CE was started in 2012 to serve my needs at my company to show how C++ constructs translated to assembly code. It started out as a tmux session with vi running in one pane and watch gcc -S foo.cc -o - running in the other. Since then, it became a public website serving the C++, Rust, Go, Haskell, Ispc and D communities and performs around 20,000 compilations per day.
This document is an attempt to capture thoughts on the future direction of Compiler Explorer.
CE‘s UI doesn’t work well with mobile clients. The editor doesn‘t work well on many mobile clients, and the layout doesn’t lend itself well to small screens.
Ideas for improving mobile support include automatically folding up all the panes into a single tab upon detection of a mobile client. This would require a bunch of fixes in the underlying UI library as this doesn't properly work with mobile and tabs.
Perhaps a read-only simplified view would work better: the main reason one brings up the CE website is to look at tweeted links rather than author content.
The UI has a number of things that need improving:
Another big ticket item is to allow executing of the user's code. This is fraught with security issues, and brings up a number of UI and API considerations. Compiling code every time to execute with different params seems wasteful, so caching seems good; but in a multi-instance setup a shared cache woudl be needed. Perhaps a backend system that caches the executables (and makes them downloadable; at least for some compilers where license allows), and stores the binaries in ephemeral, shared storage. This same backend system could also be used to store code, and could be part of a whole new way of sending and sharing code (if made permanent storage).
Most of the open tickets are to do with adding new compilers, or fixing issues with existing compilers. Continuing to add more compilers and make it easier for others to submit PRs to add new compilers is very important. A separate document explains how to add a compiler.
There‘s an inherent tension between the standalone, run-it-yourself version of CE and the scalable, AWS-backed CE instance. Care must be taken to keep the standalone version usable, not least as the majority of CE’s development is done on a laptop during a commute (with little or no internet access).
Above all, the priority is to keep the main CE site up, stable and dependable.
CE will remain ad-free, open-source and non-commercial. There's no plans at all to add “freemium” content, despite the Patreon site where folks can help support the cost of running the servers.
With all this in mind, the tentative goals for 2017 are:
These goals will be refined as time ticks on.