Style: Leverage more of Serapeum

[Ambrevar]

Serapeum has lots of cool features and it feels that Nyxt code quality could benefit from it.

I suggest we start using more of the following, in the order of the reference documentation:

Macro tools

unsplice

with-thunnk

callf

with-read-only-vars

Types

->

true

Defining Types

defunit

defunion

Binding

lret

receive (Safer.)

mvlet (Less verbose.)

Control Flow

nor

eif (Replace all ifs?)

ensure

~>

nest (Reduce indentation.)

eq*

Threads

synchronized (This one seems particularly useful for us, like in data-storage.lisp and maybe the prompt buffer.)

monitor

Iter

collecting

Conditions

ignoring (Replace all ignore-errors and some handler-case?)

Op

op (Nice shorthand for all the trivial lambdas that we have.)

Functions

eqs

partial (Replace all alex:currys?)

flip

throttle (May be useful for some of our threading stuff, like delayed serialization or downloads.)

once

Trees

walk-tree (For the GHT?)

Hash Tables

dict @ (So much nicer to read than make-hash-table and gethash!)

maphash-return

merge-tables

Box

box (We sometimes box values, but maybe always with additional data. Investigate.)

CLOS

make (Seems that Eldoc cannot show the initargs though, can we fix this?)

class-name-of (Oh, this could replace many instances of (class-name (class-of ...)).)

Packages

package-exports (Replace the manual symbol listing we are doing.)

Lists

filter-map (Replace our (delete nil...).)

in (Replace most (find ... :test #'equal) and (member ... :test #'equal).)

pop-assoc (I believe we have a few occurrences of this.)

Sequences

filter keep (Replace some remove-if-not or (remove-if (complement ...) ...).)

single (Replace our (= 1 (length ...)).)

slice (Replace our subseq that rely on the sequence length to count from the end.)

take drop (And so on. Replace some of our subseq.)

seq= (Deep sequence comparison.)

Strings

blankp

words tokens lines (Replace some of our str:split " " ... and str:split (string #\newline) ....)

fmt (Replace our (format nil ...).)

ellipsize (Replace our manual ellipsis formatting.)

Internal Definitions

local (Replace some of our deeply nested flet and labels.)

[Ambrevar]

Thoughts about this? I'd like to start using / converting some of these.

I'd get started with lines, words, fmt, in, blankp. class-name-of is mostly useful when imported, that is, without package prefix.

Thoughts?

[jmercouris]

I prefer to use standard CL whenever possible. I think converting to these constructs does not provide significant improvement to the codebase. For example, I would use (format ) whenever possible instead of str:join.

Of course if you find places in the code where you think these are better, no problem :-D.

[Ambrevar]

In some cases, Serapeum can replace other libraries, for instance words replace (str:split " " ...).

format is not really equivalent to str:join, in particular is renders objects, which can be undesirable (plus there is a performance penalty). The CL alternative to str:join is concatenate 'string which is very verbose :p

[jmercouris]

You can take a list of strings and join them with format. Format can definitely replicate the behavior of str:join :-D. I'm not saying it will be obvious or readable, but it is possible!

[Ambrevar]

I'd like to replace all the verbose (declaim (ftype ...)) with Serapeum's shorter and more natural equivalent.

Compare:

(declaim (ftype (function (&key (:id string) (:current-is-last-p boolean)))
                switch-buffer))

with

(-> switch-buffer (&key (:id string) (:current-is-last-p boolean)) t)

Cool? @jmercouris @aartaka

[jmercouris]

That's fine with me, I was never a fan of the function type declarations anyway :-D

[aartaka]

Cool!

[Ambrevar]

See https://github.com/atlas-engineer/nyxt/pull/1669.

[jmercouris]

Lots of good ideas in here. However, it should come up naturally as we refactor. There is no reason to force all of the code to use serapeum style everywhere.

[jmercouris]

In other words, we should use the best function for the job, whenever we see a place something can be replaced, lets do it, but I don't think we need an issue for it.