add a jlpm "bootstrap" entrypoint that can run during eg pip install of jupyterlab

.gitignore

@@ -87,3 +87,8 @@ ENV/
 
 # Rope project settings
 .ropeproject
+
+# ms IDE stuff
+*.code-workspace
+.history
+.vscode

jupyter_packaging/jlpmapp.py

@@ -0,0 +1,43 @@
+# coding: utf-8
+"""A Jupyter-aware wrapper for the yarn package manager"""
+
+# Copyright (c) Jupyter Development Team.
+# Distributed under the terms of the Modified BSD License.
+import sys
+
+import os
+from jupyterlab_server.process import which, subprocess
+
+HERE = os.path.dirname(os.path.abspath(__file__))
+YARN_PATH = os.path.join(HERE, 'yarn.js')
+
+
+def execvp(cmd, argv):
+    """Execvp, except on Windows where it uses Popen.
+
+    The first argument, by convention, should point to the filename
+    associated with the file being executed.
+
+    Python provides execvp on Windows, but its behavior is problematic
+    (Python bug#9148).
+    """
+    cmd = which(cmd)
+    if os.name == 'nt':
+        import signal
+        import sys
+        p = subprocess.Popen([cmd] + argv[1:])
+        # Don't raise KeyboardInterrupt in the parent process.
+        # Set this after spawning, to avoid subprocess inheriting handler.
+        signal.signal(signal.SIGINT, signal.SIG_IGN)
+        p.wait()
+        sys.exit(p.returncode)
+    else:
+        os.execvp(cmd, argv)
+
+
+def main(argv=None):
+    """Run node and return the result.
+    """
+    # Make sure node is available.
+    argv = argv or sys.argv[1:]
+    execvp('node', ['node', YARN_PATH] + argv)