Python is my day job and I'm very fluent in the common data science tools like Pandas, but you'll hardly ever catch me using Pandas because Excel is just so much easier for 2d data.

The thing is, I never use Excel because the data I work with every day is almost always multidimensional, so I work with xarray[1]. It drives me crazy to see coworkers using Excel for things that it is not suitable for and having crazy solutions like multiple tables per sheet, multiple sheets, multiple files or other hacks to describe another dimension of data.

[1] http://xarray.pydata.org

> It still takes lot of expensive human work to select the good quality works and proofread /edit/ peer review them etc before it is fit for publishing in a decent journal

Good thing that almost all of that expensive effort is done by volunteer scientist editors and reviewers. I routinely see spelling errors and english errors (going against journal guidelines for British/US english) in scientific literature. I'm convinced publishers do as little proofing as possible.

For the major journals (Nature, Science, Cell, PNAS), they don't do any. There is no copy editing. That's the author's responsibility now.

Pandoc?

That is what I use lately. It's not perfect but it's the closest thing I have found.

In my field, if you get a TT position and play nice, you usually get tenure. But yeah - it's basically all about money. You can't do good research if you don't have solid funding, you don't get solid funding unless you have good research in high impact journals.

One thing they mentioned in the 'left out' section is code review. I'm a researcher in a computation-heavy field and I think that anything I write that is intended to be used in our research group or used by other people should be code reviewed. In my experience, so many codes written by former group members or by senior group members become blackboxes that no one can read or maintain in five years.

Code review for anything more complicated than a script has helped the quality of what I write. It also ensures that there are other people who have at least seen the code that I wrote. Even if they don't fully understand it, they are at least empowered enough to wade through it if need be.

Apparently google rewite all code every three years.

https://arxiv.org/ftp/arxiv/papers/1702/1702.01715.pdf

2.11

I am there. We learn to please our master as our ancestors taught us in our rich oral tradition.

I hope that this undocumented and fragile big codebase dies before killing the PhD of some poor guy.

This is more likely to be applied to topology optimization imo

Retain means pay monthly.

Are you sure about that? My understanding is it simply means you pay a certain amount up front.

That money is held by the firm and Billings are placed against it until time to top it off again.

> Python in some cases can be 100x slower than C

Yeah, the Python implemented version is. But people doing serious computing in Python that requires speed are doing it with NumPy or even Cython or just straight up calling C/Fortran libraries in Python.

It's the same in R - you can have good performance of you use vectorized routines, because those routines are written in C/C++/Fortran.

Maybe that's just the language, not the support?

To me the Hardee's burgers are just like Burger King.

have no way of telling if that's a compliment or complaint...