fast-json VS tech.ml.dataset

There are many more things there but you can read the README for more information. There is timing information at the bottom of the README and the fast-json project includes a comparison. My hope is that ham-fisted becomes a platform to try out ideas and build faster datastructures and algorithms that follow in the typeless functional pathway so we can carefully and measurably improve the platform we all are standing on. So please take it, try some things out, and let me know how it goes :-).

Tablecloth by generateme is a friendly & expressive table-processing library built on top of tech.ml.dataset & dtype-next, Chris Nuernberger's high-performance data libraries.

For 1: This ns of tech.ml.dataset supports reading of multiple worksheets per file https://github.com/techascent/tech.ml.dataset/blob/master/src/tech/v3/libs/fastexcel.clj

After building tech.ml.dataset and charred I wanted to take the lessons learned there and apply them back into the base Clojure substrate of persistent maps, persistent vectors, and algorithmic primitives like group-by and frequencies.

Did you try tech.ml.dataset?

Checkout babashka for scritping and clj-python to use numpy from clojure, or https://github.com/techascent/tech.ml.dataset for pure clojure

Really high quality libraries for deep learning, dataset manipulation, and more

tech.ml.dataset has seen some major upgrades for discoverability - specifically the tech.ml.dataset main namespace has been revamped. If you use Cursive or Calva your intellisense will now work with the main namespaces.

It is based on a state-of-the art , high performance tabular dataset implementation, tech.ml.dataset and combines it with a innovative pipeline approach build with idiomatic and functional Clojure concepts in mind. Machine Learning models get pulled in as plugins from existing ecosystems, so are available from the start. Please find user guides and example code in GitHub at https://github.com/scicloj/scicloj.ml