blog/content-org/blog.org
2024-01-25 06:43:59 +01:00

1465 lines
63 KiB
Org Mode
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# -*- eval: (require 'ox-hugo) -*-
#+title: Phundraks Blog
#+author: Lucien “Phundrak” Cartier-Tilet
#+hugo_base_dir: ../
#+hugo_section: ./
#+hugo_categories: emacs linux conlanging orgmode
#+startup: content
* [EN] Open-Sourcing ALYS :ALYS:
:PROPERTIES:
:EXPORT_FILE_NAME: open-sourcing-alys
:EXPORT_DATE: 2021-12-15
:export_hugo_menu: :menu "main"
:END:
#+TOC: headlines 1 local
** Too Long, Didnt Read
VoxWave no longer exists as a company, ALYS lives on as an open-source
project under the GPL-3.0 and the BY-CC-4.0 license. You can find it
at the following address:
+[[https://labs.phundrak.com/phundrak/ALYS]]+
EDIT: The main repository moved to [[https://labs.phundrak.com/ALYS/ALYS]]
and vocal libraries are now separated in different repositories linked
from the main one.
** What happened?
You might have noticed it, but VoxWave became quite silent over the
last months. This is because we at the head of VoxWave chose to close
the company, a decision which came in effect in early September 2021.
Theres not much else to say.
*However*, the good news is the rest still goes on! ALYS as a project is
still alive and well! As her creator, I decided to step in and
continue its technical support. Since the company no longer exists,
and as a free and open-source software supporter, I also decided to
open-source ALYS as much as possible. As a result:
#+begin_center
ALYS for Alter/Ego is now free software as in /free beer/.
ALYS for UTAU, including its previously unreleased UTAU prototype, is
now free as in /free beer/ and in /freedom/.
#+end_center
You can find the installer of ALYS for Alter/Ego on the repository
linked above as well as a free licence file. Regarding its UTAU
version, its prototype is already configured with ~oto.ini~ files, but
the source file for its Alter/Ego version are stripped of any
configuration.
** Whats New?
Therefore, ALYS is now available under three different licences:
- The character is now under the free [[https://creativecommons.org/licenses/by-nc/4.0/][CC-BY-NC-4.0 licence]]
- The UTAU vocal libraries are now under the [[https://www.gnu.org/licenses/gpl-3.0.en.html][GPL-3.0license ]] ([[https://choosealicense.com/licenses/gpl-3.0/][short
readable version]])
- The Alter/Ego is still under a proprietary licence under my name, but
it is now available free of charge
Basically, this means you can do whatever you wish with the character
as long as it is non-commercial, and you credit [[https://www.instagram.com/hsaphirya/][Saphirya]], ALYS
designer. The UTAU vocal libraries can be used, modified, and
redistributed as much as you wish as long as it stays under the
GPL-3.0 licence. And you are free to use the Alter/Ego vocal libraries
as much as you wish, but you cannot redistribute or modify them.
I also decided to release ALYS very first, secret, unreleased,
unheard French vocal library. It was scrapped no long after recording
it due to quality issues and was replaced by its French UTAU prototype
people could hear through ALYS first songs. It is released more as a
way of preserving the fact it existed rather preserving a usable vocal
library. (I dont even remember what it sounds like.)
If you have any question, you are free to email me at
[[mailto:lucien@phundrak.com][lucien@phundrak.com]] or open an issue on the repository mentioned
above.
* Conlanging :@conlang:
** TODO Writing my conlanging docs with Emacs :emacs:conlanging:
* Development :@dev:
** [FR] Mettre à niveau mes sites org-mode :dev:emacs:
:PROPERTIES:
:EXPORT_FILE_NAME: mettre-a-nivea-mes-sites-org-mode
:EXPORT_DATE: 2022-08-15
:export_hugo_menu: :menu "main"
:END:
*** Le Problème
Cela fait quelques temps que je réfléchis à une nouvelle manière de
gérer deux de mes sites web, [[https://conlang.phundrak.com][conlang.phundrak.com]] et
[[https://config.phundrak.com][config.phundrak.com]].
Les deux sites sont actuellement générés via un export depuis org-mode
(un des nombreux modes dEmacs) directement vers du HTML. Sauf que
lorganisation du fichier HTML de sortie de me plaît pas, et depuis
plus de deux ans jutilise un script rédigé en Dart et compilé vers
Javascript pour réorganiser les fichiers. En soit ce ne serait pas
trop grave si mes pages web nétaient pas forcément lourdes. Mais
elles le sont! La plus lourde page de mon site de linguistique fait
232Ko (la page francophone sur le Proto-Ñyqy) et celle de mon site de
configuration fait 5,5Mo (configuration Emacs)! Je parle bien de
fichiers HTML! Il faut vraiment que ça change!
*** Nouveau Framework pour le front-end
À la base je métais lancé pour écrire un exporteur personnalisé pour
exporter mes fichiers org-mode vers des fichiers JSX qui seraient
utilisés par un projet [[https://reactjs.org/][Reac]]t, ou même [[https://nextjs.org/][Next.js]]. Mais jai récemment
découvert quelque chose qui pourrait être bien plus pratique pour
moi: [[https://vuejs.org/][Vue]] et tout particulièrement [[https://v3.nuxtjs.org/][Nuxt]]!
En effet, Nuxt lit le [[https://content.nuxtjs.org/guide/writing/mdc/][MDC]], ou Markdown Components. De fait, il est
possible avec MDC et Nuxt dinsérer dans du Markdown des composants
Vue soit en blocs soit en inline. Et pour moi, ça change tout! Je
peux maintenant écrire un exporteur minimal qui se chargera simplement
dexporter quelques éléments personnalisés vers des composants Vue,
voire même de simples macros org-mode pour exporter les composants
inline.
Et bien sûr, pour pallier au problème de fichiers HTML trop lourds, il
me faudra séparer mes fichiers actuels en plusieurs fichiers, mais
cela devrait être plus simple à gérer une fois la transition vers le
nouveau framework effectuée.
*** Et pour le backend ?
Mais ce nest pas tout: un élément que jaimerais ajouter à mon site
de linguistique serait un dictionnaire entre mes langues construites
et dautres langues, quelles soient construites ou non. Ce
dictionnaire doit pouvoir être interactif, avec par exemple une
recherche, une page par mot, etc.
Je ne ferai certainement pas télécharger à mes utilisateurs
lentièreté du dictionnaire à chaque recherche dun mot dans le
dictionnaire, il ne peut donc pas être hébergé avec mon frontend, et
jaurai besoin dun backend avec une API REST pour gérer les requêtes
des visiteurs du site web. Maintenant la question est, quel type de
back-end ?
Tout dabord, je vais complexifier un peu le problème: je suis un
grand amateur de org-mode. Je pourrais gérer ça via une base de
données classique, ajoutant chaque entrée manuellement, mais je vais
plutôt essayer de gérer tout ça via org-mode. Les fichiers texte sont
plus simples à versionner que des bases de données en un seul fichier
binaire. Du coup, il va falloir que je mécrive un nouvel exporter,
mais lequel?
Je pourrais rédiger un exporteur pour mon fichier ~dictionnaire.org~ qui
lexporterait vers un fichier Json qui serait lu ensuite par mon
backend qui extraierait et enverrai à mes utilisateurs les
informations nécessaires. Lavantage serait de navoir quasiment pas
besoin de manipuler le Json et den envoyer tel quel. Mais louverture
et fermeture constante du fichier nest pas forcément la meilleure des
idées, quoi que cela pourrait permettre de remplacer le fichier
pendant que le backend tourne. Mais je suis sûr quon peut mieux
faire.
Ma solution suivante était dutiliser EmacSQL, un paquet Emacs lui
permettant dinteragir avec des bases de données SQLite, PostgreSQL et
MySQL. Au moins ce serait une véritable base de données, avec
seulement un blob binaire à mettre à jour, et ce serait
potentiellement plus performant étant donné quil ny aura quà ouvrir
une fois une connexion avec elle. Mais le problème est maintenant sa
mise à jour. Mince…
Vient enfin ma troisième solution qui, je pense, sera celle que je
vais adopter: utiliser une base de donnée type Firebase. Lidée dun
verrouillage fournisseur ne me plaît pas franchement, donc jai décidé
dutiliser une alternative open source et hébergeable: [[https://appwrite.io/][Appwrite]]! Je
peux écrire sur une de ses bases de données pendant que mes
utilisateurs peuvent la lire, donc la mise à jour nest pas un
problème, et je nai rien à mettre en ligne, seulement une série de
requêtes à faire. Cependant, un problème reste: comment communiquer
avec Appwrite?
*** La quête pour un SDK Appwrite pour Emacs
Hélas, jai beau chercher, il nexiste aucun paquet pour Emacs
permettant une communication avec Appwrite. Mais ce nest pas
franchement surprenant: Appwrite nest pas encore extrêmement
répandu, et même Firebase ne dispose pas de paquet pour Emacs.
Bien heureusement, Appwrite dispose dune API REST assez bien
documentée, et Emacs est capable de gérer des requêtes nativement via
sa bibliothèque ~url~, cest donc naturellement que jai commencé à
travailler sur ~appwrite.el~, un SDK Appwrite pour du Emacs Lisp.
Jaurais pu utiliser ~request.el~, un paquet assez populaire pour Emacs
afin de gérer les requêtes HTTP, mais je ne suis pas grand fan de son
workflow et je préfère limiter au maximum le nombre de dépendances
dans mes paquets. Ce que ce paquet fait actuellement est une
transformation des paramètres nommés que mes fonctions acceptent en un
payload Json. Par exemple, ma fonction ~appwrite-stogare-list-buckets~
accepte les mot-clefs ~search~, ~limit~, ~offset~, ~cursor~, ~cursor-direction~
et ~order-type~. Ces arguments sont transformés en du Json via la
bibliothèque native dEmacs afin de donner ceci:
#+begin_src js
{
"search": "my search request",
"limit": 30,
"offset": 0,
"cursor": "",
"cursorDirection": "before",
"orderType": "ASC",
}
#+end_src
Ce payload Json est enfin envoyé à lAPI REST correspondante, en
loccurrence ~/v1/storage/buckets~ comme on peut le voir [[https://appwrite.io/docs/server/storage?sdk=nodejs-default#storageListBuckets][sur la
documentation officielle]]. Bien sûr, les éléments optionels ne sont
pas nécessairement inclus afin déviter à avoir à envoyer trop
dinformations. Dans ce cas, tous les éléments du payload sont
optionels, ce qui ferait que le ~appwrite.el~ nenverra que src_js{{}}
comme payload à lAPI.
Pour linstant, le projet en est encore à ses débuts, mais jai
commencé à travailler sur le SDK pour Appwrite que vous pouvez trouver
sur [[https://github.com/Phundrak/appwrite.el][ce dépôt Github]].
La question maintenant est: comment exporter mon dictionnaire vers
Appwrite? La réponse me semble relativement simple; je pourrai
écrire un exporteur org-mode dépendant de ~appwrite.el~ qui exportera
pour chaque mot quil rencontrera un payload Json vers mon instance
personnelle Appwrite. Et à la différence des exporteurs org-mode
habituels, ~ox-appwrite~ nexportera aucun fichier sur mon système.
*** Conclusions
Au fur et à mesure de mon analyse du projet et de mes besoins, je me
suis rendu compte que jaurai besoin doutils plus intelligents que de
simples pages HTML exportées automatiquement via Emacs.
Ainsi, jaurai besoin de créer un site web avec Nuxt, profitant ainsi
de sa capacité à rendre du Markdown avec du contenu interactif,
agissant en tant que frontend pour mon site web. Ce Markdown sera
exporté via org-mode à partir de mes fichiers déjà existants, bien
quà fragmenter afin de réduire la taille des fichiers de sortie.
Le backend sera une instance Appwrite que jhébergerai moi-même sur
mes serveurs. Elle sera populée par un exporter org-mode custom via
Emacs, ce qui me permettra de continuer à gérer mes dictionnaires et
mes langues avec org-mode.
Ce projet est vraiment intéressant car cela ma incité à explorer de
nombreuses possibilités et technologies différentes afin de trouver ce
qui correspond le mieux à mon besoin, notamment en me rendant compte
par exemple que React nétait pas forcément loutil le plus adapté à
ce projet précisément. Cela me fera également travailler sur ma
capacité à interagir avec des backends et des API REST, tout autant du
côté front-end pour le site web que du côté SDK avec Emacs. Enfin, la
création de ce SDK ainsi que des exporteurs org-mode me sera bénéfique
afin dapprofondir ma connaissance dEmacs et du Emacs Lisp.
Maintenant, au travail!
** [EN] Writing a Dynamic Array in C :dev:C:
:PROPERTIES:
:EXPORT_FILE_NAME: writing-dynamic-vector-c
:EXPORT_DATE: 2020-11-28
:export_hugo_menu: :menu "main"
:END:
/Edit on October 28th 2023:/ This article was written on November 28th
2020, almost three years ago. Since then, I have noticed issues with
the current implementation of my dynamic C array, as noted by some
readers in the comments below. I will probably rewrite a new dynamic
array in C some time in the future addressing these issues.
Although C is a very, very popular language, it is also known to be
quite tiny: memory is handled manually, and much of what is available
in its standard library is a given in all other languages. But C being
a low level language also means it lacks a lot of other stuff other
popular languages have; for instance, dynamic arrays are present in
the library of most popular languages, be it JavaScript, C++, Rust and
so on, but Cs simplicity forbids them from being there. If you want
it in C, you have to implement it which is exactly what I did!
#+TOC: headlines 1 local
*** Introduction
When I wrote this library, I was mostly inspired by C++s ~std::vector~
and Rusts ~std::vec::Vec~, but my library lacks some features both
have: its still a simple one. Here is the list of what it is able to
do:
- Create a dynamic array, with or without an initial capacity
specified by the user
- Store a function pointer to the /destructor/ of the elements that will
be stored in the vector for when they are destroyed
- Append new elements at the end of the array
- Get elements by position, safely or not, or get the first and last
elements in the array
- Get the length of the vector as well as its capacity
- Shrink the size of the allocated array to the size of the vector
- Remove an element at a specific index, or the last element
- Completely destroy the vector and its elements
Elements that will be stored in the vector will need to be dynamically
allocated in memory since the vector will not store the elements
themselves, but rather pointers to them. This way, we avoid copying
data when inserting it to the vector, and handling these elements is
also a tad easier. And since we do not know what we will be storing,
we will be storing void pointers. The user will be able to cast them
to their desired type later on.
Before defining the vector, there are a few things I want to define.
First, there is an attribute I will often use with my functions:
#+NAME: vector-nonnull-h
#+BEGIN_SRC c
#indef NONNULL
# define NONNULL __attribute__((nonnull))
#endif
#+END_SRC
This will forbid passing to functions marked with this attribute ~NULL~
pointers, because we will use a lot of them.
We will also need to include some headers:
- ~assert.h~ :: so we can make sure memory is allocated and reallocated
correctly
- ~string.h~ :: for some memory operations such as ~memcpy~
#+NAME: vector-includes-c
#+BEGIN_SRC c
#include <assert.h>
#include <string.h>
#+END_SRC
We also need to define a type that will be used as the destructor
type. The functions we want to accept as destructors are functions
that accept a void pointer to an element and return nothing, hence
this definition:
#+NAME: vector-destructor-type-h
#+BEGIN_SRC c
typedef void (*Destructor)(void *element);
#+END_SRC
Now, onto the structure itself.
*** The Data Structure of the Vector
With our vector, we will need to keep track a couple of things:
- the size of the vector
- the capacity of the vector
- the destructor
- the array itself
With this, we can describe our structure for the vector:
#+NAME: vector-struct-def
#+BEGIN_SRC c
struct Vector_s {
size_t capacity;
size_t length;
void ** elements;
Destructor destructor;
};
typedef struct Vector_s Vector;
#+END_SRC
We have now four elements:
- ~elements~ :: an array of void pointers pointing themselves either to
elements stored in the vector or to nothing (initialized to ~NULL~)
(note this forbids storing ~NULL~ elements in the vector),
- ~length~ :: the number of elements currently stored in the vector,
- ~capacity~ :: the size of the allocated memory pointed to by ~elements~
divided by the size of a void pointer. This gives us the amount of
elements that can be stored in the vector without any reallocation /at most/,
- ~destructor~ :: pointer to the function used to free elements stored
in the vector
Now, onto the functions associated with this data structure. They are
all prefixed with ~vec_~ in order to avoid any collisions with other
libraries and functions.
*** Building Vectors
The first function for building vectors is ~vec_new()~. Here is its
definition:
#+NAME: vector-vec_new-h
#+BEGIN_SRC c
Vector *vec_new(Destructor const destructor);
#+END_SRC
It is quite straightforward: when creating a new, standard vector,
simply pass as its arguments a pointer to the destructor of this
vector, either a ~NULL~ pointer for trivial data types, or a pointer to
an existing function you declared somewhere. Once you do that, you get
yourself a pointer to the newly created vector with which you can now
store elements. Lets see how it works under the hood:
#+NAME: vector-vec_new-c
#+BEGIN_SRC c
Vector *vec_new(Destructor const destructor)
{
Vector *self;
self = (Vector *)malloc(sizeof(Vector));
assert(self);
,*self = (Vector){.length = 0,
.capacity = VEC_INITIAL_CAPACITY,
.elements = (void *)malloc(sizeof(void *) * VEC_INITIAL_CAPACITY),
.destroy = destructor};
assert(self->elements);
return self;
}
#+END_SRC
A new pointer is created, which will be the pointer returned to the
user. To this pointer, we allocate enough memory to hold a vector.
Once that is done, we initialize this new memory buffer with an actual
vector, with its members initialized as described above. An assertion
is done in order to ensure both the vector but also its storage are
correctly allocated.
The second function, ~vec_with_capacity~, is quite similar though not
the same as ~vec_new~: it allows for an initialization of
~vec_with_capacity~ with a user-defined amount of capacity in the
storage of the vector. That is, if ~vec_with_capacity(14)~ is called,
the library will return a pointer to a vector which can contain and
has the size of precisely fourteen elements. That way, if the user
knows theyll need a certain amount of elements to be stored in a
vector, theyll be able to reserve that exactly and limit the amount
of reallocations when adding new elements. Its definition is the
following:
#+NAME: vector-vec_with_capacity-h
#+BEGIN_SRC c
Vector *vec_with_capacity(Destructor const destructor, size_t const capacity);
#+END_SRC
Under the hood, it calls ~vec_new~, then it will reallocate the memory
already allocated for the member ~elemements~.
#+NAME: vector-vec_with_capacity-c
#+BEGIN_SRC c
Vector *vec_with_capacity(Destructor const t_destructor,
size_t const t_capacity)
{
Vector *self = vec_new(t_destructor);
free(self->elements);
(*self).elements = (void *)malloc(sizeof(void *) * t_capacity);
assert(self->elements);
(*self).capacity = t_capacity;
return self;
}
#+END_SRC
*** Adding Data
The main feature of vectors is to hold data, so lets make them able
to take new data from the user. But first, let me explain a bit how
this dynamic array which I call vector works in C.
As you saw earlier, a vector is initialized with a fixed amount of
memory allocated to the vector, so people can store their data in
these arrays. Now, imagine you have an array of four elements, and you
wish to add one more, what to do? You can reallocate your array with
~realloc~ with one more slot for your element, so now you have an array
for five elements with your four original elements and a free slot for
your fifth. Cool, now you can add new elements as you need them!
Except that if you want to add some tens of thousands of new elements,
you would end up calling some tens of thousands times ~realloc~, and
that is /*slow*/. Seriously, try it, youll understand what I mean. And
all these calls to ~realloc~ are an opportunity for it to fail. Lets
limit calls to this function, OK ? If we end up short on slots in our
current array, lets actually double the amount of slots in it. So, if
we have a four-slots array, lets make it an eight-slots array, and
then a sixteen-slots array. And in a couple more calls to ~realloc~,
well quickly reach our tens of thousands slots array, way faster than
by incrementing its capacity one by one.
/“But, well end up with a lot of unused memory if we need just one more element than 2^{16} elements! We dont need a 2^{17} elements array for 2^{16}+1 elements!”/
Youre completely right, but thats a tradeoff. Would you rather have
a slow but memory-efficient program, or a fast but memory-hungry
software? Plus, as youll see later, there is a function to shrink the
size of the allocated array down to the actual amount of elements you
stored in it, making it possible to temporarily have a 2^{17} elements
array, and immediately after shrink it down to 2^{16}+1, once you know
you wont be adding any other elements.
With this out of the way, lets see how to add new elements to our
vector. First, lets declare a static function that reallocates the
memory of a vector. Here is its declaration:
#+NAME: vector-vec_realloc-def-c
#+BEGIN_SRC c
static void vec_realloc(Vector *const self) NONNULL;
#+END_SRC
Its implementation is rather simple: double its capacity, and
reallocate its array twice its previous size. Of course, there is an
assertion on whether the arrays have been correctly reallocated to
ensure memory safety.
#+NAME: vector-vec_realloc-c
#+BEGIN_SRC c
void vec_realloc(Vector *const self)
{
self->capacity *= 2;
self->elements = realloc(self->elements, sizeof(void *) * vec_capacity(self));
assert(self->elements);
return;
}
#+END_SRC
Now, we can proceed to element insertion. Here is the definition of
~vec_push~, which adds a new element at the end of the vector:
#+NAME: vector-vec_push-h
#+BEGIN_SRC c
void *vec_push(Vector *const self, void *const element) NONNULL;
#+END_SRC
As you can see, it takes as its arguments a pointer to the vector (the
same returned by its constructor) as well as a pointer to the element
to be added to the vector. This is an important point: *the vector does
not store elements themselves, only their pointer*. If the function
detects there is not enough space for a new element, a call will be
made to ~vec_realloc~ described above. Once the function is done, it
will return a pointer to the newly inserted element.
#+NAME: vector-vec_push-c
#+BEGIN_SRC c
void *vec_push(Vector *const self, void *const t_element)
{
if (vec_length(self) >= vec_capacity(self)) {
vec_realloc(self);
}
self->elements[(*self).length++] = t_element;
return vec_last(self);
}
#+END_SRC
And this is it! There may be a function added later that will allow
the insertion of a new value in any valid position between the first
and last position of an array (not counting the unused slots of said
array), and if I implement this it will imply a reimplementation of
~vec_push~ so that ~vec_push~ relies on this potential new ~vec_insert~.
*** Retrieving Data
Two functions are available when retrieving data: ~vec_safe_at~ which
safely retrieves the element at a certain index, and ~vec_at~, which is
a bit more performant but without the safety of the former. Lets see
the definition of both:
#+NAME: vector-vec_at-h
#+BEGIN_SRC c
void *vec_safe_at(Vector const *const self, size_t const index) NONNULL;
void *vec_at(Vector const *const self, size_t const index) NONNULL;
#+END_SRC
Both have the same arguments: the former is a pointer to the vector we
want to manipulate, and the latter is the index at which we want to
retrieve our data. To see the difference in how both work, lets first
see the definition of ~vec_at~:
#+NAME: vector-vec_at-c
#+BEGIN_SRC c
void *vec_at(Vector const *const self, size_t const index)
{
return self->elements[index];
}
#+END_SRC
~vec_at~ is really straightforward and is just syntax sugar around the
vectors ~elements~ member and will behave exactly like the square
brackets in standard C. However, ~vec_safe_at~ performs some additional
checks as you can see below:
#+NAME: vector-vec_safe_at-c
#+BEGIN_SRC c
void *vec_safe_at(Vector const *const self, size_t const t_index)
{
return (t_index >= vec_length(self)) ? NULL : vec_at(self, t_index);
}
#+END_SRC
If the requested index is larger than the furthest index possible, a
~NULL~ pointer will be returned, otherwise the pointer to the requested
element is. With this function, it is possible to check whether an
element has been returned or not while avoiding a possible segfault or
something similar. It could be used in a loop for instance in order to
check we only have valid elements.
It is also possible to retrieve directly the last element with
~vec_last~. Here is its definition:
#+NAME: vector-vec_last-h
#+BEGIN_SRC c
void *vec_last(Vector const *const self) NONNULL;
#+END_SRC
Just as the previous functions, its declaration is really straightforward:
#+NAME: vector-vec_last-c
#+BEGIN_SRC c
void *vec_last(Vector const *const self)
{
return vec_at(self, vec_length(self) - 1);
}
#+END_SRC
For the sake of the Object-Oriented Programming paradigm, two
functions were also declared in order to retrieve some data that could
otherwise be easily accessible:
#+NAME: vector-vec_length_capacity-h
#+BEGIN_SRC c
size_t vec_length(Vector const *const self) NONNULL;
size_t vec_capacity(Vector const *const self) NONNULL;
#+END_SRC
Their implementation is extremely trivial and doesnt really need any
explanation.
#+NAME: vector-vec_length_capacity-c
#+BEGIN_SRC c
size_t vec_length(Vector const *const self)
{
return self->length;
}
size_t vec_capacity(Vector const *const self)
{
return self->capacity;
}
#+END_SRC
*** Deleting Data
While this chapter is about destroying data, this first function will
not exactly destroy data, or at least not data we care about:
~vec_shrink_to_fit~ will reallocate the memory in our vector to make it
so that the member ~elements~ is exactly large enough to store all of
our data with no more space than that. Here is its definition:
#+NAME: vector-shrink_to_fit-h
#+BEGIN_SRC c
void vec_shrink_to_fit(Vector *const self) NONNULL;
#+END_SRC
Theres nothing too exciting about its implementation: a simple
reallocation exactly the size of the number of elements currently
stored times the size of a void pointer, and we verify with an ~assert~
if it has been correctly reallocated. Nothing is returned.
#+NAME: vector-shrink_to_fit-c
#+BEGIN_SRC c
void vec_shrink_to_fit(Vector *const self)
{
if (self->length <= 0) {
return;
}
self->capacity = self->length;
self->elements = realloc(self->elements, sizeof(void *) * vec_capacity(self));
assert(self->elements);
return;
}
#+END_SRC
Notice that a check is done to see if the vector exists, because
otherwise calling ~shrink_to_fit~ on an empty vector would result in an
error while asserting the reallocation.
Next, we have two functions: ~vec_pop_at~ and ~vec_pop~. The latter relies
on the former, which can delete an element at any valid position.
*Beware*: these functions return /nothing/ and simply deletes the element.
Here is their definition:
#+NAME: vector-vec_pop-h
#+BEGIN_SRC c
void vec_pop_at(Vector *const self, size_t const index) NONNULL;
void vec_pop(Vector *const self) NONNULL;
#+END_SRC
In order to insure memory safety, a static function is declared in
~src/vector.c~ which will delete an element if a destructor has been
provided to the vector when it has been built. Its definition is the
following:
#+NAME: vector-vec_maybe_delete_element-def-c
#+BEGIN_SRC c
static void vec_maybe_delete_element(Vector const *self,
size_t const t_index) NONNULL;
#+END_SRC
Its implementation is quite simple: if a destructor exists, then the
element at the requested index will be destroyed through this
destructor. Otherwise, nothing is done with the destructor, hence the
name of the function ~vec_maybe_delete_element~. However, it should be
noted that the element will be freed from memory, so if the user needs
it before popping it, they need to retrieve it with something like
~vec_at~ and store it elsewhere.
#+NAME: vector-vec_maybe_delete_element-c
#+BEGIN_SRC c
void vec_maybe_delete_element(Vector const *self, size_t const t_index)
{
void *element = vec_at(self, t_index);
if (self->destroy) {
self->destroy(element);
}
free(element);
}
#+END_SRC
Now that we have this function sorted out, we can implement our pops.
Here is the implementation of ~vec_pop_at~:
#+NAME: vector-vec_pop_at-c
#+BEGIN_SRC c
void vec_pop_at(Vector *const t_self, size_t const t_index)
{
if (vec_safe_at(t_self, t_index) == NULL) {
return;
}
vec_maybe_delete_element(t_self, t_index);
if (t_index + 1 < vec_length(t_self)) {
memcpy(vec_at(t_self, t_index), vec_at(t_self, t_index + 1),
sizeof(void *) * (t_self->length - (t_index + 1)));
}
--(*t_self).length;
}
#+END_SRC
A check is performed at the beginning of the function: that the
element we want to pop actually exists. If it does not, the function
does nothing, otherwise the function deletes the element if needed.
The call to ~vec_maybe_delete_element~ will free the requested element.
Then, a check is performed to see if the requested element was at the
end of the array or not. If it was not, then the elements located
after the destroyed element are shifted one element closer to the
beginning of the array; otherwise, if the requested element was at the
end of the array, nothing is done particularly. Lastly, the count of
elements stored in the vector is decreased by one.
~vec_pop~ uses the above function in order to provide a simpler call if
we want to delete the last element of the array. We can see how it
relies on ~vec_pop_at~ in its implementation:
#+NAME: vector-vec_pop-c
#+BEGIN_SRC c
void vec_pop(Vector *const self)
{
vec_pop_at(self, vec_length(self));
}
#+END_SRC
Finally, ~vec_delete~ allows for the complete destruction and
deallocation of a vector, including all of its elements. Here is its
definition:
#+NAME: vector-vec_delete-h
#+BEGIN_SRC c
void vec_delete(Vector *const self) NONNULL;
#+END_SRC
In its implementation, we can see three distinct steps:
- The deletion of all its elements if a destructor exists
- The deletion of the array of the vector
- The deletion of the vector itself.
#+NAME: vector-vec_delete-c
#+BEGIN_SRC c
void vec_delete(Vector *const self)
{
if (self->destroy) {
for (size_t i = 0; i < vec_length(self); ++i) {
self->destroy(self->elements[i]);
}
}
free(self->elements);
free(self);
}
#+END_SRC
*** The Final Source Code
Finally, we can see the whole source code. Here is the header for the
library: ~vector.h~
#+BEGIN_SRC c :noweb yes
#ifndef VECTOR_H_
#define VECTOR_H_
<<vector-nonnull-h>>
<<vector-struct-def>>
<<vector-vec_new-h>>
<<vector-vec_with_capacity-h>>
<<vector-vec_push-h>>
<<vector-vec_at-h>>
<<vector-vec_last-h>>
<<vector-vec_length_capacity-h>>
<<vector-shrink_to_fit-h>>
<<vector-vec_pop-h>>
<<vector-vec_delete-h>>
#endif /* VECTOR_H_ */
#+END_SRC
And here is the implementation file: ~vector.c~
#+BEGIN_SRC c :noweb yes
#include "vector.h"
<<vector-includes-c>>
<<vector-vec_realloc-def-c>>
<<vector-vec_maybe_delete_element-def-c>>
<<vector-vec_new-c>>
<<vector-vec_with_capacity-c>>
<<vector-vec_realloc-c>>
<<vector-vec_push-c>>
<<vector-vec_at-c>>
<<vector-vec_safe_at-c>>
<<vector-vec_last-c>>
<<vector-vec_length_capacity-c>>
<<vector-shrink_to_fit-c>>
<<vector-vec_pop-c>>
<<vector-vec_maybe_delete_element-c>>
<<vector-vec_pop_at-c>>
<<vector-vec_pop-c>>
<<vector-vec_delete-c>>
#+END_SRC
And with that, we should be good! I used this library in a SOM
(Kohonen, 1982) implementation and ran it through valgrind, and there
were no memory leaks. If you find one though, dont hesitate telling
me in the comments, through social media such as Twitter, or by email.
Happy programming!
* Emacs :@emacs:
** Emacs 29 is nigh! What can we expect? :dev:emacs:
:PROPERTIES:
:EXPORT_FILE_NAME: emacs-29-what-can-we-expect
:EXPORT_DATE: 2022-11-29
:EXPORT_OPTIONS: toc:2
:export_hugo_menu: :menu "main"
:END:
It [[https://lists.gnu.org/archive/html/emacs-devel/2022-11/msg01774.html][was announced a couple of hours ago]], Emacs 29s branch is now cut
from the master branch! This means the ~emacs-29~ branch will from now
no longer receive any new feature, but only bug fixes.
So, whats new with this new major release? I skimmed over the ~NEWS~
file, and here are the changes which I find interesting and even
exciting for some.
*Article updated on December 22nd at 14:05 UTC*
*** Major features
A couple of major improvements will be most likely present, here are
the ones that stand out the most for me.
**** Eglot is now part of Emacs core
During the last couple of years, LSP has given text editors incredible
capabilities, giving them IDE-like features relatively easily. Aside
from Elisp development, most of the code I write is now done with the
help of an LSP server, running along Emacs and analysing my code,
suggesting and performing changes and actions for me.
Several integrations of LSP exist for Emacs, such as [[https://emacs-lsp.github.io/lsp-mode/][LSP Mode]], [[https://github.com/joaotavora/eglot][Eglot]],
and [[https://github.com/manateelazycat/lsp-bridge][lsp-bridge]]. Among the three, Eglot is now part of Emacs core! No
longer do you need to install a package, simply register an LSP server
and autocompletion, documentation, error detection, and other features
will become available right away!
I must admit I dont really know Eglot, I personally use LSP Mode, but
with this addition to Emacs core, I might attempt the switch.
**** Tree-Sitter is also part of Emacs core
In case you didnt know, Emacs current syntax highlighting is
currently based on a system of regexes. Although it is not the /worst/
thing to use, its not the best either, and it can become quite slow
on larger files.
Tree-Sitter parses programming languages based into a concrete syntax
tree. From there, not only can syntax highlighting can be done at high
speed, but a much deeper analysis of the code is possible and actions
such as syntax manipulation can also be achieved since the syntax tree
itself is available as an object which can be manipulated!
In case you want some more information on Tree-Sitter itself, you can
check out the [[https://tree-sitter.github.io/tree-sitter/][official Tree-Sitter website]], or you can even check this
talk out given by Tree-Sitters creator, Max Brunsfeld.
#+begin_export html
<iframe width="560" height="315" src="https://www.youtube-nocookie.com/embed/Jes3bD6P0To" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
#+end_export
Well, this is now a native solution in Emacs! Currently, Emacs
Tree-Sitter supports the current major modes:
- ~bash-ts-mode~
- ~c-ts-mode~
- ~c++-ts-mode~
- ~csharp-ts-mode~
- ~css-ts-mode~
- ~java-ts-mode~
- ~js-ts-mode~
- ~json-ts-mode~
- ~python-ts-mode~
- ~typescript-ts-mode~
Tree-Sitter also holds for now a special status in the new ~emacs-29~
branch since new features can still be added to it, as its merge with
the master branch is still recent. So we might see the list of major
modes for Emacs get a bit longer yet, especially considering
Tree-Sitter tries to make adding new languages relatively easy.
If you cant wait to test Tree-Sitter, there is already [[https://emacs-tree-sitter.github.io/][another package]]
available for Emacs you can use right now. Just be aware this is not
the same package as the one that got integrated into Emacs.
**** Install packages from source with ~package.el~
If you use [[https://github.com/radian-software/straight.el][Straight]], you might be familiar with installing packages
directly from their Git repository. Well, good news, it is now
possible to install packages from Git using Emacs built-in packaging
system ~package.el~! It can be done with the new function
~package-vc-install~, and packages installed that way can be updated
with ~package-vc-update~ or ~package-vc-update-all~.
On the topic of ~package.el~, there is also the new function
~package-report-bug~ which allows Emacs users to report bugs to the
developers of a package directly from Emacs! Be aware though, it only
works for packages installed through ~package.el~. Since Im a
[[https://github.com/jwiegley/use-package][~use-package~]] and ~straight.el~ user, there is no package listed when I
invoke the command.
**** Org mode 9.6
As confirmed by one of org-mode maintainers [[https://bzg.fr/][Bastien Guerry]] on [[https://lists.sr.ht/~bzg/emacsfr/%3C87bkophdzo.fsf%40phundrak.com%3E#%3C87tu2hqf4h.fsf@gnu.org%3E][a
French-speaking Emacs mailing list]], Org 9.6 is set to be part of
Emacs 29! There is an [[https://orgmode.org/Changes.html][official article]] on this release, which is
[[https://elpa.gnu.org/packages/org.html][already available on GNU ELPA]]!
**** use-package in Emacs core
It has also been confirmed on the [[https://lists.gnu.org/archive/html/emacs-devel/2022-11/msg01821.html][Emacs development mailing list]] that
[[https://github.com/jwiegley/use-package][~use-package~]], an awesome package manager, is set to be part of Emacs
29, although it initially wasnt included in the ~emacs-29~ branch.
**** Pure GTK Emacs is here for Wayland!
One of the major issues Emacs had on Linux was its dependency on Xorg
when running in GUI mode. When running Xorg, its not really an issue,
but Wayland has become more and more common during the last years, and
even with the existence of XWayland, this became an annoyance.
Well, fear not, for pure GTK Emacs is here! It can now be built
Xorg-free and run natively in Wayland!
Be aware though that Wayland is basically the only use-case for pure
GTK Emacs. If you dont use Wayland, Emacs will display a warning
message, as it will most likely cause issues if you are running Xorg.
In my case, I sometimes see some ghost text when the content of a
buffer updates (I still need pure GTK though, since I alternate
between Xorg and Wayland).
**** Compile EmacsLisp files ahead of time
With Emacs 28 came the ability to natively compile EmacsLisp if your
Emacs was built with the ability to do so, using GCCs Just In Time
library. This results in quite the impressive boost in performance,
which made Emacs much snappier than it was before. The only issue I
had was Emacs would only compile its EmacsLisp files when they were
loaded for the first time.
This is no longer the case! If you now compile Emacs with
~--with-native-compilation=aot~, Emacs native EmacsLisp files will be
natively compiled along with Emacs itself! Be aware though, it can be
slow on most machines, so the time you save by not compiling these
files when launching Emacs for the first time is basically transferred
to when compiling Emacs itself. Is it worth your time? In my case, I
would say yes, because when I compile Emacs, Im generally not in a
hurry. But in your case? Well, test it out and see for yourself.
**** Native access to SQLite databases
Emacs can now be built with native support for SQLite and the sqlite3
library. In fact, this is now a default behaviour, since you need to
pass ~--without-sqlite3~ to Emacs build configuration script in order
to prevent it.
This comes with a new ~sqlite-mode~ which allows you to explore SQLite
databases within Emacs and to interact with them. Check out the
~sqlite-mode-open-file~ function!
**** HaikuOS support
For all three HaikuOS users out there, good news, you now have access
to Emacs! (In all seriousness, I should check out HaikuOS one day)
Moreover, it also supports an optional window-system port to Haiku
with ~--with-be-app~. Be aware, you will need the Haiku Application Kit
development headers and a C++ compiler. Otherwise, Emacs will only run
in the terminal. If you want to also add Cairo to the mix, you can add
~--with-be-cairo~.
**** New major mode for C#
~csharp-mode~ is now a native major mode for Emacs and is based on ~cc-mode~.
*** Minor features
**** Its easier to use Emacs in scripts!
If you like to write scripts and especially writing Lisp scripts,
Emacs now supports the option ~-x~ in order to execute scripts written
in EmacsLisp. When executing such a script with ~#!/usr/bin/emacs -x~ as
its shebang, Emacs will not read its init file (like with ~-Q~) and will
instead execute the Elisp code right away and return the last value to
the caller of the script (most likely the shell you called the script
from).
**** TRAMP natively supports Docker, Podman, and Kubernetes
Three new connections are now available for TRAMP:
- ~docker~
- ~podman~
- ~kubernetes~
You will now be able to access your containerized environment right
from Emacs without the need to write custom code.
**** Custom user directory
It is now easier to launch custom Emacs profiles without the need of
tools such as [[https://github.com/plexus/chemacs2][chemacs2]] with the addition of the flag ~--init-directory~.
This can set to any directory Emacs ~user-emacs-directory~ which
includes the ~init.el~ which comes along with it. Yet another reason for
me not to use a ~.emacs~ file, but the ~init.el~ file instead.
**** Support for Webp images
For quite some time, Emacs has been able to display images, but not
webp yet. Well, this is now fixed! And in fact, support for webp
images became the default behaviour, since you need to pass
~--without-webp~ to Emacs configuration script to disable webp support.
**** C++ mode now supports the C++20 standard
Yep. Theres nothing more to say, really. Happy coding!
**** Better handling of ~.pdmp~ files
Emacs has had for a few version the ability to dump its state into a
~pdmp~ file for faster startup time. Well now, when creating such a
file, it will include in its name a fingerprint of its current state,
although it will still prioritize an ~emacs.pdmp~ file if it exists.
**** Better mouse and touchpad support
Emacs now uses XInput 2, which enables Emacs to support more input
events, such as touchpad events. For instance, by default, a pinch
gesture on a touchpad increases or decreases the text size of the
current buffer. This is thanks to the new event ~pinch~, which comes
along with ~touch-end~.
**** Unicode 15.0 and emojis
Emacs now supports [[https://www.unicode.org/versions/Unicode15.0.0/][Unicode 15.0]], which is currently the latest Unicode
version. Although this is not directly related, quite a few new
emoji-related features have been introduced. The new prefix ~C-x 8 e~
now leads to a few new commands related to emojis:
- ~C-x 8 e e~ or ~C-x 8 e i~ :: Insert an emoji (~emoji-insert~)
- ~C-x 8 e s~ :: Search an emoji (~emoji-search~)
- ~C-x 8 e l~ :: List all emojis in a new buffer (~emoji-list~)
- ~C-x 8 e r~ :: Insert a recently inserted emoji (~emoji-recent~)
- ~C-x 8 e d~ :: Describe an emoji (~emoji-describe~)
- ~C-x 8 e +~ and ~C-x 8 e -~ :: Increase and decrease the size of any
character, but especially emojis (~emoji-zoom-increase~ and ~emoji-zoom-decrease~ respectively)
There is also the new input method ~emoji~ which allows you to type for
instance ~:grin:~ in order to get the emoji 😁.
**** True background transparency
Up until recently, if you wanted transparency with Emacs, you had no
choice but to make the whole frame transparent, including text and
images.
Thanks to the frame parameter ~alpha-background~ and its related
~alphaBackground~ X resource, it is now possible to set transparency
only for the frames background without affecting any of the other
elements on screen.
**** WebKit inspector in Emacs WebKit widget browser
You can now access the WebKit inspector when using the WebKit widget
browser in Emacs, given you are using a version of Emacs which has
been compiled with it. I wish there was a keybinding or at least a
function for it, but apparently you can only open it with a right
click and select /Inspect Element/. Still nice to have.
**** Some news for Windows
Although it has been available for Linux users since Emacs 26.1,
Windows finally has access to double-buffering to reduce display
flicker. If you wish to disable it, you can set the frame parameter
~inhibit-double-buffering~ to ~nil~.
Emacs also follows Windows dark mode with Windows 10 (version 1809)
and onwards.
Emacs also now uses Windows native API to render images. This
includes BMP, GIF, JPEG, PNG, and TIFF images. Other formats, however,
still rely on other dependencies and libraries to properly work, such
as Webp images.
*** Whats next?
With Emacs 29 being cut, development on the master branch will now go
towards Emacs 30. Is there anything we can expect yet?
Its still very early to say, most stable features merged into master
went to Emacs 29, and only the ~feature/pkg~ and
~feature/improved-lock-narrowing~ branches seem to have received commits
less than a week prior to the day of writing this, and I do not know
the status of other branches that received commits during the past few
weeks such as ~feature/package+vc~ or ~feature/eglot2emacs~ (which I
assume both got merged).
+However, there are currently talks about including ~use-package~ into Emacs! Im a bit disappointed it wont make it into Emacs 29, but progress is being made on ~scratch/use-package~, and you can always check the mailing list to check its status such as [[https://lists.gnu.org/archive/html/emacs-devel/2022-11/msg01533.html][here]].+ *Update*: Rejoice! As mentioned above, ~use-package~ is actually set to land in Emacs 29!
** [EN] Automatic Meaningful Custom IDs for Org Headings :emacs:orgmode:dev:
:PROPERTIES:
:EXPORT_FILE_NAME: better-custom-ids-orgmode
:EXPORT_DATE: 2020-06-06
:export_hugo_menu: :menu "main"
:END:
Spoiler alert, I will just modify a bit of code that already exists,
go directly to the bottom if you want the solution, or read the whole
post if you are interested in how I got there.
#+TOC: headlines 1 local
**** Update 2021-11-22
Ive put the code presented here as a complete package. You can find
it in [[https://labs.phundrak.com/phundrak/org-unique-id][this repository]] or in its [[https://github.com/Phundrak/org-unique-id][GitHub mirror]] (be aware the latter may
not be as up-to-date as the former is. Installation instructions are
in the README.
*** The issue
About two to three years ago, as I was working on a project that was
meant to be published on the internet, I looked for a solution to get
fixed anchor links to my various headings when I performed HTML
exports. As some of you may know, by default when an Org file is
exported to an HTML file, a random ID will be generated for each
header, and this ID will be used as their anchor. Heres a quick
example of a simple org file:
#+caption: Example org file
#+begin_src org :exports code
,#+title: Sample org file
,* First heading
Reference to a subheading
,* Second heading
Some stuff written here
,** First subheading
Some stuff
,** Second subheading
Some other stuff
#+end_src
And this is the result once exported to HTML (with a lot of noise
removed from ~<head>~):
#+caption: Output HTML file
#+BEGIN_SRC html
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
<head>
<title>Sample org file</title>
<meta name="generator" content="Org mode" />
<meta name="author" content="Lucien Cartier-Tilet" />
</head>
<body>
<div id="content">
<h1 class="title">Sample org file</h1>
<div id="outline-container-orgd8e6238" class="outline-2">
<h2 id="orgd8e6238"><span class="section-number-2">1</span> First heading</h2>
<div class="outline-text-2" id="text-1">
<p>
Reference to a subheading
</p>
</div>
</div>
<div id="outline-container-org621c39a" class="outline-2">
<h2 id="org621c39a"><span class="section-number-2">2</span> Second heading</h2>
<div class="outline-text-2" id="text-2">
<p>
Some stuff written here
</p>
</div>
<div id="outline-container-orgae45d6b" class="outline-3">
<h3 id="orgae45d6b"><span class="section-number-3">2.1</span> First subheading</h3>
<div class="outline-text-3" id="text-2-1">
<p>
Some stuff
</p>
</div>
</div>
<div id="outline-container-org9301aa9" class="outline-3">
<h3 id="org9301aa9"><span class="section-number-3">2.2</span> Second subheading</h3>
<div class="outline-text-3" id="text-2-2">
<p>
Some other stuff
</p>
</div>
</div>
</div>
</div>
</body>
</html>
#+END_SRC
As you can see, all the anchors are in the format of ~org[a-f0-9]{7}~.
First, this is not really meaningful if you want to read the anchor
and guess where it will lead you. But secondly, these anchors will
change each time you export your Org file to HTML. If I want to share
a URL to my website and to a specific heading, … well I cant, it will
change the next time I update the document. And I dont want to have
to set a ~CUSTOM_ID~ property for each one of my headings manually. So,
what to do?
*** A first solution
A first solution I found came from [[https://writequit.org/articles/emacs-org-mode-generate-ids.html][this blog post]], where Lee Hinman
described the very same issue they had and wrote some Elisp code to
remedy that (its a great read, go take a look). And it worked, and
for some time I used their code in my Emacs configuration file in
order to generate unique custom IDs for my Org headers. Basically what
the code does is it detects if ~auto-id:t~ is set in an ~#+OPTIONS~
header. If it is, then it will iterate over all the Org headers, and
for each one of them it will insert a ~CUSTOM_ID~, which is made from a
UUID generated by Emacs. And tadah! we get for each header a
~h-[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}~ custom
ID that wont change next time we export our Org file to HTML when we
save our file, and only for headings which dont already have a
~CUSTOM_ID~ property. Wohoo!
Except…
*** These headers are not meaningful
OK, alright, thats still a huge step forward, we dont have to type
any ~CUSTOM_ID~ property manually any more, its done automatically for
us. But, when I send someone a link like
~https://langue.phundrak.com/eittland#h-76fc0b91-e41c-42ad-8652-bba029632333~,
the first reaction to this URL is often something along the lines of
“What the fuck?”. And theyre right, this URL is unreadable when it
comes to the anchor. How am I supposed to guess it links to the
description of the vowels of the Eittlandic language? (Thats a
constructed language Im working on, you wont find anything about it
outside my website. Also, this link is dead now, it got simplified
thanks to Vuepres.)
So, I went back to my configuration file for Emacs, and through some
trial and error, I finally found a way to get a consistent custom ID
which is readable and automatically set. With the current state of my
code, what you get is the complete path of the Org heading, all spaces
replaced by underscores and headings separated by dashes, with a final
unique identifier taken from an Emacs-generated UUID. Now, the same
link as above will look like
~https://langue.phundrak.com/eittland#Aperçu_structurel-Inventaire_phonétique_et_orthographe-Voyelles_pures-84f05c2c~.
It wont be more readable to you if you dont speak French, but you
can guess it is way better than what we had before. I even added a
safety net by replacing all forward slashes with dashes. The last ID
is here to ensure the path will be unique in case wed have two
identical paths in the org file for one reason or another.
The modifications I made to the first function ~eos/org-id-new~ are
minimal, where I just split the UUID and get its first part. This is
basically a way to simplify it.
#+BEGIN_SRC emacs-lisp
(defun eos/org-id-new (&optional prefix)
"Create a new globally unique ID.
An ID consists of two parts separated by a colon:
- a prefix
- a unique part that will be created according to
`org-id-method'.
PREFIX can specify the prefix, the default is given by the
variable `org-id-prefix'. However, if PREFIX is the symbol
`none', don't use any prefix even if `org-id-prefix' specifies
one.
So a typical ID could look like \"Org-4nd91V40HI\"."
(let* ((prefix (if (eq prefix 'none)
""
(concat (or prefix org-id-prefix)
"-"))) unique)
(if (equal prefix "-")
(setq prefix ""))
(cond
((memq org-id-method
'(uuidgen uuid))
(setq unique (org-trim (shell-command-to-string org-id-uuid-program)))
(unless (org-uuidgen-p unique)
(setq unique (org-id-uuid))))
((eq org-id-method 'org)
(let* ((etime (org-reverse-string (org-id-time-to-b36)))
(postfix (if org-id-include-domain
(progn
(require 'message)
(concat "@"
(message-make-fqdn))))))
(setq unique (concat etime postfix))))
(t (error "Invalid `org-id-method'")))
(concat prefix (car (split-string unique "-")))))
#+END_SRC
Next, we have here the actual generation of the custom ID. As you can
see, the ~let~ has been replaced by a ~let*~ which allowed me to create
the ID with the variables ~orgpath~ and ~heading~. The former concatenates
the path to the heading joined by dashes, and ~heading~ concatenates
~orgpath~ to the name of the current heading joined by a dash if ~orgpath~
is not empty. It will then create a slug out of the result, deleting
some elements such as forward slashes or tildes, and all whitespace is
replaced by underscores. It then passes ~heading~ as an argument to the
function described above to which the unique ID will be concatenated.
#+BEGIN_SRC emacs-lisp
(defun eos/org-custom-id-get (&optional pom create prefix)
"Get the CUSTOM_ID property of the entry at point-or-marker POM.
If POM is nil, refer to the entry at point. If the entry does not
have an CUSTOM_ID, the function returns nil. However, when CREATE
is non nil, create a CUSTOM_ID if none is present already. PREFIX
will be passed through to `eos/org-id-new'. In any case, the
CUSTOM_ID of the entry is returned."
(interactive)
(org-with-point-at pom
(let* ((orgpath (mapconcat #'identity (org-get-outline-path) "-"))
(heading (replace-regexp-in-string
"/\\|~\\|\\[\\|\\]" ""
(replace-regexp-in-string
"[[:space:]]+" "_" (if (string= orgpath "")
(org-get-heading t t t t)
(concat orgpath "-" (org-get-heading t t t t))))))
(id (org-entry-get nil "CUSTOM_ID")))
(cond
((and id
(stringp id)
(string-match "\\S-" id)) id)
(create (setq id (eos/org-id-new (concat prefix heading)))
(org-entry-put pom "CUSTOM_ID" id)
(org-id-add-location id
(buffer-file-name (buffer-base-buffer)))
id)))))
#+END_SRC
The rest of the code is unchanged, here it is anyway:
#+BEGIN_SRC emacs-lisp
(defun eos/org-add-ids-to-headlines-in-file ()
"Add CUSTOM_ID properties to all headlines in the current file
which do not already have one.
Only adds ids if the `auto-id' option is set to `t' in the file
somewhere. ie, #+OPTIONS: auto-id:t"
(interactive)
(save-excursion
(widen)
(goto-char (point-min))
(when (re-search-forward "^#\\+OPTIONS:.*auto-id:t"
(point-max)
t)
(org-map-entries (lambda ()
(eos/org-custom-id-get (point)
'create))))))
(add-hook 'org-mode-hook
(lambda ()
(add-hook 'before-save-hook
(lambda ()
(when (and (eq major-mode 'org-mode)
(eq buffer-read-only nil))
(eos/org-add-ids-to-headlines-in-file))))))
#+END_SRC
Note that you *will need* the package ~org-id~ to make this code work. You
simply need to add the following code before the code I shared above:
#+BEGIN_SRC emacs-lisp
(require 'org-id)
(setq org-id-link-to-org-use-id 'create-if-interactive-and-no-custom-id)
#+END_SRC
And thats how my links are now way more readable *and* persistent! The
only downside I found to this is when you move headings and their path
is modified, or when you modify the heading itself, the custom ID is
not automatically updated. I could fix that by regenerating the custom
ID on each save, regardless of whether a custom ID already exists or
not, but its at the risk an ID manually set will get overwritten.
* Linux :@linux:
** [EN] My YouTube subscriptions as an RSS feed :linux:dev:tutorial:
:PROPERTIES:
:EXPORT_FILE_NAME: youtube-subscriptions-rss
:EXPORT_DATE: 2022-02-04
:export_hugo_menu: :menu "main"
:END:
*** The Problem
Im sure youve been in the same situation before: you go on YouTube
because you want to watch a video, maybe two, from your subscriptions.
You open the first one. Oh great, an unskippable fifteen seconds ad.
And another one! OK, the video starts. It gets cut a couple of times
by other ads of varying length. Oh but whats this? This recommended
video looks nice! And before you know it, your whole afternoon and
evening went by painfully watching videos on YouTubes atrocious video
player. You lost focus.
*** My Solution: mpv + RSS
Wouldnt it be nice if it were possible to watch these videos with a
full-fledged video player over which you have complete control? Which
could be customized to your hearts content? Which wont secretly
track what you watch?
Oh right, [[https://mpv.io/][mpv]]! It supports most video formats you can think of, and
thanks to its interoperability with [[https://github.com/ytdl-org/youtube-dl][youtube-dl]], you can also watch
videos from [[https://ytdl-org.github.io/youtube-dl/supportedsites.html][an extremely wide variety of websites]]! So why not YouTube?
Now, the question is how to get rid of YouTubes interface. The answer
is actually quite simple: lets use an RSS feed. With the RSS feeds
from YouTube, you will receive in your RSS reader the link of the
video with its thumbnail and its description. You can then copy from
there the link and open it with mpv with a command like this:
#+begin_src bash
mpv "https://www.youtube.com/watch?v=xym2R6_Qd7c"
#+end_src
**** Channel RSS
Now the question is how to get the RSS feed of a channel? The answer
is quite simple. The base URL for a YouTube channel RSS feed is
~https://www.youtube.com/feeds/videos.xml?channel_id=~ to which you
simply have to add the channel ID. For instance, if you want to follow
Tom Scott with this, you simply have to extract the part of the
channel after ~/channel/~ in his URL and append it to the URL mentioned
above, and TADAH! you get an RSS feed to his channel!
#+begin_src text
https://www.youtube.com/feeds/videos.xml?channel_id=UCBa659QWEk1AI4Tg--mrJ2A
#+end_src
Be careful to select the channel ID only if it is after a ~/channel/~
though! The part that is after a ~/c/~ will not work. If you end up on
the URL ~https://www.youtube.com/c/TomScottGo~, simply click on a random
video, then click on the channels name. This should bring you back to
the channel but with an important difference: the URL is now
~https://www.youtube.com/channel/UCBa659QWEk1AI4Tg--mrJ2A~.
The thing that is really nice with this setup is you dont really need
to actually subscribe to a channel, your RSS feed already does that
for you! And with lots of RSS feed readers, you can categorize your
different feeds, meaning you can even categorize your subscriptions!
**** Playlist RSS
It is also possible to follow not only a channel but a playlist of
videos. For that, you will instead use
~https://www.youtube.com/feeds/videos.xml?playlist_id=~ as your base URL
to which you will add the ID of the playlist you want to follow. For
instance, with Tom Scotts playlist for Citation Needed Season 7, the
URL of the playlist is
~https://www.youtube.com/playlist?list=PL96C35uN7xGI15-QbtUD-wJ5-G8oBI-tG~,
which means you need to keep the ~PL96C35uN7xGI15-QbtUD-wJ5-G8oBI-tG~
and put it into the URL like so:
#+begin_src text
https://www.youtube.com/feeds/videos.xml?playlist_id=PL96C35uN7xGI15-QbtUD-wJ5-G8oBI-tG
#+end_src
*** Which RSS reader to go with?
If you know me, youll know I am extremely biased towards Emacs, so of
course Ill recommend Elfeed to any Emacs user ([[https://config.phundrak.com/emacs#Packages-Configuration-Applications-Elfeedoip0fl6184j0][my relevant
configuration is here]]). I even wrote an advice around
~elfeed-show-visit~ to ensure YouTube videos are open with mpv instead
of my web browser.
If youre not into Emacs, or not /that/ into Emacs, you can also try
other alternatives such as [[https://gitlab.com/news-flash/news_flash_gtk][NewsFlash]], a very nice RSS reader written
in GTK for Linux I may not always agree with DistroTube, but he made
a [[https://www.youtube.com/watch?v=KBAmviddh4A][very nice video]] presenting this piece of software. (Remember,
right-click and then ~mpv "the url here"~!)
The [[https://apps.nextcloud.com/apps/news][News app]] for Nextcloud is also very neat, I recommend you using it.
You can also get your RSS feed in your terminal with [[https://newsboat.org/][Newsboat]]. Not
really my cup of tea, but I can see why some people enjoy it.
*** Improving a bit the mpv tooling
You might have heard it, but youtube-dl hasnt been doing great
recently. The tool is becoming slow, and it lacks quite a few features
it could really benefit from. While it is important to acknowledge its
historical importance, I think it is now time to move on, and its
successor shall be [[https://github.com/yt-dlp/yt-dlp][yt-dlp]]. In my experience, this youtube-dl fork is
much faster than youtube-dl itself on top of providing additional
features such as [[https://github.com/yt-dlp/yt-dlp#sponsorblock-options][SponsorBlock integration]].
How do you replace youtube-dl with yt-dlp then? If you use Arch Linux
or one of its derivates (I hope not Manjaro though), you can simply
install ~yt-dlp-drop-in~ from the AUR.
#+begin_src bash
paru -S yt-dlp-drop-in
# or if you prefer yay
yay -S yt-dlp-drop-in
# or whichever AUR helper you prefer, as long as it is NOT yaourt
#+end_src
If you are not an Arch Linux user, check out [[https://www.funkyspacemonkey.com/replace-youtube-dl-with-yt-dlp-how-to-make-mpv-work-with-yt-dlp][this article]], it will
help you.