merlin VS base2048

Merlin: Context sensitive completion for OCaml in Vim and Emacs\ (0 comments)

Instead of searching functions based on their type structure (like Hoogle), you could search for functions that "consume"/"produce" values of given types (like OCaml's Merlin). I think Rust already computes variance of type constructors, so such a tool just would have to obtain this information.

However, after posting this question I stumbled upon this Github issue where they say it isn't supposed to work out of the box and you're supposed to bind a key to it by editing your .emacs file. Turns out the default .emacs file binds auto-complete to "backtab" which means Shift+Tab but that didn't work. I did eventually discover that I can get some kind of completion by binding backtab to completion-at-point like this:

Self-advertising: I partcipated to the writing of Merlin: A Language Server for OCaml (Experience Report), which explains the overall design of Merlin, a language server for OCaml. A key idea of Merlin are that classic lexing-parsing-typing pipelines can easily be adapted to be incremental for a Language Server, especially when they are using immutable data structures.

qntm did a fun project of using larger bases, constrained to subsets of unicode instead of ASCII like base64. It's specifically for social channels where you're constrained by the number of code points, but not bytes, so you want to maximize data per code point. base2048 is pretty impressive, and base32768 is just absurd.

Numbers don't have to loop until 32k or 2048 (or even up to 1,112,064, until such point the Unicode standard allows for more)

On a semi-related note, they mention base64 encoding messages to evade filters. There were actually other base{n} methods [1] created specifically for Twitter to be more space optimized though not as readily available to operating systems. I guess this is less useful if they are really expanding the text limit to 4k soon but figured I would add it in the event they add a parser for base64.

[1] - https://github.com/qntm/base2048

Actually, in the case of twitter they do some weird counting. It is mostly based on codepoints, true, but some codepoints are considered "heavy" and are counted twice, see https://github.com/qntm/base2048

I know of one case of Twitter doing client side validation [1]. Maybe there are more?

1 - https://github.com/qntm/base2048#note

Base 2048\ (15 comments)