t4d-gmbh / majortrack Goto Github PK

Since python 3.8 importing MajorTrack fails with:

Traceback (most recent call last):
  File ",,,/test.py", line 1, in <module>
    from majortrack import MajorTrack
  File ".../.virtualenvs/py310/lib/python3.10/site-packages/majortrack/__init__.py", line 9, in <module>
    from .tracker import MajorTrack
  File ".../.virtualenvs/py310/lib/python3.10/site-packages/majortrack/tracker.py", line 11, in <module>
    from .lazylist import lazy_list
  File ".../.virtualenvs/py310/lib/python3.10/site-packages/majortrack/lazylist.py", line 4, in <module>
    from collections import MutableSequence
ImportError: cannot import name 'MutableSequence' from 'collections' (/usr/lib/python3.10/collections/__init__.py)

Classes from collections are used in

• majortrack/lazylist.py
• majortrack/tracker.py

Solution

The alias collections.MutableSequence which was created when the class moved to collections.abc in 3.3 was removed with python 3.8. So the import needs updating, to be compatible with recent python versions, but should not break backwards compatibility.

Suggestion

try:
    from collections.abc import MutableSequence  # noqa
except ImportError:
    from collections import MutableSequence  # noqa

Currently, the only out of the box options for labels that are provided are 'groupsize' and 'group_index'.

'group_index' might be helpful for debugging, and for doing custom layout adaptations (see e.g. line 126 in the history.py example and the optional y_fix parameter in AlluvialPlot) but the group IDs can be misleading since they are a priori completely unrelated between slices.

Suggestion:

Adding the option 'dc_index' to the set of available cluster labels, such that the dynamic cluster affiliation can also be shown by the label of a group and not just it's color.

Providing an iterator with an initial element bigger than 0 leads to an IndexError when attempting to create fluxes between the clusters to display.

The error occurs in https://github.com/j-i-l/MajorTrack/blob/11f92c5fadb8a15c2672c750371df83847d90ca1/majortrack/tracker.py#L2509-L2510

Problem:

The elements to draw (clusters and fluxes) for each slice of the time series are collected in lists rather than in dictionaries. The elements are collected by iterating over the specified iterator and appending to these lists. If the iterator does not start at 0 this creates a miss-match between the index of the slices to draw and the index of the collections of elements to draw for that index.

Fix:

option a): Use the first element of the specified iterator as offset when accessing the elements in to draw.

option b): collect the elements to draw in dict's rather than list's and use the elements from the iterator parameter as keys.