See also: 
[https://github.com/nim-lang/needed-libraries/issues/77](https://github.com/nim-lang/needed-libraries/issues/77)

For me I think the pressing needs would be:

  * A jupyter kernel. It can use hot-code reloading.
  * Make NimData, NimPlotly and Arraymancer work with jupyter kernel seamlessly.
  * Being able to delegate to Python ecosystem via nimpy when we are stuck to 
the respective pandas/numpy
  * Allowing Python folks to use Nim libraries for performance via nimpy as well
  * Package all that niftiness in Pip/Anaconda so that we can share the love



An example of Nim benefit, I recently made SVD and PCA (Singular Value 
Decomposition, Principal Component Analysis) at least [2x to up to 10x 
faster](https://github.com/mratsim/Arraymancer/pull/384#issuecomment-536682906) 
than Scikit-learn and Facebook's fbpca. It also uses 2x less memory, which is 
quite useful given the size of current datasets.

Reply via email to