object of type 'generator' has no len() about 3d-bpp

[kiesoman]

Hi,
It is workable to install the newest version of OR-Tools in environment.
However, you should change the the content around 63 line in baseline.py as below:
cp_model.LinearExpr.Sum(
[fsi[s, i] * zsl[s, l] for s in range(n_superitems) for l in range(max_layers)]
)
the argument should be a list, and then it will fit len() requirement.

you also need to change 77 line and 114 line.
enjoy.

[Wadaboa]

Hi,

Thanks for opening the issue.

Could you please dump some information about your Python environment? Also, what is the actual exception raised by the code? I cannot see it in the stacktrace you pasted above.

Thanks,
Alessio.

[rlacjfjin]

You can rewrite the wrong part, which may be caused by the inconsistent version of ortools. (my ortools version is 9.3.10497)
I have found problems involving the cp_model.LinearExpr.Sum() function, which you can simply write as sum().

I hope to owner (@Wadaboa) provide an environment in which the project can run normally, such as requirements.txt or environment.yml of conda.

And I have a question:
You implemented it in the form of constraint programming, but is also can using MIP form, I think two ways are same.
Is there any reason for you to choose constraint programming? (For example, constraint programming solver is faster?)

Looking forward to your reply,
Thanks.

Zhe

[Wadaboa]

Hi @rlacjfjin,

In the init/ folder you can find 2 files: requirements.txt and environment.yml. The former is for pip environments, while the latter for Conda environments. The README also has a brief Installation section about this. As you can see in the mentioned files, the version of OR-Tools we relied on is 8.2.8710, so I expect the exception above to disappear when the environment is correctly set up. Please let me know if this isn't the case.

As for your question, we implemented the sub-problem as both CP and MIP for education purposes and also because of the additional utilities available in the CP solver, such as decision strategies and search branching methods. In our final tests, we went for MIP in the "no-placement" sub-problem and CP in the "placement" sub-problem: we experimented with different combinations and this one seemed to be the fastest, but this might vary depending on the data you use.

Let me know if you have any other question. Thanks,
Alessio.

[rlacjfjin]

@Wadaboa ,
Thank you, I have no more questions.