Command not found will no longer traceback

Previously, when a command was not found, Jupyter printed a traceback on the user:

    $ jupyter foobar
    Traceback (most recent call last):
      File "/usr/bin/jupyter", line 33, in <module>
        sys.exit(load_entry_point('jupyter-core==4.6.3', 'console_scripts', 'jupyter')())
      File "/usr/lib/python3.9/site-packages/jupyter_core/command.py", line 247, in main
        command = _jupyter_abspath(subcommand)
      File "/usr/lib/python3.9/site-packages/jupyter_core/command.py", line 133, in _jupyter_abspath
        raise Exception(
    Exception: Jupyter command `jupyter-foobar` not found.

Generally, I believe users should only see tracebacks when they are debugging
stuff or when there is an unexpected "crash".

In Fedora, such traceback is detected as crash by the automatic bug reporting tool.

Now, an error message is printed to stderr instead to provide a better UX.

Fixes #211

jupyter_core/command.py

@@ -282,7 +282,10 @@ def main():
         parser.print_usage(file=sys.stderr)
         sys.exit("subcommand is required")
 
-    command = _jupyter_abspath(subcommand)
+    try:
+        command = _jupyter_abspath(subcommand)
+    except Exception as e:
+        sys.exit(e)
 
     try:
         _execvp(command, sys.argv[1:])

jupyter_core/tests/test_command.py

@@ -4,7 +4,7 @@
 import os
 import sys
 import sysconfig
-from subprocess import check_output, CalledProcessError
+from subprocess import check_output, PIPE, CalledProcessError
 from unittest.mock import patch
 
 import pytest
@@ -21,7 +21,7 @@ def get_jupyter_output(cmd):
     """Get output of a jupyter command"""
     if not isinstance(cmd, list):
         cmd = [cmd]
-    return check_output([sys.executable, '-m', 'jupyter_core'] + cmd).decode('utf8').strip()
+    return check_output([sys.executable, '-m', 'jupyter_core'] + cmd, stderr=PIPE).decode('utf8').strip()
 
 
 def write_executable(path, source):
@@ -109,8 +109,10 @@ def test_help():
 
 
 def test_subcommand_not_found():
-    with pytest.raises(CalledProcessError):
+    with pytest.raises(CalledProcessError) as excinfo:
         get_jupyter_output('nonexistant-subcommand')
+    stderr = excinfo.value.stderr.decode('utf8')
+    assert stderr == 'Jupyter command `jupyter-nonexistant-subcommand` not found.\n'
 
 @patch.object(sys, 'argv', [__file__] + sys.argv[1:])
 def test_subcommand_list(tmpdir):