What features or libraries are you waiting for before you start using Mojo in your domain?
Just curious what the community is waiting for before they start using Mojo seriously in their domain.
I'm in the Data Eng/Backend/DevOps space and I have a few I'm waiting for. Namely an HTTP client with SSL support and Database drivers to build data services/tooling.
As for Mojo features, I'm excited to see file scope code support and better C FFI (for the above libraries). A clean way to call system
commands like
subprocess would be great too!9 Replies
Extensibility with other languages is a huge thing for me, I have a few jobs where Mojo would fill a great niche, but it just can’t be slotted into any ecosystem at the moment due to limited libraries and support
Something like a Postgres driver would really make a huge difference. But also basics like JSON serialization
Postgres is my preferred flavor of RDBMS as well! Would be great to have something there. I’ve tried my hand at wrapping the C API for DuckDB and but couldn’t figure out how to actually get query results 😂
My domain is control theory / autonomous systems. I'm very interested in how Mojo could accelerate the deployment of AOT compiled ML / RL elements onboard mobile robots.
I'm waiting for the following features:
* Better C FFI / interoperability (C++ and/or Rust would be great as well). This would allow for much easier integration with existing frameworks such as ROS2 or Zenoh.
* The ability to call statically compiled Mojo code from Python (the reverse is currently possible).
* Support for embedded systems and microcontrollers. It would be fantastic to code elements such as UAV autopilots in Mojo. It would also be great to use Mojo for AOT compiling ML elements to run on companion computers such as Raspberry Pis or Nvidia Jetsons.
* A rock solid general-purpose autodiff framework similar to JAX.
For me, better support for polymorphism is at the top of my wishlist. Dynamic traits, fields and default methods for traits would be game-changer, support for classes and inheritance would also be helpful for writing more elegant and maintainable code of course.
Basically domain less these days but I love to implement tools/frameworks which application can utilize and build upon, such as an LLM Suite to name a big one i have in mind. However, without proper polymorphism, many things are possible to do but not as flexible as I would love to implement them - at least with my current Mojo coding skills 😜
In the long run i really look forward for Mojo to become closer to be a superset of Python. So many great python libs out there which could benefit from porting them to Mojo.
I'm waiting for intel mac support and better networking utilities. Python's Socket wrapper for example
That reminds me of another thing, most of my actual work projects REQUIRE some form of web request being sent, and i’m just not comfortably relying on importing python when I could just use it instead
no being broken, damn it. I've just wasted approx 6 hours of time trying to fix a basic import problem. Another hour fixing a for loop that doesn't work. That has happened over multiple days.
The language error module kept crashing probably couple dozens of times every day which led me to not seeing my errors.
To get an icing on the cake, no documentation.
i thought I will commit to the language, but enough is enough - it slows me down by such a margin that I'll be faster off using Python.
The possibility to call Mojo code from Python to allow us to progressively migrate more and more parts of the codebase to Mojo