Adding the C-API functions.

[DanielEven]

Copied the kmeans code and added functions for the 2 module functions.

[DanielEven]

We still need to do:

• [x] Pass all the sizes (n,k,num_mapped_vectors) to the kmeans functions somehow.
• [x] Solve memory management in kmeans - returning NULL and checking return result.
• [x] Modify kmeans to get the right arguments.
• [x] Handle the errors in the C-API.

@liammohr

[DanielEven]

I changes the location of these functions, do you agree ? @liammohr

[DanielEven]

Sorry for the typing mistake.

[DanielEven]

We also need to:

• [x] Implement calculate_requested in spkmeans.c.

[DanielEven]

Don't we prefer to pass the sizes (n,k,num_mapped_vectors) through a global variable ? Because pass them many times and they are always staying the same, so we can initialze once and then use them globally.

@liammohr

[DanielEven]

The next step is to finish calculate requested (and modify create_k_eigenvectors_matrix in order to do that).

[liammohr]

The C-API works! I finished calculate_requested and adjusted create_k_eigenvectors_matrix. I also debugged the code for hours until I finally found all the bugs (hopefully), and it runs smoothly and outputs correct results for the blobs1 file. Two more things to do before merging: (additional to the ones listed above in this discussion)

• [x] check what exactly to print for ddg input
• [x] check what format to print for jacobi input - vectors as row or columns

(The output is mathematically correct, we just need to decide on the format)

[liammohr]

The C-API works! I finished calculate_requested and adjusted create_k_eigenvectors_matrix. I also debugged the code for hours until I finally found all the bugs (hopefully), and it runs smoothly and outputs correct results for the blobs1 file. Two more things to do before merging: (additional to the ones listed above in this discussion)

• [ ] check what exactly to print for ddg input
• [ ] check what format to print for jacobi input - vectors as row or columns

(The output is mathematically correct, we just need to decide on the format)

@DanielEven

[DanielEven]

Very good job! Amazing! Everything looks fine.

[DanielEven]

Also memory leak auto-testers are helpful😎😎

[DanielEven]

The C-API works! I finished calculate_requested and adjusted create_k_eigenvectors_matrix. I also debugged the code for hours until I finally found all the bugs (hopefully), and it runs smoothly and outputs correct results for the blobs1 file. Two more things to do before merging: (additional to the ones listed above in this discussion)

• [ ] check what exactly to print for ddg input
• [ ] check what format to print for jacobi input - vectors as row or columns

(The output is mathematically correct, we just need to decide on the format)

@DanielEven

We can talk and fix the 4 left tasks.

[DanielEven]

Great!👑

[DanielEven]

I fixed the testers.

[DanielEven]

Please review what I wrote. @liammohr

[DanielEven]

And please:

• [x] Write print_output.

[DanielEven]

The Failed tester is because we changed the epsilon from 10^-15 to 10^-5. In my opinion, the staff changed the pdf, but probably didn't match the expected results to it, so the results are for 10^-15 and then there are BAD DIFFs.

[DanielEven]

Also, valgrind discovered some memory leaks.

[liammohr]

We still need to:

• [x] change what our code does for goal=jacobi
• [ ] figure out if we should print -0 or 0
• [x] figure out if there should be an empty line between the eigenvalues and the eigenvectors

[DanielEven]

We still need to:

• [ ] change what our code does for goal=jacobi
• [ ] figure out if we should print -0 or 0
• [ ] figure out if there should be an empty line between the eigenvalues and the eigenvectors

👍

[DanielEven]

The C-API works! I finished calculate_requested and adjusted create_k_eigenvectors_matrix. I also debugged the code for hours until I finally found all the bugs (hopefully), and it runs smoothly and outputs correct results for the blobs1 file. Two more things to do before merging: (additional to the ones listed above in this discussion)

• [ ] check what exactly to print for ddg input
• [ ] check what format to print for jacobi input - vectors as row or columns

(The output is mathematically correct, we just need to decide on the format)

We also need to check these 2.

[liammohr]

The C-API works! I finished calculate_requested and adjusted create_k_eigenvectors_matrix. I also debugged the code for hours until I finally found all the bugs (hopefully), and it runs smoothly and outputs correct results for the blobs1 file. Two more things to do before merging: (additional to the ones listed above in this discussion)

• [ ] check what exactly to print for ddg input
• [ ] check what format to print for jacobi input - vectors as row or columns

(The output is mathematically correct, we just need to decide on the format)

We also need to check these 2.

I think both of these issues were addressed in the forum so it shouldn't be a problem to fix them tomorrow.