update_existing_symbol_data(args)

Answer 1

SIGPIPE is a POSIX thing that gets sent when a socket write operation fails. The default behavior is for the signal (this is an OS/socket thing, not a Python thing) to just kill your process. Python instead gives it to you as an exception so that it's possible to write more robust programs. But if you don't need to handle that event, which it sounds like you don't considering your use case, you can safely ignore it. There's no logic you need to do when you receive the signal, so the solution from that blog post should be fine. No try/except needed.

If your use case changes at a later date and you do need to handle the SIGPIPE, then wrapping that in a try/except and handling it there would be the way to go.