The more blog posts I write the harder its getting to find the one I am looking for at a glance. Since the filenames are sorted in lexicographical order and I name files solely by topic with no consideration of their resulting order, adding a new filename to me is tantamount to tossing a needle in a haystack. I create the post and when I save the file I think: "Where did it go?". A post I created today and one I created five minutes ago might be miles apart. Conversely, a post created five years ago and one created five minutes ago might be right next to each other. The experience is disorienting and costly as I end up spending more time searching for posts. Of course, I can always use consult-line from consult but I do not want to do this every time. Ultimately, I want posts for my blog to be in the order that they where created. To that end, I decided to prefix a timestamp at the beginning of the filename of each post.
I did not know how timestamps were normally written. Should there be some separator between certain time components? If so what is it and where should it go? I suspect the answer largely depends on use-case and preference.
In a question similar to my own the following timestamp format was generated using format-time-string.
(format-time-string "%Y-%m-%d--%H-%M-%S")
;; => "2024-04-24--14-31-16"
In contrast the timestamp used for org IDs which is much more conservative about separators. It only uses "T" to separate the year and a "." to separate anything smaller than a second.
(require 'org-id)
(org-id-new)
;; => "20240424T141308.453178"
Not wanting excessively long filenames, I was inclined to avoid using separators. It was when I actually tried using timestamps with no separators that I realized how difficult it is to distinguish time components without them. Compare with this. I think the year is easy to see but suppose I wanted the hour and minute instead. I for one would have to sit there a few seconds trying to figure it out.
(format-time-string "%Y%m%d%H%M%S%N")
;; => "20240504145659269408809"
It is apparent that the question of whether to have separators or not is a trade off between readability and brevity. Then again the question. After all, if I never need to read the timestamp myself and can manipulate them programmatically then.
Ultimately, I opted for readability by separating individual time components with either a "-" or "–". Although the minor seconds are not pretty I still leave them so I can be sure that filenames are different even if they are created in a short time span from each other.
(format-time-string "%Y-%m-%d--%H-%M-%S--%N")
;; =>; "2024-05-01--05-04-45--204039545"
Typing out the timestamp is too error prone and tedious. Instead, I wrote the following function that prompts me for a blog post title and then creates a blog post from the given name, opening the file in Emacs1 and inserting the title and author information.
(defun oo-blog-new-post (title)
"Create a new blog post."
(interactive "sBlog post title: ")
(find-file (oo--blog-post-filename title))
(insert (format "#+TITLE: %s\n" (capitalize title)))
(insert (format "#+AUTHOR: Luis Henriquez-Perez\n" (capitalize title))))
For testing it is often much easier to segregate side-effects (the creation of
the file) from computations (the generation of the path). That is why even
though the body of this function is small enough to fit in oo-blog-new-post
without making it excessively long, I place it into its own function.
(defun oo--blog-post-filename (title)
"Return a valid blog post name from TITLE.
TITLE is a string describing the topic of a post."
(let ((timestamp (format-time-string "%Y-%m-%d--%H-%M-%S--%N"))
(name (downcase (string-replace "\s" "-" title)))
(post-dir "~/Documents/blog/org/posts/"))
(expand-file-name (format "%s--%s.org" timestamp name) post-dir)))
This way I can test whether the correct filename is generated without worrying about whether inadvertently creating a file in the wrong directory in case of I get the code wrong.
(abbreviate-file-name (oo--blog-post-filename "writing a post while upside down"))
;; => "~/Documents/blog/org/posts/2024-05-02--12-11-11--127335202--writing-a-post-while-upside-down.org"
(abbreviate-file-name (oo--blog-post-filename "writing a post underwater"))
;; => "~/Documents/blog/org/posts/2024-05-02--12-11-23--036941798--writing-a-post-underwater.org"
I could generalize this function maybe for just generating a timestamp filename. I could definitely see myself wanting more such files in the future. But since I have no concrete need as of yet I will leave it as is.
After adding timestamps I have achieved what I wanted. Now blog posts are displayed in a predictable order when viewed via ls or any file viewer.
Figure 1: The dired buffer of the directory containing my blog posts.
To be more precise, it creates a buffer that visits the post file using find-file.