Hi,

I'm trying to implement data frames in D based on mir slices. I'm not sure if there's a better approach. Currently support just basic slicing and assignment.

Are there any efforts in the community toward this goal.

What features would you like to see in the data frame? Any ideas or feature request would be appreciated.

https://github.com/istmarc/typeddataframe

On Saturday, 16 August 2025 at 20:40:50 UTC, Marc wrote:

The only one I know is this one: https://github.com/navid-m/parallax

On Saturday, 16 August 2025 at 20:40:50 UTC, Marc wrote:

My impression is that this turns out to be a lot of work once you get into it. Here are a couple of previous projects that flamed out:

• https://code.dlang.org/packages/magpie
• https://github.com/aravindavk/dataframes-d

I'm not a big user of data frames, but when I want them, I just call them from R. At first it's tempting to do the NIH thing, but that quickly goes away when you realize you don't have to spend years writing and testing libraries that duplicate existing functionality. This is another one of the many areas where NIH syndrome makes D users go in circles, inevitably leading to their departure for another language.

On Saturday, 16 August 2025 at 20:40:50 UTC, Marc wrote:

I don't think you should attempt to port a python data structure 1 to 1. I'd like to see a smarter data structure I'd suggest figuring out smarter filtering on statictyped data. There's hard problems before mimicking a dymanic language will just add complexity.

On Saturday, 16 August 2025 at 20:40:50 UTC, Marc wrote:

I forgot to mention - if I was to implement dataframes, I would read polars code, and check how they did it.

For me the most important feature would be how nicely it fits into the rest of the D ecosystem. Implementation of dataframes must be idiomatic D. Range based, etc. We do not yet have replacement for std.stream, but there is talk of replacement based on iopipe/std.io. dataframes implementation should work with that too.

One of the most impressive aspects of polars is that you can get data from anywhere and save data everywhere.

On Saturday, 16 August 2025 at 20:51:35 UTC, Dejan Lekic wrote:

This one looks good overall.

On Sunday, 17 August 2025 at 01:24:51 UTC, Monkyyy wrote:

That’s why iam trying to make it with type information. But it’s very restrictive I feel and some mutations which are common when manipulating a data frame are hard to implement. Maybe it can be implemented with a different more smarter structure.

On Sunday, 17 August 2025 at 08:38:53 UTC, Dejan Lekic wrote:

polars is a great library. I will try to see what if it’s possible to implement all the operations available on polars.

On Monday, 18 August 2025 at 21:49:33 UTC, Marc wrote:

Such as? I dont like all the opinions that went into sumtype but I cant imagine it being impossible to store a myarray!(sumtype!(int,float)) just working naturally