[python] How to catch exception output from Python subprocess.check_output()?

I'm trying to do a Bitcoin payment from within Python. In bash I would normally do this:

bitcoin sendtoaddress <bitcoin address> <amount>

so for example:

bitcoin sendtoaddress 1HoCUcbK9RbVnuaGQwiyaJGGAG6xrTPC9y 1.4214

if it is successfull I get a transaction id as output but if I try to transfer an amount larger than my bitcoin balance, I get the following output:

error: {"code":-4,"message":"Insufficient funds"}

In my Python program I now try to do the payment as follows:

import subprocess

try:
    output = subprocess.check_output(['bitcoin', 'sendtoaddress', address, str(amount)])
except:
    print "Unexpected error:", sys.exc_info()

If there's enough balance it works fine, but if there's not enough balance sys.exc_info() prints out this:

(<class 'subprocess.CalledProcessError'>, CalledProcessError(), <traceback object at 0x7f339599ac68>)

It doesn't include the error which I get on the command line though. So my question is; how can I get the outputted error ({"code":-4,"message":"Insufficient funds"}) from within Python?

All tips are welcome!

Trying to "transfer an amount larger than my bitcoin balance" is not an unexpected error. You could use Popen.communicate() directly instead of check_output() to avoid raising an exception unnecessarily:

from subprocess import Popen, PIPE

p = Popen(['bitcoin', 'sendtoaddress', ..], stdout=PIPE)
output = p.communicate()[0]
if p.returncode != 0: 
   print("bitcoin failed %d %s" % (p.returncode, output))

As mentioned by @Sebastian the default solution should aim to use run(): https://docs.python.org/3/library/subprocess.html#subprocess.run

Here a convenient implementation (feel free to change the log class with print statements or what ever other logging functionality you are using):

import subprocess

def _run_command(command):
    log.debug("Command: {}".format(command))
    result = subprocess.run(command, shell=True, capture_output=True)
    if result.stderr:
        raise subprocess.CalledProcessError(
                returncode = result.returncode,
                cmd = result.args,
                stderr = result.stderr
                )
    if result.stdout:
        log.debug("Command Result: {}".format(result.stdout.decode('utf-8')))
    return result

And sample usage (code is unrelated, but I think it serves as example of how readable and easy to work with errors it is with this simple implementation):

try:
    # Unlock PIN Card
    _run_command(
        "sudo qmicli --device=/dev/cdc-wdm0 -p --uim-verify-pin=PIN1,{}"
        .format(pin)
    )

except subprocess.CalledProcessError as error:
    if "couldn't verify PIN" in error.stderr.decode("utf-8"):
        log.error(
                "SIM card could not be unlocked. "
                "Either the PIN is wrong or the card is not properly connected. "
                "Resetting module..."
                )
        _reset_4g_hat()
        return

Based on the answer of @macetw I print the exception directly to stderr in a decorator.

Python 3

from functools import wraps
from sys import stderr
from traceback import format_exc
from typing import Callable, Collection, Any, Mapping


def force_error_output(func: Callable):
    @wraps(func)
    def forced_error_output(*args: Collection[Any], **kwargs: Mapping[str, Any]):
        nonlocal func

        try:
            func(*args, **kwargs)
        except Exception as exception:
            stderr.write(format_exc())
            stderr.write("\n")
            stderr.flush()

            raise exception

    return forced_error_output

Python 2

from functools import wraps
from sys import stderr
from traceback import format_exc


def force_error_output(func):
    @wraps(func)
    def forced_error_output(*args, **kwargs):
        try:
            func(*args, **kwargs)
        except Exception as exception:
            stderr.write(format_exc())
            stderr.write("\n")
            stderr.flush()

            raise exception

    return forced_error_output

Then in your worker just use the decorator

@force_error_output
def da_worker(arg1: int, arg2: str):
    pass

This did the trick for me. It captures all the stdout output from the subprocess(For python 3.8):

from subprocess import check_output, STDOUT
cmd = "Your Command goes here"
try:
    cmd_stdout = check_output(cmd, stderr=STDOUT, shell=True).decode()
except Exception as e:
    print(e.output.decode()) # print out the stdout messages up to the exception
    print(e) # To print out the exception message

I don't think the accepted solution handles the case where the error text is reported on stderr. From my testing the exception's output attribute did not contain the results from stderr and the docs warn against using stderr=PIPE in check_output(). Instead, I would suggest one small improvement to J.F Sebastian's solution by adding stderr support. We are, after all, trying to handle errors and stderr is where they are often reported.

from subprocess import Popen, PIPE

p = Popen(['bitcoin', 'sendtoaddress', ..], stdout=PIPE, stderr=PIPE)
output, error = p.communicate()
if p.returncode != 0: 
   print("bitcoin failed %d %s %s" % (p.returncode, output, error))

There are good answers here, but in these answers, there has not been an answer that comes up with the text from the stack-trace output, which is the default behavior of an exception.

If you wish to use that formatted traceback information, you might wish to:

import traceback

try:
    check_call( args )
except CalledProcessError:
    tb = traceback.format_exc()
    tb = tb.replace(passwd, "******")
    print(tb)
    exit(1)

As you might be able to tell, the above is useful in case you have a password in the check_call( args ) that you wish to prevent from displaying.