Schema fails when unhashable values within lists

[tcmoore3]

Description

When statepoints contain values that are lists containing dicts, a TypeError is raised when detect_schema() is called on the project.

To reproduce

import signac


pr = signac.init_project('gonna_fail')
j = pr.open_job({'oh': [{'hi': 'mark'}]}).init()
pr.detect_schema()

Error output

Traceback (most recent call last):
  File "reproduce-error.py", line 6, in <module>
    pr.detect_schema()
  File "/Users/tcmoore3/programs/signac/signac/contrib/project.py", line 507, in detect_schema
    return ProjectSchema.detect(statepoint_index)
  File "/Users/tcmoore3/programs/signac/signac/contrib/schema.py", line 63, in detect
    return cls({k: _collect_by_type(v) for k, v in statepoint_index})
  File "/Users/tcmoore3/programs/signac/signac/contrib/schema.py", line 63, in <dictcomp>
    return cls({k: _collect_by_type(v) for k, v in statepoint_index})
  File "/Users/tcmoore3/programs/signac/signac/contrib/project.py", line 481, in build_job_statepoint_index
    for x in _build_job_statepoint_index(jobs=self, exclude_const=exclude_const, index=index):
  File "/Users/tcmoore3/programs/signac/signac/contrib/schema.py", line 38, in _build_job_statepoint_index
    collection.index(key, build=True)
  File "/Users/tcmoore3/programs/signac/signac/contrib/collection.py", line 458, in index
    self._build_index(key)
  File "/Users/tcmoore3/programs/signac/signac/contrib/collection.py", line 415, in _build_index
    self._indexes[key] = _build_index(self._docs.values(), key, self._primary_key)
  File "/Users/tcmoore3/programs/signac/signac/contrib/collection.py", line 193, in _build_index
    index[_to_tuples(v)].add(doc[primary_key])
  File "/Users/tcmoore3/programs/signac/signac/contrib/collection.py", line 155, in __getitem__
    _float(key) if type(key) is float else key)
TypeError: unhashable type: 'dict'

System configuration

Please complete the following information:

• Operating System: macOS 10.13.6
• Version of Python: 3.5.5
• Version of signac: 1.2.0

[tcmoore3]

A TypeError is also raised if a value contains a set within a list, but not if the value contains a list within a list.

import signac


pr = signac.init_project('gonna_fail')
j = pr.open_job({'oh': [{0, 0, 1, 2, 3, 3, 3}]}).init()
pr.detect_schema()

Traceback (most recent call last):
  File "reproduce-error.py", line 5, in <module>
    j = pr.open_job({'oh': [{0, 0, 1, 2, 3, 3, 3}]}).init()
  File "/Users/tcmoore3/programs/signac/signac/contrib/project.py", line 368, in open_job
    job = self.Job(project=self, statepoint=statepoint)
  File "/Users/tcmoore3/programs/signac/signac/contrib/job.py", line 61, in __init__
    self._id = calc_id(self._statepoint()) if _id is None else _id
  File "/Users/tcmoore3/programs/signac/signac/contrib/hashing.py", line 12, in calc_id
    blob = json.dumps(spec, sort_keys=True)
  File "/Users/tcmoore3/anaconda/lib/python3.5/json/__init__.py", line 237, in dumps
    **kw).encode(obj)
  File "/Users/tcmoore3/anaconda/lib/python3.5/json/encoder.py", line 198, in encode
    chunks = self.iterencode(o, _one_shot=True)
  File "/Users/tcmoore3/anaconda/lib/python3.5/json/encoder.py", line 256, in iterencode
    return _iterencode(o, 0)
  File "/Users/tcmoore3/anaconda/lib/python3.5/json/encoder.py", line 179, in default
    raise TypeError(repr(o) + " is not JSON serializable")
TypeError: {0, 1, 2, 3} is not JSON serializable

[vyasr]

@tcmoore3 you're tearing me apart