Merge tag 'v0.5' into debian

The first release after major refactoring:

- Refactored into a proper `heudiconv` Python module
  - `heuristics` is now a `heudiconv.heuristics` submodule
  - you can specify shipped heuristics by name (e.g. `-f reproin`)
    without providing full path to their files
  - you need to use `--files` (not just positional argument(s)) if not
    using `--dicom_dir_templates` or `--subjects` to point to data files
    or directories with input DICOMs
- `Dockerfile` is generated by [neurodocker](https://github.com/kaczmarj/neurodocker)
- Logging verbosity reduced
- Increased leniency with missing DICOM fields
- `dbic_bids` heuristic renamed into reproin
- [LICENSE](https://github.com/nipy/heudiconv/blob/master/LICENSE)
  with Apache 2.0 license for the project
- [CHANGELOG.md](https://github.com/nipy/heudiconv/blob/master/CHANGELOG.md)
- [Regression testing](https://github.com/nipy/heudiconv/blob/master/tests/test_regression.py) on real data (using datalad)
- A dedicated [ReproIn](https://github.com/repronim/reproin) project
  with details about ReproIn setup/specification and operation using
  `reproin` heuristic shipped with heudiconv
- [utils/test-compare-two-versions.sh](utils/test-compare-two-versions.sh)
  helper to compare conversions with two different versions of heudiconv
- Support for converters other than `dcm2niix`, which is now the default.
  Explicitly specify `-c none` to only prepare conversion specification
  files without performing actual conversion
- Compatibility with Nipype 1.0, PyDicom 1.0, and upcoming DataLad 0.10
- Consistency with converted files permissions
- Ensured subject id for BIDS conversions will be BIDS compliant
- Re-add `seqinfo` fields as column names in generated `dicominfo`
- More robust sanity check of the regex reformatted .json file to avoid
  numeric precision issues
- Many other various issues

* tag 'v0.5':
  ENH: skip test_conversion if test data is not fetchable
  BF: skip testing the monitor code if tinydb is not available
  BF: do not ship "test" package in binary distribution

Author: yarikoptic

setup.py

@@ -31,6 +31,11 @@ def findsome(subdir, extensions):
             f.split(op.sep, 1)[1] for f in findall(subdir)
             if op.splitext(f)[-1].lstrip('.') in extensions
         ]
+    # Only recentish versions of find_packages support include
+    # heudiconv_pkgs = find_packages('.', include=['heudiconv*'])
+    # so we will filter manually for maximal compatibility
+    heudiconv_pkgs = [pkg for pkg in find_packages('.') if pkg.startswith('heudiconv')]
+
 
     setup(
         name=ldict['__packagename__'],
@@ -40,7 +45,7 @@ def findsome(subdir, extensions):
         description=ldict['__description__'],
         long_description=ldict['__longdesc__'],
         license=ldict['__license__'],
-        packages=find_packages(),
+        packages=heudiconv_pkgs,
         entry_points={'console_scripts': [
             'heudiconv=heudiconv.cli.run:main',
             'heudiconv_monitor=heudiconv.cli.monitor:main',

tests/test_monitor.py

@@ -3,7 +3,10 @@
 import os.path as op
 import pytest
 from mock import patch
-from tinydb import TinyDB, Query
+try:
+    from tinydb import TinyDB, Query
+except ImportError:
+    pytest.importorskip("tinydb")
 from subprocess import CalledProcessError
 
 try:

tests/test_regression.py

@@ -7,6 +7,7 @@
 have_datalad = True
 try:
     from datalad import api # to pull and grab data
+    from datalad.support.exceptions import IncompleteResultsError
 except ImportError:
     have_datalad = False
 
@@ -21,7 +22,10 @@
 @pytest.mark.skipif(not have_datalad, reason="no datalad")
 def test_conversion(tmpdir, subject, heuristic):
     tmpdir.chdir()
-    datadir = fetch_data(tmpdir.strpath, subject)
+    try:
+        datadir = fetch_data(tmpdir.strpath, subject)
+    except IncompleteResultsError as exc:
+        pytest.skip("Failed to fetch test data: %s" % str(exc))
     outdir = tmpdir.mkdir('out').strpath
 
     args = gen_heudiconv_args(datadir, outdir, subject, heuristic)