Python3 about construe

[tomas-teijeiro]

Unfortunately, we do not have the resources to maintain both Python 2 and Python 3 implementations, but in the near future we plan to migrate all the codebase to Python 3.

[lkjell]

I will port the code to python3 with 2to3 and fix some import problems. In addition pep8 formatting too.

[krivenkoz]

Hi guys! So what about Python3 support? Is it exist now?

[lkjell]

I should have a branch that support python3 at my repository.

[krivenkoz]

I didn't find. Could you provide a link to it?

[lkjell]

https://github.com/lkjell/construe/tree/python3

[krivenkoz]

Great, thanks

[krivenkoz]

When I ran run_example.sh I've got next error:
ModuleNotFoundError: No module named 'construe.utils.signal_processing.dtw.dtw
I've read about this issue in Know issues but I can't resolve it by now. My OS is Linux Debian 9. I could not install mlpy package due to 'Failed building wheel for mlpy' error. Any suggestions?

[lkjell]

Have you read the readme file in that folder?

[lkjell]

https://github.com/lkjell/construe/blob/652396a3378277cb473e858af1b2f7cb2be9cab1/construe/utils/signal_processing/dtw/readme

[krivenkoz]

My bad, thank you. It works for me now after replacing some methods in plotter.py to modern one(nodes_iter() to nodes(), set_axis_bgcolor() to set_facecolor()). What about the Windows 10 x64? Do I have a chance to launch Construe? At the moment I have errors with dtw,pygraphviz and some else.

[hsiaocy]

Hi @lkjell, thanks for your job on Python3 version!
I want to help and use the MIT-LCP's wfdb-python repository as a pure python environment, and decrease the dependency problems.
While I ran the construe_ecg.py in your repository, the error pop up with a "Not Implemented Error".
Could you please help me to fix this issue?

The error context happens from this function "process_record_rhythm()" in construe_ecg.py module and so on:
`File "/Users/AppleUser/PycharmProjects/construe-python3/construe_ecg.py", line 256, in
Rec.exclude_twaves, Rec.verbose)

File "/Users/AppleUser/PycharmProjects/construe-python3/record_processing.py", line 498, in process_record_rhythm
cntr.step(filt)

File "/Users/AppleUser/PycharmProjects/construe-python3/construe/inference/searching.py", line 182, in step
self._update_closed(newclosed)

File "/Users/AppleUser/PycharmProjects/construe-python3/construe/inference/searching.py", line 132, in _update_closed
self.closed.append(tmplst.pop(0))

File "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/sortedcontainers/sortedlist.py", line 1279, in append
raise NotImplementedError('use sl.add(value) instead')
NotImplementedError: use sl.add(value) instead`

Please ask if any question. Thanks!

[lkjell]

You could try to monkey patch th code and use the method add instead of append for self.closed. Not sure why that happen. Could be later version of sortedcontainer breaks the api.

[hsiaocy]

Sorry, I didn't notice the replacement message 'use sl.add(value) instead'. It works with following:
self.closed.append(tmplst.pop(0))
change to --> self.closed.add(tmplst.pop(0))

[karimkhanp]

@lkjell Thanks for python3 version of construe.

I installed all dependencies in my Mac machine python3 venv.

Now when I try to run

./run_example.sh fig4

It gives me an error with python2 code lines -

Running: python -i fragment_processing.py -r examples/fig4
Traceback (most recent call last):
  File "fragment_processing.py", line 12, in <module>
    import construe.utils.plotting.plotter as plotter
  File "/Users/construe/construe/construe/utils/plotting/plotter.py", line 11, in <module>
    import construe.knowledge.observables as o
  File "/Users/construe/construe/construe/knowledge/observables/__init__.py", line 7, in <module>
    from .CardiacCycles import *
  File "/Users/construe/construe/construe/knowledge/observables/CardiacCycles.py", line 11, in <module>
    from construe.model import Observable, singleton_observable
  File "/Users/construe/construe/construe/model/__init__.py", line 5, in <module>
    from .abstraction_pattern import AbstractionPattern
  File "/Users/construe/construe/construe/model/abstraction_pattern.py", line 13, in <module>
    from .constraint_network import ConstraintNetwork, InconsistencyError, verify
  File "/Users/construe/construe/construe/model/constraint_network.py", line 429
    print 'v0:' + str(v0) + ' v1:' + str(v1) +\
              ^
SyntaxError: invalid syntax

I converted to python3 print syntax and run it. So it gave another error

Traceback (most recent call last): File "fragment_processing.py", line 12, in <module> import construe.utils.plotting.plotter as plotter File "/Users/construe/construe/construe/utils/plotting/plotter.py", line 12, in <module> import construe.knowledge.abstraction_patterns as ap File "/Users/construe/construe/construe/knowledge/abstraction_patterns/__init__.py", line 6, in <module> from .rhythm import (RHYTHMSTART_PATTERN, SINUS_PATTERN, TACHYCARDIA_PATTERN, File "/Users/construe/construe/construe/knowledge/abstraction_patterns/rhythm/__init__.py", line 11, in <module> from .patterns import * File "/Users/construe/construe/construe/knowledge/abstraction_patterns/rhythm/patterns.py", line 13, in <module> from construe.knowledge.constants import (PW_DURATION, ST_INTERVAL, File "/Users/construe/construe/construe/knowledge/constants.py", line 12, in <module> import construe.acquisition.signal_buffer as SIG File "/Users/construe/construe/construe/acquisition/signal_buffer.py", line 11, in <module> import segmentation.wavelets.wavelet_filter as wf ModuleNotFoundError: No module named 'segmentation'

[lkjell]

@karimkhanp perhaps you are not on the python3 branch.

[lkjell]

@karimkhanp It is hard to tell. But you might try python 3.5 and use packages which was released at Dec 10, 2017. I suspect that some dependency packages changed their API. Beside the classifier for detecting P wave is quite old. Last time I talked to @tomas-teijeiro he did not have the training data.

[tomas-teijeiro]

Dear all,

We just released a new version of Construe, with all code translated to Python 3. The interpretation algorithms have also been refactored to reduce the computing requirements, so we suggest you to move to the new version. @karimkhanp our last tests were made with python 3.6 and everything worked as expected, but I will try with python 3.7 and see if there are some regression issues. Thanks @lkjell for your help!