Iteratively increase sample size for good/bad classification.

scripts/slave/recipe_modules/auto_bisect/resources/significantly_different.py

+"""Launches an anaconda environment to run some scipy hypothesis tests."""
+
+import json
+import os
+import subprocess
+import sys
+
+class ScipyNotInstalledError(Exception):
+  pass
+
+def main(argv, anaconda_path=None):
+  if not anaconda_path:
+    if os.name == 'nt':
+      anaconda_path = r'c:\conda-py-scientific\bin\python'

[Vadim Sh., 2016/02/12 00:17:12]

It is actually C:\conda-py-scientific\python.exe (no 'bin' on windows).

[Vadim Sh., 2016/02/12 00:36:14]

Oh, and also clear PYTHONPATH environ variable when calling another python, just
to avoid weird stuff when one environment tries to use modules from another
environment.

+    else:
+      anaconda_path = '/opt/conda-py-scientific/bin/python'
+  if not os.path.exists(anaconda_path):
+    raise ScipyNotInstalledError()
+
+  _, list_a, list_b, significance  = argv
+
+  inner_script_location = os.path.join(os.path.dirname(os.path.realpath(
+      __file__)), 'significantly_different_inner.py')
+
+  return json.loads(subprocess.check_output([
+      anaconda_path, inner_script_location,list_a, list_b, significance]))
+
+if __name__ == '__main__':
+  result = main(sys.argv)
+  if result:
+    print json.dumps(result, indent=4)
+    sys.exit(0)
+  sys.exit(1)