Hey everyone, I took a job a few months ago with a team that is building an inventory management app. Pandas is being used in a fair amount of back end data processing (not reports). I've never seen pandas being used in this way and to me and I feel like this can't be good practice. I've brought it up once or twice but nothing has changed and design patterns haven't changed.
I'm not anti-Pandas, in fact I think if you are building reports or 2D arrays with extensive modeling it's the way to go. I just don't think it's appropriate for preprocessing. Am I off base here?
$75 shipped.
PAA The battle of NYC set (c-6 used once)
Will sell both for 35 shipped.
Fine Lavender pour homme set (used once)
Chiseled Face Panda soap (75%)
is there any update on the software for the drop panda? i know there was something like 3 months ago but ive seen nothing since then
Just considering getting this setup myself, but don't want to if it feels awkward and clunky and riddled with input delay?
i try using fn+upkey but it doesnt seem to switch colors just how the duration of the wave of white colors there is
Hi all, intermediate programmer here, pondering script efficiency and how to improve ETLs.
Like many data programmers, I've been using Pandas for awhile. But some of you mentioned here that Pandas has too much overhead and eats up memory compared to NumPy.
I also worked with a software engineer this year who felt strongly that people should learn NumPy properly and not use Pandas as a crutch.
I'd love to hear more views and considerations on this. I've been developing ETLs for low-volume data pipelines so it probably doesn't matter much now, but if we scale up then I want to lay the groundwork properly.
Side note: most of our data is text, not numbers. So then does that mean NumPy vs. Pandas doesn't matter?
Add to this a Python vs. SQL vs. other tools question that I've been thinking about. I want to do more OOP - love the elegance and efficiency - but I believe that Python is always gonna be slower than SQL (we're using Redshift - definitely a bigger, stronger engine than Python scripts). Throw your two cents in if you like!
Edit: Yes, I'm aware that NumPy runs under the hood of Pandas! Good to know for newbies though.
Also: Some good informative responses. Others seem to think I'm saying "NumPy is better than everything else all the time and I love it". No. Obviously there are many considerations, and that's why I wanted to ask for your input.
Panda walks into a restaurent & orders a sandiwch. After finishing the sandwich he gets up, pulls out a gun, shoots the waiter and turns to the manager and says ...'I am a Panda, look it up'!
Managers checks the dictonary that reads.. - 'Panda- Large mamal found in China, who eats shoots and leaves'!
There are many low code / no code Data science libraries / tools in the market. But one stark difference I find using them vs say SPSS or R or even Python statsmodel is that the latter clearly feels that they were designed by statisticians, for statisticians.
For e.g sklearn's default L2 regularization comes to mind. Blog link: https://ryxcommar.com/2019/08/30/scikit-learns-defaults-are-wrong/
On requesting correction, the developers reply " scikit-learn is a machine learning package. Don’t expect it to be like a statistics package."
Given this context, My belief is that the developer of any software / tool designed for statisticians have statistics / Maths background.
What do you think ?
Edit: My goal is not to bash sklearn. I use it to a good degree. Rather my larger intent was to highlight the attitude that some developers will brow beat statisticians for not knowing production grade coding. Yet when they develop statistics modules, nobody points it out to them that they need to know statistical concepts really well.
Just got an email from Kickstarter saying my pledge has been refunded. Big shoutouts to the guys at panda for being so honest!
Edit 1: link to full update from Panda
https://docs.google.com/document/d/1Swl0i_lmZz18aX11W3zk-gWTRvJMT0fjlchFDVNSi-I/edit?usp=sharing
I'm a cook from Panda Express and as a passion project, I've been putting together a cookbook of Panda Express recipes. They aren't 100% accurate because each sauce to some capacity relies on a premade sauce that I can only approximate based on the ingredients labels (feedback welcome). Aside from that, this is how the dishes are made in the restaurant, with some recipes being resized for home use. I hope you enjoy!
Hi
We just released an open source library to calculate Krippendorff's Alpha on a pandas Dataframe. We wanted a simple API that would accomodate the funny shapes that data takes in the wild.
Repo is here and [Blog Post]
**edit - formatting
Hadley is employed by RStudio and Wes is employed by 2sigma, both for-profit firms. It's interesting because their main job is to contribute to non-profit open source software, despite being employed in industry.
For RStudio, the business prop is clear. Hadley made R better -> more people use R -> more people use RStudio.
For 2sigma, a relatively secretive quantitative trading firm, it's not so clear. I guess that 2sigma just uses pandas a lot and wants to fund it? Maybe Wes has duties outside of OSS development, contributing to internal tooling or something, but he has mentioned on twitter that OSS is really his full time gig.
What other examples of firms indirectly sponsoring open source DS software exist via employment?
It's dangerous to open windows in space
)
