Compilation error and dependencies

[sgyzetrov]

Hi @bmmalone :

I am currently working through the paper Learning Optimal Bayesian Networks: A Shortest Path Perspective, but code included in your paper (urlearning-cpp-8-24-2014 -> urlearning-cpp-673f203f44b9) can not calculate score. So I did some digging and found a bitbucket repo and this Github repo. Few questions:

N.B. I am using a docker image for urlearning testing, the image is Ubuntu 18.04

1. This repo no longer seem active, and so was the bitbucket repo. Which one is the latest version? I am guessing this repo and rather than the one hosting on bitbucket?
2. Got the same compilation errors stated in Issues compiling score #27 and some more, @bmmalone could you be kind enough to specify dependencies? I got boost_1_63 from your bazel file but did not catch your gcc/g++ version for compilation. Which version did you use?
3. I did managed to compile the whole bitbucket project before (through replacing buggy files with files on Github). But I was unable to calculate score. Some weird boost error appears:

   $> score soybean.csv soybean.pss
   ...
   score: /usr/include/boost/dynamic_bitset/dynamic_bitset.hpp: boost::dynamic_bitset<Block, Allocator>::size_type, bool) [with Block = long unsigned int; Allocator = std::allocator<long unsigned int>; boost::dynamic_bitset<Block, Allocator>::size_type = long unsigned int]: Assertion `pos < m_num_bits' failed.
   Aborted
   

   Have you ever been able to use score and had no trouble calculating .csv files at all? Which .csv data did you try? Because it seems to me score is currently unusable...

Looking forward for your reply, thanks in advance!

[bmmalone]

Hey @sgyzetrov ,

As you noticed, we are no longer maintaining these repositories, although the github version is more recent. The latest "stable" version is available here: http://urlearning.org/download.html, but it is quite old (also, it is "research stable").

but did not catch your gcc/g++ version for compilation

Unfortunately, I don't remember which version, and I no longer have access to the servers where I ran experiments. Based on the release history (https://gcc.gnu.org/releases.html, around 2015), I would guess we were using something in the 4.8 series, or one of the early version 5s.

Have you ever been able to use score and had no trouble calculating .csv files at all?

At the time, yes, it worked fine. Please see the example (data/hepatitis.clean.csv) for one that should work. Internally, we use dynamic bitsets to represent variable sets, but I'm not sure what that exact error message means.

---

I'm sorry I can't offer more help. If you haven't already tried it, you can also try the dev branch; it includes some more changes, and I believe it is more stable than the master branch.

I won't be able to help much with the code, but if you do have specific questions, issues, etc., please don't hesitate to contact me. I will try to help as much as possible.

[sgyzetrov]

Thank you for the quick response @bmmalone !

The latest "stable" version is available here: http://urlearning.org/download.html

Just to clarify, by that you mean this tar file http://urlearning.org/urlearning-files/urlearning-cpp.tar.gz right?

Thanks again!

[bmmalone]

Hey @sgyzetrov ,

Yes, that one should be stable.

Also, in case you are looking for implementations of things in our later papers, many of them are here: https://bitbucket.org/bmmalone/urlearning-python. The code is again not terribly robust, but at least it doesn't need to be compiled ;)