This Week’s Program: Apr 3 - Apr 7

My commits are dense and all over the place this week. Just as I hit my stride in my GIR library, I hit a wall with how to represent something in the API. I need the weekend to chew on a game plan.

Thankfully there’s a nice distraction to be had in Mastodon. The new federated social network has captured the imagination of those of us who are more than a little burnt on Twitter and more than a little interested in networks built on standard protocols and technologies. I am @mw@mastodon.social.

I find Mastodon fascinating, not so much because of its potential as a Twitter escape hatch, but because the protocols that drive a federated publication/syndication network are ripe for exploitation. Could you make an experience like Snapchat Stories on the Mastodon/GNU Social protocols? The potential for new experiences and interfaces is so interesting. It reminds me of peak RSS or when I first discovered tumblelogs. When I reach a conclusion with overscan, I’ll be eyeballing Mastodon for my next project.

Despite my fascination with Mastodon, I’m finding myself more and more fatigued with social networking. There’s been a downward trend in my “engagement”. My tweets aren’t liked or RT’d. My subscription count is stagnant if not down for this newsletter. Am I losing touch? Are people just not into this whole Racket thing? What does my (tiny) audience expect from me? What do you, dear reader, expect?

I hope you expect the sickest code.

f1438547294b2dc41e7845e6a5c242921032d24f

I create a base struct gi-base, for all the other GObject Introspection structs to inherit from. I’m trying to get my codebase to be a better match to what GIR actually does. I also make it so that when invoking a function returns a pointer, I cast it to another pointer that represents a structure that can be introspected. This way, I can call methods on GObject instances.

7e1a037439add2a1b453aae67808c09bc830f8eb

I alter a handful of foreign function signatures to accept or return a _symbol type. C represents these as Strings (char * more precisely) but Racket will transform them into interned symbols.

725d6ee7f3e5c7fe06f3531fe4c014bf9387de8a

At this point, I stumble across something in the Racket FFI docs that show that I can actually alter the definition of some of these C functions. So that:

(define-gir g_base_info_get_name (_fun _gi-base-info -> _string))

Can be renamed in Racket but refer to the same function like so:

(define-gir gi-base-name (_fun _gi-base-info -> _string)
  #:c-id g_base_info_get_name)

Before, I was creating my own Racket functions that just wrapped these C functions. I don’t need those wrappers at all! I’m really impressed with the flexibility that Racket’s FFI offers.

aadb78788439577bdef21b019a8cd266fa0c859a

This flexibility allows me to update the _gi-base-info base type. Instead of just an empty pointer value, I can do some intelligent conversions here. Now, whenever a C function returns a _gi-base-info type, Racket will unwrap that pointer into the appropriate struct.

Here’s where the wheels fell off a little bit. Let’s say I have a C function that I’ve wrapped with GIR and it returns a pointer to a C Struct. What do I unwrap that struct into? I know what fields that struct has, what methods it can call, its name… I tried to make a new (Racket) struct type on the fly, but that’s not quite right. I think I have to turn to Macros here so that whenever a Registered Type is introspected, it creates definitions for instances of that type for Racket. Alternatively I can create a struct-instance struct that holds the details, but that API is less than ideal. It’s driving me a bit mad.

Toot me on Mastodon if you have some thoughts or suggestions.

– Mark