Hacker News Reader

Comment by hardwaresofton | Hacker News Reader

I am also that age, and kebab-case is the best case for filenames.

2021-01-01-some-important-document.pdf gives me the warm fuzzies. On the off chance that some more differentiation is needed, throw in an underscore and a whole new world opens up

View on HN

I go one step further: 2021-11-11_client_project-name.ext

2021-11-11_client_projectName.ext is also OK. But underscore separates fields, hyphens for space replacement.

this is the way

but the extra Shifts, no thank you

you gotta involve your pinky or it'll atrophy

Cut most mine off in an unsupervised Halloween pumpkin carving accident when I was a kid. I think the lack of length actually allows me to type faster.

I see and applaud your use of the underscore there, but I must reject the premise!

work/client/project/2021-11-11-file.ext is more or less how I lay stuff out. I’d say client/project is a folder level distinction (arguably dates too).

[EDIT] Realistically most of the stuff under <project> is git repos and I usually make a “home” repo where I keep org files for tracking hours, notes, and resources related to the engagement.

work/client/project/2021-11-11-file.ext is great until you've got a '2021-11-11-project-status.txt' in a few directories and you need to find one quickly! I do a combination: clients/client/project/2021-11-11-client-project-update.txt

I just store it as a content hash and then when I want to find the file, I just have to recreate its content and I can then just get the hash.

It sounds like what everyone in this thread needs is a database file system. This was always my favorite proposed feature of Windows Longhorn that never made the cut. Almost 2 decades later and Microsoft's latest OS still doesn't have this feature.

I wrote about what I perceived as deficiencies of hierarchical file systems, and proposed an alternative organization based on tags and hashes. It was discussed on Hacker News last week and many years ago.

https://www.nayuki.io/page/designing-better-file-organizatio... ; https://news.ycombinator.com/item?id=29141800

Have you used BeOS?

For sure! I actually used Be before I ever used Linux.

I'll be the opposite voice: the file system isn't for precise organisation, it's just for storing. For organisation, the ideal thing to use is tags. Since most file systems don't have tags and using software for that would be a pain, the best way to do this is to list the tags in the file name.

I've always thought that personal files, photos, or any other kind of just needed more connections between them to improve my information retrieval experience. That's how I had become a Zettelkasten evangelist. I believed it would be the cure for the information overload disease of our era.

But life made me use Emacs org-mode more and more, and I'm now in love with tags. Retrieving information has become so easy, especially with org-mode's tags inheritance, that I hardly think making connections between headings or notes is necessary anymore[1]. And I believe that applying tags to filenames (a la Karl Voit [2]) will create the same effect

[1] A Zettelkasten-like system is still unbeatable imo when it comes to ideas repositories, i.e. a second brain you can talk to and get new insights. It's just not that great for personal knowledge management or project management.

[2] https://github.com/novoid/filetags

Maybe you mean `2021-11-11_client_project-name_v2_final.ext`

2021-11-11_client_project-name_v2_final_ridaj(1).ext

Copy (2) of 2021-11-11_client_project-name_v2_final_ridaj(1)__FINAL-v2.ext

> But underscore separates fields, hyphens for space replacement

But why not the other way, hyphen-minus for separating fields and underscore for space replacement? That seems to me more consistent with how underscores and dashes are used.

I fully agree, that's how I do it :)

my_project-some_activity-this_document-20210923-v02.txt

I've recently shifted sharply toward the dash from the underscore. I find it more readable, and it doesn't require the shift key. However, I do find it useful to use underscores to create groups, e.g. test-001_2021-10-11.log. Including hours, minutes, seconds is still awkward.

Burn the witch!

Brother in arms. I just posted similar thing below.

There's a customer for everything. I've just never liked the aesthetics of the underscore. Also if your underscored thing gets put in some document and then underlined the underscores can become invisible.

A lot of this is personal aesthetics, for sure. Personally, I am not a big fan of camel casing. In code, I only use it for class names, generally. I don't find it particularly readable, and for filenames, not all filesystems are case sensitive, so best not to rely on case to differentiate files. Camel case does have the nice property of being more compact, as no character is required. That's its main benefit.

R traditionally uses the . as a legal character in identifiers. Once you get it used to not being syntactic, I found I actually prefer them to underscores.

> kebab-case

I hadn't heard that before and I love it.

If you hadn't heard kebob-case called that before there's a chance you haven't heard SCREAMING_SNAKE_CASE called that before, and I couldn't live myself if I didn't let you know.

that's hilarious, thanks for sharing that.

Perennially relevant xkcd: https://xkcd.com/1053/

Awe, in turn I have never seen that particular xkcd—it's great! I learned to call it "feigning surprise" and I always try and be conscious of it (though I still catch myself doing it from time-to-time).

Google considers it too violent apparently. In one of their recent changes to their style guide, they started recommending "dash-case" instead.

https://developers.google.com/style/word-list#letter-k

This guide is stupid. They recommend not using “janky.”

Tbf, dash-case is more descriptive. Kebab doesn't mean skewer everywhere

Same. I had tears in my eyes from laughing. For some inexplicable reason it seems incredibly funny.

I'm of the opinion that kebab-case is the best case for all identifiers, because it's easy to read and to type. As always, Lispers were right all along.

Kebab case is the often overlooked benefit of prefix notation and semantic white space in programming languages. Honestly the best case of all cases imo.

One glorious day we'll accept programming languages that require spaces around infix arithmetic operators so that we can make kebab case a reality!

Lisps, especially Scheme with its `x->something-else` convention, have ruined naming in other languages for me.

Maybe Raku[1] is for you!

[1] https://raku.guide/#_syntax_overview (see section 1.7.1)

Forth does something like this, by virtue of its reverse Polish notation.

In Forth, 'words' (which are roughly analogous to functions and operators) must always be separated by whitespace, as Forth doesn't parse out operators the way most languages do. In exchange, you get the ability to use symbols in identifiers, as Forth has no reason to single out symbols like + as being syntactically special. You can even use a number for the first character. (For that matter, Forth will even let you override the usual interpretation of a numerical literal, but that's always struck me as going a bit far.)

It gives you a + word, analogous to the + operator of most languages [0]. It also gives you a 1+ word, as an (admittedly slight) abbreviation of the sequence 1 +. [1] If you wanted a 2+ word, you could easily define it yourself.

(This property of Forth evidently wasn't enough to get it to take over the world, but it's still neat.)

[0] https://www.complang.tuwien.ac.at/forth/ansforth-cvs/documen...

[1] https://www.complang.tuwien.ac.at/forth/ansforth-cvs/documen...

I found that some_document_2021-01-01_v03.pdf works best because it keeps the same document next to its other versions alphabetically, keeps them in date order, and keeps them in a sub-day version order.

In my work, today's date would be 21K11, to save space over the longer date.

How do you distinguish 21K111 and 21K111?

Are you trying to catch GP on differentiating hours, were it to be appended to his time format (1st @ 11 vs 11th @ 1am)?

Notably he didn't promise any, but presumably one'd need a separator... Maybe, per his "K" usage of the month, one'd use the alphabet again. 11am would be "K" again... or lowercase just for giggles?

I don't think it reads very well, but I also think one'd get used to it pretty quickly.

I was thinking January 11th vs November 1st. Maybe their "date" doesn't need/support day-of-month? Or they typod and I should just focus on my work.

I imagine January is A and November is K, so 21A11 vs. 21K1 (or maybe 21K01).

Ah yes, I missed that K was a month.

Are you working in some embedded system with tiny memory space or something? What's the use of saving one character? Just make it YYMMDD!

As a side note, in the good ol' times of ISO9660 level 1-4 and the various mkisofs parameters, an underscore _ which is a CAPITAL -, may have given issues, only for the record/as a curiosity:

https://web.archive.org/web/20151007005513/http://www.911cd....

P.S. should anyone want to see/run the actual batch, a copy has been uploaded here:

http://reboot.pro/index.php?showtopic=18962&page=29#entry204...

> 2021-01-01

Yes on the date format.

Saves you so much time.

Agreed on dates ordering problem but 20210101 is so much easier to type.

But much less easy to read!

Years that end in a 1 are awful when doing this, especially in October and November. We've had 20211001, 20211010, 20211101, 20211110, and now today 20211111.

I just tend to use $(date -Is) so I don't need to think what date it happens to be today. I guess -Id would work if you don't want the time part.

I don't bother with the century or the dashes, saves time...

211111_foobar_v1.txt

I am old enough that I still save before printing. I think it was Lotus 123 that engrained it for me.

This used to be my default, and then I used Matlab, and "-" was interpreted as subtraction.

I use this style:

2021-01-01_what-happened_who-did-it_possible-reason

Reply on news.ycombinator.com