Thread by @arthurturrell

Ok, we're all tired of the neverending count, so here's something completely different:

10 less well-known code packages; the fun, the useful, and the unexpected.

🧵below.

#EconTwitter #DataScience #RStats #Python #PyData

1. Jazzit. “Ever wanted your scripts to play music while running/ on erroring out? Of course you didn’t. But here it is anyway”

Yes, Jazzit laughs at your expense when your code hits an error, here via a clip from curb your enthusiasm. github.com/Sangarshanan/j…

(see also: beepy)

2. Handcalcs

Handcalcs renders maths in latex in your console/notebook and can also export latex equations to file. *Really* useful for writing papers where you want your code and the mathematics in your paper to be consistent.

github.com/connorferster/…

(see also: sympy)

3. Pandas profiling

Exploratory data analysis is important, often a bit overlooked, and also saves time in the long-run. So a library creating beautiful HTML EDA reports (missing variables, cardinality, distributions, & correlations) is super useful.

github.com/pandas-profili…

4. Matplotlib!?

Alright, you’ve probably heard of @matplotlib & might be surprised to see it on the list. But there’s a new feature of matplotlib that you might not be aware of: figure placement using ASCII & lists. Can be useful for plots in papers.

matplotlib.org

5. Pandera data validation

You don't want nasty surprises with the data you're reading in (think floats as strings), & this package validates data types, & more, as it arrives. Fantastically useful.

pandera.readthedocs.io/en/stable/

See also: Great Expectations, greatexpectations.io

6. Tenacity

If at first you don’t succeed, try and try again. Tenacity has several different ways to keep trying a function, even if execution fails.

tenacity.readthedocs.io/en/latest/

See also: purrr’s 'insistently'

7. Streamlit

streamlit sells itself as the fastest way to build data apps (dashboards). You can do an impressive amount with a very simple set of commands. But there’s also depth--it can, for example, serve up 'explainable' AI models.

streamlit.io

Also: RShiny, Dash

8. Black

Black is an automatic code formatter (“you can have it any colour you want, as long as it’s black”). If you just want to write code & forget about formatting, it's brilliant-though some won't like its opinionated style.

black.readthedocs.io/en/stable/

Also: Google's yapf

9. Scalene for profiling

Profiling finds bottlenecks in your code-but with scalene, you can find bottlenecks in your data too. Yes, it does memory profiling! Super useful for economics & data science, where *memory* is often limiting.

github.com/emeryberger/sc…

Also: PyInstrument

10. alive-progress for progress bars

Waiting for code to run doesn't *need* to be tedious. alive-progress gives a wealth of options for glitzing your progress bar. Better still is the realistic ETA (I'm looking at you, Windows File Transfer)

github.com/rsalmei/alive-…

Also: tdqm

Bonus: R style analysis in Python!?

R users swear by two of R’s most loved declarative-style packages, dplyr for data analysis & ggplot2 for plotting, & miss them when working in Python. Here are two great stand-ins:

plotnine, a port of ggplot2:
plotnine.readthedocs.io/en/stable/inde…

...and here's Siuba, a Python port of dplyr in R. (Yes, it has slightly awkward syntax for now, but it's new & great for R users wanting to get going quickly in Python.)

github.com/machow/siuba

Well I hope you've enjoyed this non-election related content. You can find all of the packages, the 'see also' packages, and more, in this blog post:

aeturrell.com//2020/10/24/10…

I see the gifs inexplicably didn't work on this one! You'll have to go onto the homepage.