search

convexengineering / gpkit

Geometric programming for engineers
http://gpkit.readthedocs.org
MIT License
203 stars 40 forks source link

Vector substitution unit test error #1576

Open whoburg opened 9 months ago

whoburg commented 9 months ago

Found two failing tests on a fresh install. One of is already captured in #1573; the other one in test_vector_sweep seems new.

Per the instructions printed after running import gpkit.tests, here is the output:

Found no installed solvers, beginning a build.
# Building GPkit version 1.1
# Moving to the directory from which GPkit was imported.

Attempting to find and build solvers:

# Looking for `mosek_cli`
#   (A "success" is if mskexpopt complains that
#    we haven't specified a file for it to open.)
#     Calling 'mskexpopt'
##
### CALL BEGINS
/bin/sh: mskexpopt: command not found
### CALL ENDS
##
# Looks like `mskexpopt` was not found in the default PATH,
#  so let's try locating that binary ourselves.
# expected MOSEK directory not found: /Users/serenity/mosek
# Did not find
# mosek_cli

# Looking for `mosek_conif`
#   Trying to import mosek...
# Did not find
# mosek_conif

# Looking for `cvxopt`
#   Trying to import cvxopt...

Found cvxopt in the default PYTHONPATH

Found the following solvers: cvxopt
#     Replacing directory env
..........................E...............................................................................................................................................E......................................
======================================================================
ERROR: test_vector_sweep (gpkit.tests.t_sub.TestModelSubs)
Test sweep involving VectorVariables
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/serenity/code/gpkit/gpkit/tests/t_sub.py", line 205, in test_vector_sweep
    sol = m.solve(verbosity=0)
  File "/Users/serenity/code/gpkit/gpkit/constraints/prog_factories.py", line 123, in solvefn
    constants, sweep, linked = parse_subs(self.varkeys, self.substitutions)
  File "/Users/serenity/code/gpkit/gpkit/nomials/substitution.py", line 25, in parse_subs
    append_sub(sub, keys, constants, sweep, linkedsweep)
  File "/Users/serenity/code/gpkit/gpkit/nomials/substitution.py", line 47, in append_sub
    sub = np.array(sub) if not hasattr(sub, "shape") else sub
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part.

======================================================================
ERROR: test_breakdowns_cvxopt (gpkit.tests.t_examples.TestExamples)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/serenity/code/gpkit/gpkit/tests/helpers.py", line 55, in test
    testfn(name, import_dict, path)(self)
  File "/Users/serenity/code/gpkit/gpkit/tests/helpers.py", line 90, in test
    imported[name] = importlib.import_module(name)
  File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.8/lib/python3.8/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1014, in _gcd_import
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 783, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/Users/serenity/code/gpkit/docs/source/examples/breakdowns.py", line 11, in <module>
    sol = pickle.load(open(dirpath+"solar_13.p", "rb"))
ModuleNotFoundError: No module named 'pint.quantity'

----------------------------------------------------------------------
Ran 209 tests in 20.246s

FAILED (errors=2)

GPkit is now installed with solver(s) ['cvxopt']
To incorporate new solvers at a later date, run `gpkit.build()`.

If any tests didn't pass, please post the output above
(starting from "Found no installed solvers, beginning a build.")
to gpkit@mit.edu or https://github.com/convexengineering/gpkit/issues/new
so we can prevent others from having these errors.
whoburg commented 9 months ago

Updated title to be about the vector substitution failure only -- #1573 captures the other failure.

I can work on fixing this but I'm unclear on the intent of how it should work. The model is 

x = Variable("x")
x_min = Variable("x_min", 1)
y = VectorVariable(2, "y")
m = Model(x, [x >= y.prod()])

There are several different substitutions in test_vector_sweep; the first few are:

{y: ('sweep', [[2, 3], [5, 7], [9, 11]])}

(three different values for the vector of length 2, produces costs [6, 15, 27, 14, 35, 63, 22, 55, 99])

{y: ('sweep', [[2, 3], [5, 7, 11]])}

(the one that's failing with ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part. -- intended to produce costs [10, 15, 14, 21, 22, 33]),

{y: ('sweep', [[2, 3, 9], [5, 7, 11]])}

(wrong shape, intended to throw a ValueError) ... and a few more.

I'd support a solution that's more restrictive on shapes of substitutions that are supported - seems surprising to allow the inhomogeneous shape.