（评论）

（评论）
(comments)

原始链接: https://news.ycombinator.com/item?id=41154616

为了总结您的文章，您描述了如何使用 TxTAI Python 库来处理和分析 fastAPI 日历应用程序中的时间关系。这涉及对时间注释和上下文细节进行训练和索引。您提到了一些示例，例如查找周三发生的具有特定着装要求的社交活动。此外，您还讨论了在移动设备上管理事件时面临的挑战，例如使用语音命令调整约会和删除多个警报。您过去的实践包括维护基于文本的日历和日记，其中每天分别包含相关事件和注释。随着时间的推移，您会强调在没有结构化文件夹和链接的情况下组织文件的复杂性。最终，您提出了以时间线格式实现动态层次结构的想法，以更好地组织数据。您分享了使用 Emacs/Org-Mode/Org-Roam 进行日常笔记和文件附件管理的个人经验，但也承认其缺陷。最后，您建议探索用于长期数据存储和检索的文件系统方法的可能性。

> Why Did You Do This?

> I used this project as an opportunity to learn about Rust and FUSE file systems. I also think it's hilarious.

> Visitors interested in the code should note that this is an irredeemably messy codebase—it's full of hacks, unidiomatic code, and wildly poor design decisions.

> However, visitors should also note that that's okay. The best way to learn something new is to try it out for yourself—and creating a mess is a vital part of that process.

I absolutely love this sentiment. Why do this? Because it’s fun. And messy. And sometimes, that’s okay. Not everything needs to be a product… sometimes it’s enough to do something because you can… or want to… or because $RANDOM reason. I’m very grateful to the author and whoever decided to submit it today.

As a college student trying to figure out where I wanted to go next (industry vs academia) I had an epiphany that there was intrinsic value in doing things. Even things that were neither novel (academia), profitable (industry), nor impactful (nonprofits).

While my career has found a balance between impact and financial sustainability, I have a special place in my heart for projects that are otherwise unnecessary but happen just because it brings the creator joy.

Even if it makes just you happy, it's impactful. You are the person others have to put up with. Making yourself happy improves the world for those around you.

>Even things that were neither novel (academia), profitable (industry), nor impactful (nonprofits).

You classifications are good but it's a bit simplistic. Generally across the spectrum people is looking for breakthrough (e.g. deep learning) and/or game changing (e.g. transformer) contributions regardless if you're in academia, industry, non-profit, government labs, etc. But for pure joy, it can be from any simple creative projects as you've mentioned or it can be giving up seats for the elderly in a crowded metro trains.

But there is no greater pleasure than to finish something, deliver it, get users in front of it, and get feedback you never expected from either grumpy or happy users. I think it's way more rewarding to try to work for profit, because while money is nice, profit also means it's so useful someone is ready to give you credit for future work (money) for what you did.

Not a universal experience. I receive little joy in money, beyond the basic needs it relieves. Shutting down a business that cleared $800M profit annul. was the best decision I ever made.

Profit doesn't inherently mean you've made the median experience of individual humans better, and is just as often to effect significant degradation. My biz was slightly on the positive, but I could have pegged an extra zero in profit if I had forgone scruples.

Aw, that's not what I hoped it would be. An actual file-based calendar UI is a neat idea, where you can (for instance) echo into a file to create a calendar event on the command line.

Not sure I hoped, but from the title I expected.

I don't have more than 10 entries per week. Partly because I prefer real work over meetings, so I work at a place that does not have a lot of meetings. And once you don't have many, it's easier to remember other appointments, no need to use a calendar for everything. So even if this had been a useful mapping of calendar entries into a directory tree, I would not have felt an urge to use it.

This is completely me talking out my but:

Could one use the txtai python libe to be told to train on the temporal relationships between events that pass through it as an fastAPI endpoint to a calender... then as the events head to whatever is holding them in the calendar - txtai is training and indexing, first on temporal notes, then on context. so you ould easly as it to give you a

"Show all the social events that happen on wednesdays with a dress code"

(but it learns the nature of certain events, when they occur and social clues around what /who /wheres are typically occuring...

"This venue typically holds events on W F S and the clientelle is typicalaly this, attendance that, rsvps x, cost ~$$"

Good tab expansion in a very virtual filesystem would work wonders:

    $ cat /cal/nexttues
    $ cat /cal/2024/08/13/

    $ cat /cal/nextweek
    $ cat /cal/2024/w32/

I seem to recall that I had lunch with somebody last July:

    $ grep -ir somebody /cal/2023/07/
    /cal/2023/07/12/events:13:00-14:00 Lunch with somebody

the most amazing feature of the Google calendar web UI that has since been dropped was a way to schedule events with a human message - "30 meeting with Bob next Tuesday at 10", rather than clicking a bunch of buttons in a UI.

I remember that! I think they have just collapsed to handling what most people try, evidently '[time] description'.

A couple of weeks ago I was surprised and annoyed that adding an event with description "10:30-11:00 Pointless Meeting" no longer scheduled my Pointless Meeting. Tried again very carefully. Wasn't working. Worked again the next day. Did anyone else experience this?

I can't wait for them to feed my description to an LLM to generate the event. "Schedule drinks at 7 at the bar behind the office next Tuesday. Invite my favourite colleagues and their teammates. Make them curious but not suspicious."

Well in a CLI you're not writing full English sentences. You don't say "make a directory called xyz in the current directory," you're saying "mkdir xyz". I'd rather just use a GUI than type out a full natural-language sentence.

With Google Assistant you can still do this, but you need to speak it, with the prefix "Add calendar event..."(well, Google Assistant on the phone also accepts keyboard input). I find it much faster than the button clicking.

Similarly in iOS there's some surprising: "Move my 3pm meeting from today to tomorrow 10am", and a clutch tip: "Hey Siri, Delete All My Alarms", b/c there's no UI to delete and confirm the 100 alarms that accrue of "7:00am, 7:05am, 7:07am, 7:15am, ...etc..." ;-)

Not this exactly, but I used to keep a calendar as a text file, where each day was a single line containing the date, week day and possibly some events.

Something like:

    4h sun: 1500 dentist's appointment, 1800 dinner with friends

Where a=jan, b=feb, c=marc, ... l=dec.

The initial file was autogenerated with a 10-line Python script.

You can have variations on this, like allowing multiple rows per day, one row per event.

I also used to keep an unstructured "journal.txt" file, where notes were separated by two blank lines, and you could reference topics (like #toRead) and dates (like #4h24) with #, like social media hashtags.

I was thinking about this and came to the conclusion that I'd want to define some events per date, like you illustrated (`"12:30 dentist" > /cal/2024/08/04`), but others I'd like to define per event, like `2024-08-23 12:30 > /cal/massage`, which seemed messy. I didn't really resolve this question for myself.

I don't see where it's remotely useful, the idea sounds fun but it's not praticle.

Explain how it is better vs having a simple cli that parse args and store that into SQLite.

discoverability - but then proc is not the best example for that - a lot of proc seems to be inpenetrable without documentation

access - for discovery and poking around you won't need a dedicated tool, just `cat` or `less` or `vim`

I appreciate that they made the calender system pluggable. Don't want to get locked into a single calendar vendor with something as important as your file system!

Yeah, I’ll update the README to be more specific about this. This is also why I didn’t provide instructions on how to use the tool :)

It stores file data as calendar events, sort of like that YouTube FS from a while back

This is probably considered abuse under the ToS since you're using more storage than any typical user, and they can't use it for whatever they normally sell customer data for

I seriously doubt this kind of stuff has ever messed with their bottom line. I think they just really resent users trying to do more with their service than they intended.

This is unhinged and I love every part of it! Apart from the repo having an attitude I often miss from the modern internet I also learnt something reading the code on how FUSE works!

In exploration terms I suggest considering typical serious desktop users file taxonomy, most after at certain amount of time choose a timeline taxonomy of some kind simply there is no much easy way with files and folder to collect and access data, even with {sym,hard}links.

Essentially time passing is a common thing for anything, all of the rest quickly became dirty or at least have some dirty dump of information here and there. Long story short it's a nice potential continuation of the experiment start taking notes in timeline, storing files in timeline and figure out how to make "dynamic hierarchies" that works well in this model.

I've done something myself with Emacs/org-mode/org-roam notes, with kind-of daily notes and file attachments, it's not perfect nor general but scale a bit. A filesystem approach it's uncharted but seems to be equally possible.

（评论） (comments)

（评论）
(comments)