Idea: User interface inspired by git

[michelk]

Following suggestions for the user-interface

• Switch program name from neuron to zk for faster typing
• Have a default Zettelkasten directory per project e.g. .zk/
• Use per default the Zettelkasten (.zk) directory, by searching in the descending directory tree
• Each project then could have its own zettelkasten and zk automatically finds it.

[srid]

I like the idea of obviating the -d .. argument when working on multiple Zettelkastens, but I'm not sure about this approach. What would the workflow be like for someone that uses neuron to manage a single zettelkasten, at ~/zettelkasten (i.e., no project directories)?

[maralorn]

What would the workflow be like for someone that uses neuron to manage a single zettelkasten, at ~/zettelkasten (i.e., no project directories)?

One idea in two different versions:

1. .zk can contain a link to a different root for the project. Then ~/.zk can somehow point to ~/zettelkasten.
2. Configure a default zettelkasten to use when not finding a .zk.

[srid]

Here's another point of confusion: zettelkasten notes are not necessarily associated with a (software?) project.

For example, here are two zettelkastens:

• https://github.com/srid/haskell.zettel.page
• https://github.com/srid/emacs.zettel.page

They are just git repos with note files. Where would the .zk directory fit in here?

---

As an aside, there is the optional neuron.dhall configuration file, which could perhaps be used to automatically designate a directory to be a zettelkasten directory. Though it is not clear how this would solve the goal of obviating -d.

[michelk]

I would still keep the (then optional) -dflag. So the current workflow doesn't change. Just when it is omitted, search for a conventional zettelkasten directory name (.zk or .neuron) in the descending directory tree, like git does.

I think neuron is extremely helpful in (collaborative) project work. And each of my software and client-projects could contain a zettelkasten-directory to initially document bits and aspects of the work done. This then easily compiles to a technical report or software-tutorial or whatever. Thanks @srid here for the great work.

Therefor, dealing with multiple zettelkästen, my .zshrc has multiple statements like e.g.

alias zk="neuron -d ~/zettelkasten"
alias zkhs="neuron -d ~/doc/haskell-zettelkasten"

But this gets tiresome, when project-zettelkästen are added. Because of that, it would be more comfortable to be able to do e.g.

cd projects/path/to/somewhere
do some work
zk new -e "Done this and that"

with zettelkasten-directory projects/.zk (or projects/.neuron if you like)

It would also be neat, if we had a ~/.config/neuron.dhall to specify e.g. the editor.

In my opinion we also should make the -e flag in new and search the default and use another flag to just emit the zettel-id-file.

[srid]

with zettelkasten-directory projects/.zk (or projects/.neuron if you like)

The proposal is not yet fully clear to me. There is already a .neuron directory, which holds the internal Shake database as well as the generated HTML files (which neuron open uses). Can you provide a detailed example, like the one below?

$ mkdir /tmp/zk1
$ neuron -d /tmp/zk1 new "First zettel"
/tmp/zk1/cbcebe33.md
$ neuron -d /tmp/zk1 rib
[Rib] Generating /tmp/zk1/ (rebuildAll=False)
# /tmp/zk1/cbcebe33.md
+ /tmp/zk1/.neuron/output/cbcebe33.html
+ /tmp/zk1/.neuron/output/z-index.html
+ /tmp/zk1/.neuron/output/search.html
+ /tmp/zk1/.neuron/output/index.html
$ tree /tmp/zk1/.neuron
/tmp/zk1/.neuron
└── output
    ├── cbcebe33.html
    ├── index.html
    ├── search.html
    └── z-index.html

1 directory, 4 files
$

[srid]

Take the neuron project as an example. Its zettelkasten is at the ./guide directory (with the associated .neuron directory at ./guide/.neuron). What would be the exact algorithm that neuron should use in order to automatically "detect" the default zettelkasten it should use?

[maralorn]

I understood it this way. When ¨neuron¨ is called (don‘t know if I care about renaming it too ¨zk¨), the following algorithm looks for the zettelkasten.

1. Use the directory given with -d, if given.
2. Check if the current directory contains a .neuron directory. If yes use this directory.
3. Go one directory up and go to 2. If / is reached continue.
4. Check if there is a default directory defined in a neuron.dhall config file somewhere. Use that if given.
5. Use ~/zettelkasten.

Take the neuron project as an example. Its zettelkasten is at the ./guide directory (with the associated .neuron directory at ./guide/.neuron). What would be the exact algorithm that neuron should use in order to automatically "detect" the default zettelkasten it should use?

That‘s I think something which does not really fit the git analogy.

But what you do think about using symlinks? You could have a .neuron -> guide/.neuron symlink in the repo and neuron understands the symlink and finds the Zettelkasten that way?

This could also be an additional or alternative way for point 4. in my List. Just have a .neuron symlink in your home-dir pointing to the .neuron of your Zettelkasten. (Has the disadvantage of not working outside of your home.)

Just an idea. Maybe we can do better.

We would also need something like neuron init for this to work nicely.

[srid]

Okay, step 1,2,3,5 make sense. Not step 4 - because neuron.dhall is a per-zettelkasten config that lies in the very zettelkasten directory we are trying to locate.

But what you do think about using symlinks? You could have a .neuron -> guide/.neuron symlink in the repo and neuron understands the symlink and finds the Zettelkasten that way?

Yes, I think that's a good idea.

We would also need something like neuron init for this to work nicely.

Right; neuron init should create the first zettel file, as well as

• run neuron rib on it, so that .neuron gets created (for the first time), or
• alternatively, it could generate the neuron.dhall file; in which case, the proposal receives the option of searching for neuron.dhall (file or symlink) instead of .neuron.

What's left in the proposal I think is to decide between .neuron and neuron.dhall as the thing to locate.

PR welcome!

[maralorn]

Okay sorry for mixing up the neuron.dhall, I actually haven‘t used neuron productively so I didn‘t know how it works.

So as far as I understand .neuron only contains cached data until now and therefore is kinda an implementation detail the user shouldn‘t care about?

So I think we should center around the neuron.dhall file? But if we wanna go full git mode we could make it .neuron/config.dhall. I kinda like that idea. But it‘s your call.

[maralorn]

Also maybe @michelk wants to voice an opinion here. I kinda hijacked their suggestion.^^

[michelk]

Well my original thought was to have something like below, where the .neuron directory contains all zettels and the generated site. And calling neuron say from /tmp/project1/app finds /tmp/project1/.neuron as zettel-directory.

/tmp/project1
├── app
├── .neuron
│   ├── 47c953f2.md
│   ├── 78377b7b.md
│   ├── 99d3a398.md
│   ├── index.md
│   ├── neuron.dhall
│   ├── .shake
│   │   ├── .shake.database
│   │   └── .shake.lock
│   └── _site
│       ├── 47c953f2.html
│       ├── 78377b7b.html
│       ├── 99d3a398.html
│       ├── index.html
│       ├── search.html
│       └── z-index.html
├── nix
├── src
└── test
/tmp/project2
├── app
├── .neuron
│   ├── 0a9e0b82.md
│   ├── 1b5549b1.md
│   ├── e7db5bf1.md
│   ├── index.md
│   ├── neuron.dhall
│   ├── .shake
│   │   ├── .shake.database
│   │   └── .shake.lock
│   └── _site
│       ├── 0a9e0b82.html
│       ├── 1b5549b1.html
│       ├── e7db5bf1.html
│       ├── index.html
│       ├── search.html
│       └── z-index.html
├── nix
├── src
└── test
/tmp/project3
├── app
├── .neuron
│   ├── 66274493.md
│   ├── a7dc6556.md
│   ├── c2c76b5e.md
│   ├── index.md
│   ├── neuron.dhall
│   ├── .shake
│   │   ├── .shake.database
│   │   └── .shake.lock
│   └── _site
│       ├── 66274493.html
│       ├── a7dc6556.html
│       ├── c2c76b5e.html
│       ├── index.html
│       ├── search.html
│       └── z-index.html
├── nix
├── src
└── test

[Nadrieril]

I don't like that the directory should start with a dot. I think what we want is more akin to a doc/ folder than .git.
I quite like the idea of searching for a neuron.dhall, this would be similar to many other systems that search for a particular config file, like package.json, Cargo.toml, Makefile, etc