OT: I have recently fallen in love with gnuplot + kitty. I am still figuring out if and how to wrap my config with a Python module to allow for simpler notebook-style coding. I kinda dislike matplotlib interface, but I have successfully convinced it to plot to kitty as well. Anyway, it’s pretty magical to see crispy plots inside a teletype emulator.
“kitty” is a hard term to google but I take it you mean the terminal. Does look interesting, unlike many of the gpu-accelerated terminals which sell on speed (gnome-terminal seems fast enough to me…) this one also has some interesting feature ideas. Nice.
I dislike matplotlib’s interface too. I find it very idiosyncratic and unintuitive. I also dislike the fact that the image buffer is implicit global state. The data mungers at my workplace (a group of which I am not a part) prefer plotly and Tableau (which is an Enterprise thing). I do tend to use either matplotlib or gnuplot, depending on how complicated the graph needs to be.
It’s so handy to spin-up a “proper” database as a single, temporary, in-memory process like this. I also like choosing appropriate tools for the job (from a well-stocked toolbox): I’ve not used DuckDB, but will keep it in mind for columnar data; I’ve also found clickhouse-local to be useful, e.g. for joining and querying a bunch of standalone files.
Shell pipelines are also underappreciated; the common sentiment of avoiding Bash scripts in favour of e.g. Python makes sense for scripts which have a bunch of logic and parsing, but Bash is much nicer for piping large amounts of data.
OT: I have recently fallen in love with
gnuplot+kitty. I am still figuring out if and how to wrap my config with a Python module to allow for simpler notebook-style coding. I kinda dislikematplotlibinterface, but I have successfully convinced it to plot tokittyas well. Anyway, it’s pretty magical to see crispy plots inside a teletype emulator.“kitty” is a hard term to google but I take it you mean the terminal. Does look interesting, unlike many of the gpu-accelerated terminals which sell on speed (gnome-terminal seems fast enough to me…) this one also has some interesting feature ideas. Nice.
I dislike matplotlib’s interface too. I find it very idiosyncratic and unintuitive. I also dislike the fact that the image buffer is implicit global state. The data mungers at my workplace (a group of which I am not a part) prefer plotly and Tableau (which is an Enterprise thing). I do tend to use either matplotlib or gnuplot, depending on how complicated the graph needs to be.
Plotnine is a python implementation of ggplot2. It uses matplotlib as the backend. However, it is way more expressive and intuitive to use.
https://www.visidata.org/ is a very powerful tool in this space as well, interactive gui, python scripting, good defaults
https://github.com/multiprocessio/dsq allows one to run sql queries against all kind of data files.
It’s so handy to spin-up a “proper” database as a single, temporary, in-memory process like this. I also like choosing appropriate tools for the job (from a well-stocked toolbox): I’ve not used DuckDB, but will keep it in mind for columnar data; I’ve also found clickhouse-local to be useful, e.g. for joining and querying a bunch of standalone files.
Shell pipelines are also underappreciated; the common sentiment of avoiding Bash scripts in favour of e.g. Python makes sense for scripts which have a bunch of logic and parsing, but Bash is much nicer for piping large amounts of data.
I wasn’t aware of this variant and am now looking into it - thanks.
I would guess that it fits into a similar niche as duckdb - ie a local, OLAP db.
Bash here was only meant to be an enabling technology for allowing people to run the examples! But yes I love it for ad-hoc stuff :)