windows-server-2003

If this is caused by the same issue as #44, it should be fixed in v0.3.0. Can you try and see if it's fixed or not ?

I'm back! I agree that invidius is better in terms of stability. However, YouTube parsing has little impact on performance, memory consumption or battery life, as it is only done...

P.S. Test on old 3DS Current: 4.1k calls/s = 68 characters/frame Binary search: 133k calls/s = 2.2k characters/frame

This is probably because it failed to connect to YouTube I admit the error message is weird, so I'll fix this

I'm not sure if this belongs to MAINT or BUG. Let me know if you have any idea on this.

Thank you. I'd like to rebase/force-push to edit the commit messages from MAINT to ENH and squash the "Comply with lint" one.

@jjerphan I know you are busy, but look at the 14x speedup in the most fundamental function in graph area! It will surely improve scipy a lot!

Okay, always thanks for working on scipy and related things!

I just remembered that C++ STL can be used in Cython std::priority_queue should replace the Fibonacci Heap, and - It reduces the maintenance cost - priority_queue doesn't have decrease_val but...

Here's the results of the benchmark(3 runs each, I'm only able to use the one-time benchmark `dev.py bench` because I haven't found the proper way of using libcpp from Cython...