Re: [PATCH 1/1] xdiff: provide indirection to git functions

From: Edward Thomson <ethomson@edwardthomson.com>
Date: Feb 17, 2022, 22:58 UTC
Message-ID: <20220217225804.GC7@edef91d97c94>
In-Reply-To: <220217.86ee41izpq.gmgdl@evledraar.gmail.com>

On Thu, Feb 17, 2022 at 10:29:23AM +0100, Ævar Arnfjörð Bjarmason wrote:

> 
> [I'm assuming that dropping the list from CC was a mistake, re-CC-ing]

It was; many apologies, I don't use mutt very often any more.  Thanks!

> As for the "new person to our codebase..." I don't think you're wrong
> there, but that's an asthetic preference, not something that's required
> for the stated aims of this series of dropping in compatibility shims.

Sure, but avoiding a prefix is also not a technical decision but an aesthetic and ergonomic one.

Is using a prefix here great? No, it's not great, it's shit. But it's shit that's easy to reason about.

If somebody sees a call to `xdl_free` in some code, they say "wtf is this `xdl_free` nonsense?" And they grep around and figure it out and understand the way that this project handles heap allocations. It's very transparent.

If somebody sees a call to `free` in their code, they say "great, `free`". But it merely *appears* very transparent; in fact, there's some magic behind the scenes that turns a `free` into a `git__free` without you knowing it. You've not learned the way that this project handles heap allocations, but you also don't know that there's anything that you needed to learn. These are the sorts of things that you think you understand but only discover when you _need_ to discover it because something's gone very wrong.

In my experience, calling a function what it _isn't_ is the sort of thing that a developer discovers the hard way, and that often leads to them not trusting the codebase because it doesn't do what it says it does.

Cheers- -ed