Tinymist

[oak quest]

The code currently is very ugly and I will clean code when I implement the block coverage correctly.

[marsh lance]

I think memoization could affect coverage

[buoyant sand]

would coverage colleciton not be possible form typst-eval directly without adding anything to the AST?

since that was the problem last time

[marsh lance]

it would, but it would require more design and bikeshedding

[oak quest]

I thought a typst debugger was also affected by memorization, as well as the layout iteration. But I found we could keep stupid. The control flow with memoization and the layout iteration and so on is exactly what we should understand.

[marsh lance]

I think iterations should be visible but memoization is something a user shouldn't have to be concerned with generally

[oak quest]

Or you mean some coverage will be faked by the comemo.

[marsh lance]

yeah, coverage would be TrackedMut

like the Sink

or it would be in the sink

not sure how much overhead that would incur. could be a lot.

[oak quest]

will it be a problem if we only collect coverage about cold compilation.

[marsh lance]

it's a reasonable approach for tinymist I think

for a compiler API it would to handle it

[floral adder]

[[language]]
language-servers = ["tinymist"]
name = "typst"

[language-server.tinymist]
command = "tinymist"

[language-server.tinymist.config]
formatterMode = "typstyle"

[language-server.tinymist.config.preview.background]
args = ["--invert-colors=never"]
enabled = true

I'm trying to get the live preview to work in helix with tinymist, what am I doing wrong?

[somber drum]

You might have to specify the address where to host the preview? You might also be on an older version (background preview is not released yet)

[oak quest]

Maybe we should maintain documentation for each release

pdf is great

[somber drum]

How long between the 13.6 rc and the 13.6 stable release? Sadly the zed extension ignores pre-releases :/

[oak quest]

No explicit due, but when we feel confident enough. I'm waiting some feedback from some rc users. Till now, two people said they didn't find regressions in the rc.

[oak quest]

if i could run and report it in ci

[oak quest]

I find some not pretty case, that the unreachable branches are reported as uncovered. There is still a long way to go..

[cursive bough]

What are toy showing us 😃?

[buoyant sand]

Experimental coverage info

(which code is actually run and how often)

This helps increase test coverage

[cursive bough]

Oh.. I think I get it. That's nice.

[cerulean tendon]

hm im trying to use helix for typst, is there a way to make it automatically open the pdf preview?

[somber drum]

https://github.com/Myriad-Dreamin/tinymist/releases

Have a look at the latest release notes

[somber drum]

I think the background preview is what you mean

[cerulean tendon]

im not sure sure

it doesn't really open it

this is my helix languages.toml

[[language]]
name = "typst"
language-servers = ["tinymist"]

[language-server.tinymist]
command = "tinymist"
config = { exportPdf = "onType", preview.background.enabled= true, startDefaultPreview=true}

[somber drum]

Unless you explicitly installed the pre release that feature isn't available yet

[cerulean tendon]

ah i see

what if i changed command?

would something like tinymist && mupdf

[somber drum]

Nah

Do you just want the pdf?

[cerulean tendon]

i just want the pdf viewer to launch at the same time

[somber drum]

There might be some helix way of doing this

Never used it so can't tell

[cerulean tendon]

hm but not really in tinymist?

or if it opened the preview automatically that would work as well

i don't really use the click features, but they are nice to have

[somber drum]

I'm not quite sure what exactly you mean, but you might find a solution here https://github.com/Myriad-Dreamin/tinymist/blob/main/editors/neovim/Configuration.md

You can have the pdfs generated automatically

But tinymist can't open them automatically, although the preview might be able to?

[oak quest]

by default export is independent from open, I hope helix could have a way to combine them. tinymist's export does have an option to open after write, but helix doesn't support to enable that option.

[somber drum]

I am trying to use the new update in Zed but I can't seem to get the background preview to work

"tinymist": {
      "initialization_options": {
        "exportPdf": "onType",
        "formatterMode": "typstyle",
        "outputPath": "$root/output/$dir/$name",
        "preview": {
          "background": {
            "enabled": true
          }
        }
      }
    }

This is the tinymist lsp settings, did I do something wrong here?

[oak quest]

what did the log say?

[oak quest]

I translated all titles of commands in vscode using LLM.

https://github.com/Myriad-Dreamin/tinymist/pull/1501

Looks great but I don't really know how actually good it is.

[cursive bough]

da = "Eksporter den åbnede fil som PDF"
Hmm.......
maybe da = "Eksporter den nuværende fil som PDF".

DeepL says:

I'd try to use DeepL for the translation if you can. In Danish, both of these suggestions are more accurate than what's in the PR. But this is just for Danish.. I don't know anything about the rest.

[oak quest]

I use LLM because it is pretty quick. I was simply telling LLM to generate them and I wrote 0 lines of code. Might correct them later.

perhaps I will only translate the code that I know well, and left rest of languages for people who use them.

[vital kayak]

I'm using tiny mist to create layouts, these layouts import JSON files from sys.inputs and build a pdf, I would like to provide a JSON schema that tiny mist can then use for type validation.
Is there anything similar on the roadmap

[oak quest]

Couldn't you enforce them by installing JSON scheme support in your editor? To validate them when building PDF, you could make some typst packages or plugins by yourself.

[vital kayak]

The pdf is built at runtime, actually it's an app that creates pdfs for invoice data. Each invoice has different data of course but the shape of the data remains the same.
I could provide an example JSON file but that won't allow me to cover the whole range of possibilities the data can have

I want to validate that the Typst code is properly accessing this JSON data based on the schema (doesn't have to be JSON schema, can be any typed schema)

[oak quest]

I meant you could validate the input by making some code, like https://typst.app/universe/package/valkyrie/. To check that your code correctly implements the validation, you could write some tests.

[vital kayak]

Does tiny mist infer the schema by Valkyrie and does type checking

Then I could define the expected schema inline and further code could be typechecked

[cursive bough]

No worries! Just wanted to provide the little feedback I have.. 🤷‍♂️ Translations are always a process to get right.

[oak quest]

I can but I haven't. I would rather make type checking when typst introduce official syntax and semantics of type annotations. Currently it is painful to make type checking work without any annotations.

[vital kayak]

Ok, makes sense yes. Right now I do see tiny mist inferring types if I provide a sample JSON file, so that's already great for auto completing fields

[hoary stag]

@oak quest many thanks for the ongoing work on supporting simple editors! I'm trying the background preview feature with helix and so far it's awesome!

For other helix users who want to try it: you can use something like this:

[language-server.tinymist]
command = "tinymist"
config.preview.background.enabled = true
config.preview.background.args = ["--data-plane-host=127.0.0.1:23635", "--invert-colors=never", "--open"]

in your languages.toml and it wil automatically open a browser tab with instant preview when you open a typst file

while startDefaultPreview is an LSP command you can emit manually from the editor, by running :lsp-workspace-command (this wasn't immediately clear to me)

[harsh wing]

Since now on NixOS MuPDF doesn't update instantly for some reason, I think more and more about creating a Tauri app that will be minimal, potentially statically linked and have the minimal UI to maximize the free screen real estate. And it should consume much less resources compared to a full-fledged browser (or electron).

Since browsers are bulky and it isn't easy or it's impossible to make it only show the content without any UI (fullscreen-like), and the next browser session will open it in with the same geometry (half-screen), I don't wanna use any browser-based previews.

[hoary stag]

Another solution would be to use a barebones browser for the preview. Something like qutebrowser, you can configure it to hide the tab bar and status bar and then it's just the content without any UI... And it should be easy to configure the window manager to always open those windows with the geometry that you like

But I don't want to talk you out of making a cool Tauri app that does the same thing with fewer resources and in a way that's cleaner and easier to package 🙂

[harsh wing]

I mean, Tauri is kinda a bare-bones browser, but you can make it a better UX by providing a good CLI. Though without much thinking I don't know if there is actually something that needs to be implemented, since it already uses WebKitGTK (I think), which is 99% of work done.

[cerulean tendon]

hm idk why but it doesn't open a preview at all for me

[[language]]
name = "typst"
language-servers = ["tinymist"]

[language-server.tinymist]
command = "tinymist"

config.preview.background.enabled = true
config.preview.background.args = ["--data-plane-host=127.0.0.1:23635", "--invert-colors=never", "--open"]
config.exportPdf = "onType"
config.formatterMode="typstfmt"
config.preview.cursorIndicator=true
config.review.scrollSync=true

[somber drum]

Also doesn't work for me in Zed :/

Update: works on latest Zed version

[modern falcon]

thanks for suggesting qutebrowser! i absolutely love it

[oak quest]

what's the version of tinymist you are using? You could also send logs of tinymist to give more information if you find them.

[oak quest]

does it work after upgrading tinymist?

[oak quest]

Nice, I didn't think of --open flag working great. You did invent the good trick for helix.

but you'll get a new preview tab every time you open the helix editor. I was thinking that you could keep using a single tab listening the port on 23635 where the background preview will be on.

if you love the open flag, I guess you could open multiple background preview with telling to use a random port:

[language-server.tinymist]
command = "tinymist"
config.preview.background.enabled = true
config.preview.background.args = ["--data-plane-host=127.0.0.1:0", "--invert-colors=never", "--open"]

[hoary stag]

ah nice, yes I think both of your versions are good options to have. I'll try them and see what works best for me

[harsh wing]

Can tinymist work with TYPST_ROOT? Right now it's my only problem, when I have a root that is not where the main file is.

[somber drum]

It does somewhat work now, not particularly reliable but most of the time. It doesn't work at all with the --open flag though. Document switching also only works some of the time

It is pretty slow with Safari but works better in firefox, maybe chrome would be even better? (Don't really use chrome at all tho)

[somber drum]

@oak quest this seems to add window/showDocument to Zed, so implementing it would benefit pretty much all editors https://github.com/zed-industries/zed/pull/25007/

[fathom glade]

should this not be in a rustfmt.toml?

[oak quest]

vscode ignores rustfmt.toml

[fathom glade]

._. why

damn looks like the lsp format protocol doesn't always give a path maybe

[coarse hearth]

Hello, I'm testing new text editors with Tinymist.

I've successfully tested neovim and helix but I'm having trouble getting zed to work.

Can you help me?
I can only get the LSP to work with the marketplace extension, not manually. As a result, I can't configure the LSP options. Additionally, how can I access the preview window?
Thank you

[coarse hearth]

I tried

"lsp": {
    "tinymist": {
      "initialization_options": {
        "exportPdf": "onType",
        "outputPath": "$root/$name",
        "preview": {
          "background": {
            "enable": "true"
          }
        }
      }
    }
  }

I have compliation on type, I don't have the preview feature.

[coarse hearth]

Indeed, I encounter some problems on helix.

LSP lauch, exportPDF on type works, but inverse search from web view doesn't work. and formatter doesnt work...

Here is my config

[[language]]
name = "typst"
language-servers = ["tinymist"]

[language-server.tinymist]
command = "tinymist"
config.preview.background.enabled = true
config.preview.background.args = ["--data-plane-host=127.0.0.1:23635", "--invert-colors=never", "--open"]
config.exportPdf = "onType"
config.formatterMode="typstyle"
config.preview.cursorIndicator=true
config.review.scrollSync=true

[somber drum]

I have pretty much the same config. You need to open the website manually

Make sure to only have those settings, it doesn't work at all when I override the args

It does work most of the times for me but it is still a little inconsistent

I am on MacOS and installed Zed via nixpkgs and used home manager for configuration

If you are on Linux (or Windows?) there might be some platform inconsistencies

[coarse hearth]

It doesn't work at all.
I'm on mac too.
The lsp is launched, but i cant compile...

[somber drum]

You mean the pdf compilation doesn't work?

Also make sure to use the latest Zed version, it only started to work for me with 0.177.7

[coarse hearth]

yes the pdf compilation doesn't work. I have the latest version I think.
I installed zed with homebrew

Did you install market expansion for typst ?

[harsh wing]

https://github.com/Myriad-Dreamin/tinymist/actions/runs/13886754840/workflow

[somber drum]

What do you mean?

[harsh wing]

I was about to say that GitHub's notification become ridiculous, but apparently it's due to a config error. I guess they didn't know that you better add this to the reasoning in the email.

[coarse hearth]

from my helix config

can't make the formatter work

[oak quest]

iirc helix cannot use the formatter in a language server so you should install and setup typstyle separately. typstyle is the formatter used by tinymist.

[oak quest]

the jump functionality can be only used by vscode, neovim, and emacs. the support in other editors is in todo list, there is an opened issue describing the plan.

[oak quest]

the github action told me to remove a quote from the condition. I doubted it but I still followed his advice. Then, my ci got broken😇. it also lied me a great checked mark since github will ignore workflow file containing grammar errors..

[harsh wing]

When I was trying to get the MRE for the latest issue, I found that I get all the relevant items on top if I remove the closing parenthesis before triggering the autocompletion.

I still think that it has something to do with tinymist.

[oak quest]

it is very challenging to make completion on the file containing syntax errors. not all cases are handled, and it can be an improvement.

[harsh wing]

I'm talking about completing #set text(|).

[oak quest]

but you missed a right paren.

[harsh wing]

Because this way it shows fields on the top.

#1260973804637786224 message

[oak quest]

okay I misread it. you mean the problem is that all completions are shown, though I'm not getting how I can improve it.

[harsh wing]

The server gives the order of the items?

[oak quest]

I see a content completion which is a positional argument, which I should remove in future.

[oak quest]

some neovim plugins will re order items. if they don't, tinymist does gives the order.

[harsh wing]

I tried with only nvim-cmp.

[oak quest]

i think I get it a bit. could you capture and tell the nvim requests. This is to find the difference between adding a lparen and removing a rparen.

[harsh wing]

you mean the mirror.json?

[oak quest]

by your skill. It is fine by capturing it using --mirror.

[harsh wing]

by your skill.
What?

I did imply using --mirror mirror.json.

[oak quest]

{"method":"textDocument\/completion","params":{"position":{"character":7,"line":0},"context":{"triggerKind":2,"triggerCharacter":"("},"textDocument":{"uri":"file:\/\/\/tmp\/a.typ"}},"id":21,"jsonrpc":"2.0"}
{"method":"textDocument\/completion","params":{"position":{"character":7,"line":0},"context":{"triggerKind":1},"textDocument":{"uri":"file:\/\/\/tmp\/a.typ"}},"id":24,"jsonrpc":"2.0"}

This is the difference:

- "context":{"triggerKind":2,"triggerCharacter":"("}
+ "context":{"triggerKind":1}

When adding a lparen, the completion is triggered by "triggerKind": "triggerChar" and triggerCharacter: "(". When removing a rparen, it is triggered by "triggerKind": "invoke" and triggerCharacter: null.

[harsh wing]

@oak quest, do you know if a server can send a warning message to a client? In Neovim you can send different levels of messages to display it.

Oh, wait, I think the warning from logs would probably be displayed, yeah.

[oak quest]

do you mean explicit messages instead of log? https://microsoft.github.io/language-server-protocol/specifications/specification-3-14/#window_showMessage

[oak quest]

bad news: window/showDocument will replace the preview panel to the document to show, instead of show it beside the preview panel...

good news: when you open preview in browser, this doesn't bother anymore.

[coarse hearth]

Thank you. But I can't make typstyle work in Helix. It works perfectly in terminal but not in helix.
Do someone know how to configure typstyle in Helix ?

[oak quest]

#1261622502598377545

[harsh wing]

That's not the lsp command.

But the lsp command doesn't have that root dir argument, which is very confusing. It uses LspArgs. I did find that there is some weird connection between them, but it was soo hard to undrestand what is beign called that I gave up. The problem is that in the printed config it shows the same root in multiple places, which means I either have to fix all of them, or find some magical place from where they are all being copied and fix it once.

[oak quest]

If you are not familiar with programming or background, you usually only check the related code instead of starting with the main function.

• rootUri and workspaceFolders on initialization: https://github.com/microsoft/language-server-protocol/issues/298
• initialization request spec: https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#initialize
• Handling LSP initialization request: https://github.com/Myriad-Dreamin/tinymist/blob/157703ef671806039090a201894021589ea58393/crates/tinymist/src/lsp/init.rs#L35-L51

[harsh wing]

So you're saying that the main function for LSP command is not related code?

[oak quest]

I don't want to talk more about such not important things. You don't have to read code of the main function if you only care about the root configuration.

[harsh wing]

If I'm not familiar with a code base, that is the only sensible way to get to the bottom of this (with LSP jumping and/or debugger), at least if (in this case) the root arg is used in a bunch of different places and it's not clear where it's being grabbed into the init config for the server. This is partially how I recently figured out how a few chunks of Typst logic work.

I think you are talking with a bias of knowing exactly how your project is constructed and where all things live. I have no idea where the root configuration is taking place nor do I know how (exactly) everything else work in tinymist.

But the link to the related lines of code looks helpful, so thanks for that.

[oak quest]

sorry I usually become less patient when I'm on my real life's work time. I personally never read code from main function but only do regex or type search on code base, so I felt it is not important to discuss what the main function actually does.

[harsh wing]

maybe you're just a 10x engineer?

[modern falcon]

Is that really a question

[harsh wing]

suggestion? a guess? a hunch?

[somber drum]

Does the implementation of window/showDocument aim to make the switching between documents better?

[oak quest]

surely, but it is a standard lsp request, so we can only improve it to fit for all language servers.

[somber drum]

That's pretty cool, wouldn't that sort of make background preview redundant or in the very least not necessary anymore for helix/zed?

Also, is there some way to lock the preview to one file in those editors? It's quite annoying when developing a template

[harsh wing]

I use this:

    PinMain = {
      function()
        vim.lsp.buf.execute_command {
          command = "tinymist.pinMain",
          arguments = { vim.api.nvim_buf_get_name(0) },
        }
      end,
      description = "Pin main file for better autocompletion/error messages",
    },

Though not always.

[oak quest]

default background preview is suitable for single-file documents whose imports are all package imports. for multiple-file documents, locking isn't an ultimate solution. you perhaps need to combine this lock feature. https://github.com/Myriad-Dreamin/tinymist/issues/1527#issuecomment-2729215164

Using the lock database will produce a lock file in the root, so I don't enable it by default.

[oak quest]

sleepy is not using neovim.

[harsh wing]

I showed the LSP command

[oak quest]

i understood your example but the lsp command is not usable in current helix or zed which is the problem of not being using neovim.

[harsh wing]

not usable?

because those editors don't implement some client side thing?

I thought it would work anywhere.

[oak quest]

helix doesn't allow command with arguments, and zed doesn't support to call lsp command

[somber drum]

Hey, that looks good! When will it release? This command doesn't seem to exist in 0.13.8 and there is nothing in the 0.13.10 changelog about it

That lock file is also respected by the preview, right? Or is some extra argument needed?

[oak quest]

some refined docs should be here, but I haven't cook one for this experimental feature. you could try the following steps:

1. turn the configuration resolution into lockDatabase and restart editor
2. tinymist compile main.typ and restart editor
3. change the file to template.typ and see if it is pinned to main.typ

[oak quest]

in theory it is :). let me know if it doesn't work.

[somber drum]

It does not work

The tinymist CLI doesn't use the options I set in Zed, right?

I have only set this in the project settings

  "lsp": {
    "tinymist": {
      "initialization_options": {
        "projectResolution": "lockDatabase"
      }
    }
  }

(Alongside my global settings:

    "tinymist": {
      "initialization_options": {
        "exportPdf": "onType",
        "formatterMode": "typstyle",
        "outputPath": "$root/output/$dir/$name",
        "preview": {
          "background": {
            "enabled": true
          }
        }
      }
    }

One thing I had before was some longer update delay when using preview in zed + safari, that seems to be completely fixed now

tinymist.lock is supposed to be a file in the project, right? Because there doesn't seem to be one

Will try it in helix

[somber drum]

I also doesn't work in helix, there doesn't seem to be any lock file

[oak quest]

try this option perhaps --save-lock

I'm in bed so I cannot test it soon for you.

[oak quest]

how did it get fixed? I usually know nothing about safari or webkit.

[somber drum]

Will try asap

[somber drum]

Don't know, could also be some change in Zed or the extensions I use. Good the problem's no longer there, finding out the cause would habe been a huge undertaking 😅

Using --save-lock does indeed generate this lock file:

# This file is automatically @generated by tinymist.
# It is not intended for manual editing.
version = "0.1.0-beta0"

[[document]]
id = "file:elliptic-curves/main.typ"
inputs = []
main = "file:elliptic-curves/main.typ"
root = "file:."
system-fonts = true

[[task]]
document = "file:elliptic-curves/main.typ"
id = "file:elliptic-curves/main.typ"
type = "export-pdf"
when = "never"

But preview doesn't seem to use this, it still switches between files

Works in neither Zed nor Helix

Also can't find any preview options to force usage of the lock file or specify its path or something

One more thing to note: setting tinymist.background.preview.args completely breaks the preview in Zed, even with the default setting. Would be cool to set the --open flag.

[oak quest]

We don't need an option. Once we switch to the solution using the lock file, all the things in the tinymist should use the lock file for consistency.

@somber drum could you open some issues on the https://github.com/Myriad-Dreamin/tinymist/issues ? So that I can track them and won't forget them.

[celest lily]

Has anyone noticed freezes (the editor becoming unresponsive) when entering a new line ? Hitting Enter produces nothing for several seconds and then it suddently works. If you spam Enter out of impatience, the newlines eventually appear about a minute later, slowly (one every few seconds or so). It sometimes goes away for hours, sometimes comes back. When it happens, it makes Tinymist utterly unusable. It never happens with Tinymist disabled. I will test removing other extensions by bisection this afternoon, as this sounds like a me problem, but has it ever happened to anyone else ?

[wheat garnet]

What editor is this happening in? And just as a sanity check: you don't have 100% CPU or RAM usage at that time, right? Exorbitant delays sounds like lots of swapping, but if the rest of the system runs smoothly that doesn't sound likely.

[celest lily]

VS code, and yeah the system is not particularly stressed, but it's on a corporate laptop and I never trust these machines and their shitty antiviruses

[hoary stag]

@oak quest thanks for another great release, I love the new showDocument feature

it works great in helix

Do you still have plans for a solution for jumping in the other direction? (I see that https://github.com/Myriad-Dreamin/tinymist/pull/1236 has been closed)

Maybe tinymist could react to changes in the document, and move the preview to always have new changes visible? This is more or less what https://github.com/tmistele/typst-lsp was doing, and it worked well for me: I just start typing somewhere in my source file, and the preview automatically jumps to show the region of the document where I am making changes

[oak quest]

It's not closed but in draft.

[oak quest]

Sounds like a much better solution, although It also has the problem that the language server cannot distinguish the editor request to changes, like a rename request will sometimes hit to change all source files of your document, from that made by real typing the keyboard.

I may consider the latter one rather than providing a button as the latter one is easier

[hoary stag]

maybe this feature can ignore any change that hits several files (or several locations in the same file)?

[oak quest]

I'll try as long as it can, editor might also split changes into parts tho. I'm not really like we have to depend on inferring user behaviors from the LSP requests or notifications, which is hacking the LSP protocol.

[hoary stag]

yeah I wish helix had better LSP support

Unrelated question: what's the difference between tinymist-linux-x64 and tinymist-x86_64-unknown-linux-gnu?

both seem to link the gcc libraries

[oak quest]

v0.12.14 - [2024-12-25]

We now use cargo-dist to build, attest, and release binaries. The package publishers should prefer to use these binaries and validate them along with checksum. If you have any problems with changes, please report them to GitHub Issues.

These following suffixes are used for VS Code (or VSCodium) extensions: win32-x64, win32-arm64, linux-x64, linux-arm64, linux-armhf, darwin-x64, darwin-arm64, alpine-x64, alpine-arm64, and web. The rest are used for other editors (e.g. Neovim).

We also start to release VS Code extension targeting universal platform. The universal target only doesn't bundle the binary. This is suitable for other platforms like RISC-V (riscv64) or LoongArch (loong64). Users of that must install tinymist by themselves. Note it introduces risk to use unaligned version of the extension and the binary, but we can mitigate it in future.

https://github.com/Myriad-Dreamin/tinymist/blob/main/CHANGELOG/CHANGELOG-2024.md

No intended difference, the linux-x64 one is what we exact pre-build for vscode.

And since v0.13.12, tinymist-linux-x64 is aliased (a copy) from tinymist-x86_64-unknown-linux-gnu.

[hoary stag]

they're not quite the same though:

~/Downloads $ du -b tinymist-x86_64-unknown-linux-gnu/tinymist tinymist-linux-x64
58406600        tinymist-x86_64-unknown-linux-gnu/tinymist
57183184        tinymist-linux-x64

[oak quest]

v0.13.12 is not released

[hoary stag]

ah sorry I missed that, all clear now!

[somber drum]

@oak quest opened an issue about the lock file/preview

[jaunty kiln]

does tinymist support live share in vscode?

[oak quest]

no. it requires the language server to access the virtual file system but we only support to access the physical one.

[jaunty kiln]

I'm just asking because latex workshop added this https://github.com/James-Yu/LaTeX-Workshop/pull/4308. But I'm sure they have different challenges

[oak quest]

to support vfs, we will have similar code changes.

around 1k code changes.

looks like they have problem on the synctex, but we will not if we use the integrated typst preview

[jaunty kiln]

I'm not particularly interested in live share, so no need to spend development time on it for my sake 🙂

was just curious

[rugged copper]

Is there any way to have tinymist also suggest local packages in custom namespaces?

[harsh wing]

Custom namespaces are not supported. Though I've never tried.

[rugged copper]

not supported by tinymist you mean?

it does display every information and also suggests imports from packages in a custom namespace, it just doesnt suggest packages when writing smf like "#import "@custom/"

[harsh wing]

wow, it does work... I guess this is why you can override preview namespace locally.

[rugged copper]

yep, everything besides the import suggestions work flawelessly (using tinymist in zed, just as a context)

[harsh wing]

yeah, then it is a feature request, I guess

[oak quest]

It is not encouraged so we even didn't support it.

The @local package completion was moved from typst-sync package. I didn't check it but it should be easy to be adopted for completing packages in @custom namespace.

[rugged copper]

that's sad to hear as i think this is an awesome "feature" as it allows me to separate templates and packages i use personally (@personal) from templates and packages i use for eg. for work (@work) etc.

[oak quest]

I guess in future the namespaces will be exact for some organizations or personal names. I don't know whether you have to finally rename your packages from @work/awesome to @your-name/awesome.

It is welcomed to make a feature request to complete all namespaces tho. Nothing is preventing it, but there is simply lack of a person cooking the feature.

[rugged copper]

ah, okay, got it, I'll try to make one when I have a bit more time :)

[coarse hearth]

Hello, I'm having some issues with reverse search in Neovim.

It lacks precision, sometimes there can be a discrepancy of several lines.
Additionally, in certain places, clicking for reverse search does nothing. I think this is due to the lack of precision in the reverse search.

In VSCode, there are no issues.
Surprisingly, there are also no issues in Helix. The reverse search is character-precise.

lspconfig.tinymist.setup {
  single_file_support = true,
  root_dir = function()
    return vim.fn.getcwd()
  end,
  on_attach = nvlsp.on_attach,
  on_init = nvlsp.on_init,
  capabilities = nvlsp.capabilities,
  settings = {
    exportPdf = "onType",
    outputPath = "$root/$dir/$name",
    formatterMode = "typstyle",
  },
}

{
    "chomosuke/typst-preview.nvim",
    lazy = false,
    config = function()
      require("typst-preview").setup {
        open_cmd = "open -a Firefox %s",
      }
    end,
  },

I'm on Mac and tinymist is installed in its latest version via Homebrew.

[harsh wing]

There is now vim.fs.abspath(), which I think means that I no longer need plenary to set the root dir.

[coarse hearth]

Ok. Solve when using Tinymist from Mason, not from homebrew...

[urban grove]

🥲

[lilac galleon]

Hi guys, I just try to switch from web app to tinymist + vscode, somehow auto complettion does work within math mode, any suggestion?

[toxic minnow]

does or doesn't?

[jaunty kiln]

Doesn't. But they said in another channel that they resolved it. It was the quarto extension interfering with Tinymist

[oak quest]

How do you view a "stepless" symbol completion? Would you like to make it opt-in or opt-out? Currently, this will be opt-in.

https://github.com/Myriad-Dreamin/tinymist/pull/1313

[harsh wing]

Depends on how accurate it is. Also how accurate the suggestions if you want to search through dot notation. I would think that they can be worse. But if it's opt-in, then it definitely would be nice for the longer ones.

[oak quest]

Sounds like some order should be determined for better experience.

While it is stepless, it prioritizes shorter steps.

[buoyant sand]

That's usually called fuzzy search btw

[oak quest]

it depends on your editor. I only provide a list of expanded symbols.

[oak quest]

The first try to make a linter using static analysis.

[oak quest]

will it annoy somebody who doesn't care such "bugs"? All the following f function always return none and don't affect outer context.

[vast rock]

It could be difficult to make a distinction between the lints you want the user to see

But the discarded returns are logic errors so that seems like it's relevant

cargo check would show that

cargo clippy is more about idiomatic paradigms

[oak quest]

I consider making some linter configuration before enabling it by default.

[native plover]

While exporting to HTML, this is thrown.

Any help?

[oak quest]

you could find and debug errors in the html export by changing the exportTarget to html.

the specific error in the screenshot is very much related to some set page is applied when export ing html.

[native plover]

There is this Touying exporter by @iron dust , which works well and exports the same in html.

I did some config-pages though

Like changing margins and stuff

I am total noob, lol

[oak quest]

that is not the same way. touying exporter embeds svg images into html files while exporting html means to use the experimental html feature.

[native plover]

Oh, I'll see what can be done about the error and if it bothers me much, I'll ask again

I was fine with the pdf but this presentations have some gifs into it, which pdf doesn't support so have to use html

[oak quest]

why not using the touying exporter when finishing the writing?

[native plover]

I am using it for now but I saw that Tinymist has it by default so thought would be nice to just use Tinymist

[vast rock]

isnt touying-exporter stuck at typst 0.12 currently?

yea

though probably just need to change the typst-py version in pyproject.toml

https://github.com/touying-typ/touying-exporter/pull/10

[oak quest]

im not using touying tbh. my slide documents still stick to polylux 😂

[vast rock]

just need to merge this ig

[vast rock]

same lol

but the exporter works with polylux too

[oak quest]

didn't know it.

[vast rock]

yea i tried it and it worked. since its just svgs crammed into impress.js, the package used doesnt really matter

[jovial saffron]

Not sure if there is anything you can do about it, but Cortex XDR (a security app monitoring anomalies on my work MacBook Pro) terminated VSCode repeatedly during the weekend immediately when I typed a character in a Typst document while having Tinymist active. Disabling Tinymist stopped it.

Ie. it thought Tinymist is malware.

It uses some heuristics for detecting possible malware. Now it works just fine, but I'm not sure if Tinymist was whitelisted or not. I was contacted by our IT and reported that Tinymist can be whitelisted - if they didn't do it already.

Anyway, just to let you know if similar thing is being reported later.

[oak quest]

There is a similar issue: https://github.com/Myriad-Dreamin/tinymist/issues/1394
No particular good solution I guess, other than I pay some money to ask Apple and Microsoft for stopping spam it as a malware, by sign it with some trusted certificate.

[oak quest]

Today's random linter stuff.

I only matched the pattern x logic_op y, there would be perhaps some other pattern like type(t) in ("string", "integer").

[jovial saffron]

Yea, and I have no idea how the mentioned Cortex XDR works and if there even is a possibility to globally whitelist it.

[oak quest]

Then I finally view the lint as a "task", like exporting a pdf, and you can determine that whether run lint checks upon Save or Type. Then I set the default lint setting to lint on save, which wouldn't aggressively use your CPU resource per entering a character.

[oak quest]

@vale axle do you have knowledge about cargo? If so, you could simply rebase my branch to typst/typst:main and change these patch https://github.com/Myriad-Dreamin/tinymist/blob/258cf85601fd111833316d9756bf094fba2f798c/Cargo.toml#L266-L274

Somebody requested to make new release in nightly channel, but the maintainer of the nightly channel is busy now. https://github.com/Myriad-Dreamin/tinymist/issues/1677

[vale axle]

Ah, I already did that (although I'm changing to my own branch of typst). It seems to be not compiling for lots of reasons, although I'm not at my pc right now

Although, after this, I'm then supposed to build the vsix in contrib / .... / vs-code and it will use that tinymist.exe automatically?

[oak quest]

you could start a drafting PR, so that we could discuss and solve it concretely. The only reason we are not actively publishing nightly releases is simply caused by we don't have bandwidth to do that. If you start a PR, volunteers may participate to make a nightly release together.

[vale axle]

Yes. It doesn't seem fundamentally impossible and I may have time in the coming weeks although unfortunately I have lots of work to do

[oak quest]

I think you don't have to build it locally, you could only setup cargo to check compile errors. When you start a PR, the GitHub will build artifacts for you, which is the simplest way.

[vale axle]

Yes. Although maybe it would be easier for me to build locally.

Thanks for the help btw

[oak quest]

Clicking the checkmark on the PRs, and drag down to the summary page, and you will see them, .

[harsh wing]

https://github.com/Myriad-Dreamin/tinymist/issues/1687
Not the trailing whitespaces!

[oak quest]

I'm communicating with prettypst to replace typstfmt. https://github.com/antonWetzel/prettypst/issues/9

typstfmt is awesome and has stayed with us for 2 years, but it should be unfortunately replaced as it is eating my time to maintain tools.

because as typst changing its syntax, it becomes non-trivial to upgrade typstfmt to support new typst.
prettypst is another formatter made in early times, it was proposed to get merged in typst/typst but that PR was not merged successfully. I'm not sure its code quality, but it should be much more easy to migrate some features from typstfmt to prettypst, especially those missing in the typstyle. This could be an alternative solution of typstyle.

[jaunty kiln]

I'm curious about topiary

It can use tree sitter grammar for formatting

[oak quest]

sounds like a promising (to be complete) but heavy solution. Is it ever used by someone?

[oak quest]

For the ML languages. OCaml is a language Topiary can handle, but the sheer complexity of languages like Haskell place them well out of scope for a "Good enough" formatter. I.e. the effort to make a "good enough" formatter is not far removed from making a "good" formatter.

We are still in a phase where the allowed capture groups aren't set in stone yet, and maybe Topiary will become powerful enough to take on larger challenges like Haskell in the future, but for now we our focusing our effort on what we consider languages with less complicated style preferences.

topiary has to handle code style for languages. typst is likely not suitable. topiary on markdown could answer this, but I didn't find it from quick search.

https://www.reddit.com/r/rust/comments/11mt3dn/topiary_a_code_formatting_engine_leveraging/

[jaunty kiln]

I know the devs showed up here early on and mentioned that they'd be interested in adding typst support if someone else did the grammar

I'll see if I can find the thread

It's in #old-showcase. For whatever reason I'm not able to link to it

[cursive bough]

since typstyle has reached full maturity, and is being deligently maintained, why not stick with typstyle? even make it the default option?

[jaunty kiln]

Isn't it already the default?

Anyway having options is always good

[cursive bough]

Not really. There is always time squandered away for little benefit. Plus, once you make something an option, it is very hard to take that away.

[oak quest]

it is set as default recently

[oak quest]

it is hard to get rid of feature s that many people rely on, but I think it is feasible to deprecate and remove things in years, like typstfmt. adding a lightweight formatter doesn't cost anything so i'm willing to add two to three well maintained ones if there is any, since people really have different preference on code style.

[cursive bough]

Sounds great to me.

It is amazing how mature tinymist + others have become so fast...

[earnest grove]

hello! i'm looking for some help with a multi-file project, specifically importing styling, e.g. headings, from parent directories into a subfile
i've looked at the resources mentioned in this support post https://discord.com/channels/1054443721975922748/1318273312383565934
but still can't figure it out
currently i have the output of LspTinymistGetServerInfo giving the proper root dir, as is obtained by nvim-lspconfig by the presence of a .git folder
but the relative imports from subfile say:
failed to load file (access denied); Cannot read file outside of project root.
end goal is to have a shared preamble
if anyone has any place to redirect me for support, that would be much appreciated; thank you!

[oak quest]

what's the structure of your project? Where is the preamble file?

[earnest grove]

root/class/premable.typ
root/homework/.../hw1.typ

[oak quest]

• root/class/preamble.typ
• root/homework/x/hw1.typ
  How do you import the preamble.typ in hw1.typ? If the root is set to root/, it should be either or:

// in /homework/x/hw1.typ
#import "/class/preamble.typ" // or
#import "../../class/preamble.typ"

[earnest grove]

the first way tries to path from the dir of hw1.typ, the latter way gives access denied error
altho, as above, the root dir environment seems right, but then i wouldnt need relative pathing so it's probably not actually

[oak quest]

Then, I only know you haven't set the root correctly, which is not a problem about multiple-file projects. This should be handled by nvim-lspconfig tho.

[earnest grove]

hmm, ok; ill keep trying.

thank you for the help!

[earnest grove]

just for the record, deemed the appropriate way to manage styling like this is to create a local package and import it
(am just symlinking to local pkgs from parent dir)

[oak quest]

I'm not sure whether it is a tinymist issue or a misconfiguration in emacs. The emacs users could help it: https://github.com/Myriad-Dreamin/tinymist/issues/1685

[jaunty kiln]

@oak quest the link to "harper" on the github page is broken

[oak quest]

ic

[earnest grove]

happy to report all works well on helix!

[lapis scarab]

is there a way to live preview in a pdf viewer (sioyek) instead of the browser? ( i use helix as my editor )

[lapis scarab]

ok so i read the docs and i used config = { exportPdf = "onType" }. but sioyek seems to crash onType

fish: Job 1, 'sioyek a1.pdf' terminated by signal SIGSEGV (Address boundary error)

[lapis scarab]

nvm i just switched to zathura

works flawlessly

[buoyant sand]

Lol

Sioyek seems a little unstable at times

I get random crashes with it even without export on type

[lapis scarab]

thats sad, i was kinda liking sioyek

[hoary stag]

Can I configure tinymist to treat the first opened Typst file as the project file (so that it's used for preview until I quit the editor)?

Currently I set typstExtraArgs = ["main-file.typ"] in .helix/languages.toml, but I have many projects with several documents in the same directory (for example lecture notes, exercises and slides).

It's annoying having to edit .helix/languages.toml to work on the lecture notes, then edit it again when I need to work on the exercises. I'd prefer if I can open the lecture notes in a helix instance, the exercises in another instance

[harsh wing]

Have you tried tinymist.pinMain command?

In Neovim you can totally do this, but I guess it could be also possible to deligate this to the server with a new option.

[hoary stag]

doesn't work in helix (it takes an argument which helix doen't provide)

[harsh wing]

Well, Neovim also doesn't provide this, but I can set it up for this.

  commands = {
    PinMain = {
      function()
        vim.lsp.buf.execute_command {
          command = "tinymist.pinMain",
          arguments = { vim.api.nvim_buf_get_name(0) },
        }
      end,
      description = "Pin main file for better autocompletion/error messages",
    },
  },

[hoary stag]

you're spoiled with neovim 🙂 helix doesn't let me specify an argument, it's the nightmare for @oak quest

[harsh wing]

I'm sorry that I use a superior editor. 😎 ||just more mature||

[hoary stag]

I guess that's one point in favor of the clunky mess that is neovim 🙂

[harsh wing]

Vimscript ≠ Lua. But the internals: maybe...

[oak quest]

very much believe that the clean solution will be using the projectResolution = lockDatabase. I have finished most of it, but it lacks of testing. I will continue developing it in near future. some related issue: https://github.com/Myriad-Dreamin/tinymist/issues/1697

[hoary stag]

OK I'll test it when it's ready!

something I wonder since it's based on lockfiles: would this allow me to open lecture-notes.typ in one helix instance and exercises.typ in another helix at the same time? (when lecture-notes.typ and exercises.typ are two "main" files in the same directory)

[oak quest]

exactly. tools (not limited to the language server) write compilation commands into lock files and the language server picks the commands to compile the docs accordingly. this works pretty like clangd understand c header files.

[vast rock]

does this PR solve the helix argument issue? https://github.com/helix-editor/helix/pull/12527

from what i understand, it does

youd have to build from source though

[hoary stag]

oh sweet it does indeed

with the following helix config I can press Ctrl-P to pin the current buffer as main file:

[keys.normal]
"C-p" = ':lsp-workspace-command tinymist.pinMain "%sh{pwd}/%{buffer_name}"'

[vast rock]

nice

[hoary stag]

Now I'm trying to call scrollPreview from helix but it seems I must give a string as first argument (before the panelScrollTo event)

any idea what this string is?

[flint pollen]

hi. if I write a function docstring like this:
/// Each row is passed as one markup block (`[...]` syntax)

Then the [] brackets are not shown in the hover text. Is this a neovim side issue do you think or is it something tinymist does, that the ] get lost? With [[...]] in the source text they come through as [...] in the doc text

[buoyant sand]

the typst docs itself use `[...]` to render markup, `{...}` to render code and `$...$` to render math

could be that tinymist follows that convention

to be specific it's used to show values as source code in those modes, not necessarily to render the content itself

i.e. `{11pt}` shows 11pt in that red tone the webapp uses for length literals

[flint pollen]

aha that's neat

there's way more smarts there than I thought

sounds like double [[ is the right answer then

hmm no, {[...]} would be the one because it's illustrating an argument like table([...])

thanks!

[earnest grove]

probs not exactly the right thread for this, but my 2¢ related to the above convo:
nvim better than helix for typesetting until helix gets support for unicode concealment and snippets (properly out of the box)

[buoyant sand]

It wasn't really related to that though was it, it was about doc comment syntax supported by tinymist

[scenic wolf]

Hi.
How I can format my files within helix?

[vast rock]

[language-server.tinymist]
command = "tinymist"

[[language]]
name = "typst"
language-servers = ["tinymist"]
formatter = { command = "typstyle" }

should work

you might need to install typstyle separately but i think it's part of tinymist when u install it

then you can do :fmt

(or whatever other settings you need to make it format on save)

[scenic wolf]

Thanks!

[harsh wing]

I noticed that #1.1.to-bytes() is not completed. Can float.to-bytes.size be completed with 4/8?

[oak quest]

it is on todo list (but low prio because this is not a bug)

not a critical bug

Next release will integrate lightweight conversion to LaTeX/DocX/Text features based on typst-html (HTML export). I think reliable conversion can improve experience on submitting papers (and encourage people to write papers in typst).

• Reimplemented typlite by html export by @high cypress and @Myriad-Dreamin in https://github.com/Myriad-Dreamin/tinymist/pull/1684, https://github.com/Myriad-Dreamin/tinymist/pull/1720, https://github.com/Myriad-Dreamin/tinymist/pull/1728, https://github.com/Myriad-Dreamin/tinymist/pull/1729, https://github.com/Myriad-Dreamin/tinymist/pull/1754, https://github.com/Myriad-Dreamin/tinymist/pull/1758, https://github.com/Myriad-Dreamin/tinymist/pull/1759, https://github.com/Myriad-Dreamin/tinymist/pull/1762, https://github.com/Myriad-Dreamin/tinymist/pull/1779, https://github.com/Myriad-Dreamin/tinymist/pull/1767, https://github.com/Myriad-Dreamin/tinymist/pull/1771, and https://github.com/Myriad-Dreamin/tinymist/pull/1772
• Supporting LaTeX export in https://github.com/Myriad-Dreamin/tinymist/pull/1718
• Supporting Docx export in https://github.com/Myriad-Dreamin/tinymist/pull/1727
• Supporting Text output support by @high cypress in https://github.com/Myriad-Dreamin/tinymist/pull/1731

[jaunty kiln]

Casually dropping latex export

[harsh wing]

@oak quest, I think it's time.

[harsh wing]

Sorry for the hustle..

[oak quest]

Latest release has last remaining todo list:

• the markdown converter (typlite) doesn't like examples in fletcher docs, but I should be able to make it happy with few LoC changes.
• I would like to make it work, Sample workspace to Make and Prepare for Submitting IEEE Papers
• I found many compile errors in cetz and fletcher docs, it shouldn't prevent release but I can send issues to them.

[flint pollen]

There's a couple of incompatibilities between tinymist and tidy's doc syntaxes. That you know of course, but even if using old-syntax: true in tidy, they disagree on these points:

• Tidy wants documentation for parameter named ..args while tinymist wants this parameter to be named args in the list. Worse, tidy trips an assertion if you name this parameter args (because it doesn't find the corresponding thing in the function signature)
• Tinymist uses {...} and [...] conventions and tidy does not. Compatible option that works with both: [ A B C ] (extra spaces, yes I want to show the [] things in the raw code)

/// - args (arguments): Rows like `{[A & B & C]}` and other positional or named table function parameters.

Example above of a good doc comment for tidymist.

I know it doesn't necessarily make sense to "complain" here. I want to hear if others are looking at the same thing and have a good way to work with both in harmony

[oak quest]

for the first one, args may be better than ..args, to make less surprise while learning syntax.

[oak quest]

For the second one, this can be a bug, because tinymist processes docs from packages by the one used for processing docs from official typst. I could fix it quickly. this has two considerations:

• people can apply this convention easily by show rules.
• if I remove it, there will be again reduce a point to learn for docstrings.

[flint pollen]

I'm setting up doc generation with tidy now. The end result is very nice, it looks good. I still think that LSP/hover available documentation is more important than a nice pdf manual, so I'm trying to have both work as well as possible.

In many places I've updated to use for example This syntax: `typst none` (triple backtick) which works fine in both places and adds syntax colors to none in tidy.

[oak quest]

I'll be happy if you "complain" about that, and you are the second one complaining about it since the last year (zero till last month, and two since this month lol). I only kept it almost follow tidy syntax as I look on tidy docs again and again.

may be there should be a formal syntax to help achieve flawless consensus.

[sturdy surge]

What Tinymist setting disables inline argument name previews? I find them a bit annoying.

[harsh wing]

I think it's only editor-specific setting.

[sturdy surge]

Ah. They're called "inlay hints" and are a global VSCode setting

[oak quest]

I think it is disabled by default.

[buoyant sand]

Speaking of, does LSP provide any way to configure which of them are shown? I think it has a distinction between type and parameter inlay hints but all other hints are just whatever to LSP right?

I wonder how tightly one might control them because I like some of them but find other irritating and I have a keybind to turn them all on and off as needed

[oak quest]

it doesn't have way to show part of inlay hints.

i recent find some inlay hints could be useful for most people. like content preview of label links, it is useful for note taking.

and resolve autos and place the value around.

[harsh wing]

I assume Typlite is https://github.com/igumnoff/shiva, but not as versatile? I thought that it would be better to integrate that if possible.

[oak quest]

totally not the same. shiva parses other markup and converts the markup into typst source code and generates pdf by typst.

[harsh wing]

What? I thought it converts between formats.

[flint pollen]

these don't build for me: cargo +stable install [email protected] and cargo +stable install typlite. Do I need to use a nightly compiler or something?

the first error for the rc2 build is like this

error[E0432]: unresolved import `tinymist_derive::TypliteAttr`
 --> /home/user/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/typlite-0.13.14-rc2/src/attributes.rs:4:5
  |
4 | use tinymist_derive::TypliteAttr;
  |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ no `TypliteAttr` in the root

[high cypress]

try cargo +stable install --git https://github.com/Myriad-Dreamin/tinymist.git typlite instead

[buoyant sand]

semver violation? --locked maybe?

[flint pollen]

yeah good call, I should have tried locked too. I'll try both 🙂

one at a time...

[high cypress]

well according to the log at https://github.com/Myriad-Dreamin/tinymist/actions/runs/15527495046/job/43709795010, 0.13.14-rc2 does indeed fail to compile 🤣

[buoyant sand]

Lol

The publishing step seems to have been fine though

[flint pollen]

(locked did not help, stable rust, current "stable" typlite published)

I see that you added typlite to dist.toml - so I guess it builds binaries, where do I find those?

[flint pollen]

builds 🙂

[high cypress]

should build in next release (rc3)

[flint pollen]

if it's ok I'll report the first bug I found in .typ -> md conversion 🙂

[oak quest]

https://github.com/Myriad-Dreamin/tinymist/actions/runs/15672250005/job/44145339076#step:9:15

I think it is built per ci

[hardy grotto]

As instructed, let me move the discussion from the discussion channel (the one that is not meant for discussions).

I was telling to @devout tree :

While you’re here answering my questions (which is much appreciated): another thing I couldn’t find how to do with tinymist is changing the surrounding environment. A very basic operation is: you are editing a lemma and suddenly feel the need to turn it into a theorem. In vimtex that would be cse (change surrounding environment) anywhere in the lemma and then type theorem. Again I was able to use treesitter to get that, but I couldn’t see how to do that with tinymist.

and

Even more basic: I couldn’t even find how to open a table of content for my typst file (this I’m not yet done emulating it).

[devout tree]

I get the feeling tinymist doesn't currently offer that at all. the lsp is primarily designed for vscode, so any vim exclusives probably aren't (yet) implemented

although toc I feel should be doable one sec

[cursive bough]

You can quote yourself with > followed by the quote.

[hardy grotto]

Thanks, I’ll try to remember for next time.

[devout tree]

you can also forward messages now

[vast rock]

for quite a while actually

[devout tree]

all features are new features until they aren't

[hardy grotto]

That’s fair, but then I really don’t get how people can write that Typst is easier to setup in nvim than LaTeX then (I know this one wasn’t from you) or that “should be as easy as tinymist LSP + some sort of preview plugin”.

[cursive bough]

I want to be your friend.. We can talk about discord features!

[devout tree]

because I don't use those features when writing latex :P

[hardy grotto]

Anyway, the interesting part is really to learn how to have a nvim+typst setup that is (at least!) as good as nvim+latex.

Do you have nice auto-completion when using typst? For instance for references or citations?

What do you use for preview? I see in your config exportPdf = "never",

[cursive bough]

Neovim plus typst sounds nice actually.. Atleast when I watch neovim users on YouTube I think it looks cool.. But maybe it would be natural with typst to do the whole terminal warrior bit.

It's good that you're trying this out @hardy grotto

[flint pollen]

in neovim I use chomosuke/typst-preview.nvim for preview

[hardy grotto]

I’m sure nvim+typst can be glorious, I simply don’t have the right config yet.

[devout tree]

I don't use references or citations myself :P
I'm not writing papers, just doing maths/logic

[hardy grotto]

And you never refer to a previous lemma or equation?

[devout tree]

uncommon enough that i do it manually i think?
don't really remember, tinymist may have autocompleted it

[flint pollen]

also I've done next to no typst specific configuration of the LSP, the generic config seems to work well

[devout tree]

i use my own typst-concealer plugin, which is not suitable for most workflows

[cursive bough]

This is disturbing.

[flint pollen]

magic. Can it do tables?

[devout tree]

i try my best

[hardy grotto]

My computer doesn’t like your config 😢

[devout tree]

it can do anything 👍

[devout tree]

haha yeah its freaky and nixxy

[hardy grotto]

It says “E5113: Error while calling lua chunk: /home/pmassot/.config/wumpy/lua/config/colorscheme.lua:1: attempt to call global 'nixCats' (a nil value)”

[devout tree]

yeah requires all my nix stuff

if you have nix installed you can just nix run it

includes all LSP deps and stuff, makes me happy

[hardy grotto]

This is from simply putting you nvim folder on my machine and telling nvim to use it.

[devout tree]

yeah that won't work

[hardy grotto]

I don’t use nix, so I’ll try other ways to investigate.

Because I still get the lsp getting out of sync.

Which version of nvim are you using?

[devout tree]

some random nightly v0.11

v0.11.0-nightly+1351383

[hardy grotto]

I’m on v0.11.0 so it probably doesn’t make a big difference.

[devout tree]

likely not

you're possibly on a newer version even

[hardy grotto]

I’ll try chomosuke/typst-preview.nvim to see if it fixes my issue. I discarded it when I searched for typst related stuff since I really don’t care for instant preview (I only find it distracting if something is blinking next to my text editor when I type).

[hardy grotto]

Do you also have stuff in your lsp-config or mason config? Or does this plugin handles everything? It’s not super clear in the README.

[devout tree]

look it can even show LaTeX properly :p

[hardy grotto]

typst-preview is cool, but it doesn’t fix my LSP synchronisation issue…

I should have installed it earlier: today my 11 years old daughter prepared her first Typst presentation and I told her she needed to save the typst file to see the updated pdf.

I cheated by writing the preamble but then she found that writing the slide was way easier than in Libreoffice Impress!

[flint pollen]

nah, it just does the preview, yes I have lspconfig and mason

I mean I just had an existing lsp configuration for neovim when I started using tinymist so nothing really needed configuring except preview

[hardy grotto]

Is your neovim config publicly available?

[flint pollen]

nope, sorry

[hardy grotto]

Are you on neovim 0.11? I think it used to work in 0.10, but then I had no time to try more typst before I upgraded.

[devout tree]

just to clarify: what is your lsp sync issue?

only updating on save?

[hardy grotto]

No, it’s the LSP getting out of sync. Say I make some mistake and there is an error message. Then I fix the mistake and, very often, the error message sticks around until I do :LspStop and LspStart.

And when it does that the syntax highlighting also gets crazy.

[devout tree]

syntax highlighting also getting weird implies treesitter is broken no?

[hardy grotto]

Probably yes.

But InspectTree is still working, so it’s really weird.

I must have a crazy conflict somewhere.

I should try to minimize my config to see when issues start appearing.

[hardy grotto]

Ok, I cloned the kickstart config and added tinymist, and things seem much better. So there is a clear action path for tomorrow night: I need to clean up my main config. This is a good thing to do from time to time anyway.

Thanks everybody!

[oak quest]

not primarily designed for vscode because I try to abstract things into layers. and docment outline info can be also subscribed by nvim. there is somebody who would like to implement these things but they as a programming language developer was also keeping busy finding their job and has disappeared for half a year.

[harsh wing]

Last time I wrote in that language, it probably was on the same level or worse than Typst. So it's all subjective.

[harsh wing]

Tinymist supports reference completion, but a lot of times I don't get autocompletion, or a good list of suggestions, because of something. Like I need to sometimes re-insert @ for it to work. And when completing function arguments, they always appear at the very bottom of the list, but above colors, I think. When completing strings, it often will insert first character and ", so it becomes "s"something", which I have to manually fix.

I don't think I had these issues with other LSs, but it's definitely not ideal yet.

[oak quest]

it is probably more or less because you have many settings different from the usual ones like you say "because of something."

[harsh wing]

What are "usual ones" supposed to be?

[oak quest]

for example I remember you say the order of completion items are not perfect, but that is because neovim client rearranges it by itself.

[harsh wing]

The only settings that are different for Typst are

    vim.opt.iskeyword = "a-z,A-Z,48-57,_,-,192-255"
    vim.opt.spell = true
    vim.opt.colorcolumn = { 80 }
    vim.opt.backupcopy = "yes" -- `typst watch` notify issue

And the stuff like --mirror, root_dir, PinMain, and typst_syntax_highlighting_colors().

[harsh wing]

I use the same LSP client and its settings for all LSs.

[oak quest]

So why should it rearrange the items?

[harsh wing]

That's what I'm trying to understand.

I'm not into LSP, so I only know that it's somewhere between Neovim, hrsh7th/nvim-cmp(, neovim/nvim-lspconfig), and Tinymist.

[oak quest]

From my knowledge nvim-cmp has to reorder items because it accepts multiple completion sources and combines them into one, but for the reason it reorders the items in the end.

[harsh wing]

It does accept multiple completions sources, but they all provide very few items, so like 95% of items are from LS.

There are usually snippets near the top, a bunch of different "variables" at the very bottom, and fields/parameters are right above those variables, basically near the very bottom.

Here is the top

Here is the bottom

[harsh wing]

Is this copilot?

[oak quest]

this is lsp

[harsh wing]

So you just placed cursor to the left of above:?

or is this how your completion behaves?

[oak quest]

the transparent text is "inline completion" (or you can imagine it is previewing completion)

[harsh wing]

hm, ok

[oak quest]

@harsh wing I'm not going to be offensive, but I'm judging your settings from my little knowledge about neovim. It is great if you find lsp bugs and let me know.

[harsh wing]

If I make a test case with Nix, will you look into it?

I don't like the docker, because it's a big hustle and overhead. Nix is nicer, and more customizable.

[oak quest]

I'm a fun of nix, but I have no time to setup nixos because it might consume my time. I may use nix as my next OS but I don't know.

[harsh wing]

You would basically run nix run or something like this.

[harsh wing]

I didn't say anything about NixOS.

[oak quest]

The same reason for looking into neovim settings. This is slightly better because I have neovim installed in my PC.

[harsh wing]

Especially since you don't need a window server, unlike for Firefox, VSCodium, etc., it should just work on any UNIX-like OS.

[oak quest]

I don't have nix env either. I just googled it and find nix can be run in some container.

[harsh wing]

You can install Nix with basically one prompt on any OS.

[oak quest]

I thought it is an os

[harsh wing]

nah, that's an OS overlay on top of Nix.

Nix with (a lot of) extra steps.

https://nixos.org/download/

[oak quest]

what specific thing should I look into?

[harsh wing]

Well, nothing. I just wanna make a reproducible Nix config so that we know exactly what is used.

Not everyone is into Nix.

With a flake.nix and flake.lock you can do almost anything.

[oak quest]

I just ran hello world so I feel confident. I can take a look if you make a test case in nix, but I don't ensure the due date of seeing it.

[harsh wing]

https://nixos.wiki/wiki/flakes:

Other Distros, without Home-Manager
Add the following to ~/.config/nix/nix.conf or /etc/nix/nix.conf:

experimental-features = nix-command flakes

[flint pollen]

If you're using kickstart - I have something similar, then gO (LSP document symbols) is almost a ToC

[vale axle]

is there still interest on doing work to get tinymist to work with latest version of typst / has such stuff been done yet?

[oak quest]

there is a nightly release channe but the related maintainer is preparing the final exams.

[white solar]

i have no idea how to set invert-colors=never in the helix language config. did anyone manage to set it?

[buoyant sand]

how is it set in the vscode json config? It should be the same thing, just as toml, really.

[oak quest]

how do you start the preview?

[white solar]

using the lsp

[white solar]

ill check it!

i still didnt get it working :(

i tried adding this to the config

preview.background.args = ["--data-plane-host=127.0.0.1:23635", "--invert-colors=never"]

[buoyant sand]

I have a proprietary doc which I can't share for which tinymist preview seems to be in a waiting for message document is not ready loop

is this a known issue?

Using ```
tinymist
Build Timestamp: 1980-01-01T00:00:00.000000000Z
Build Git Describe: VERGEN_IDEMPOTENT_OUTPUT
Commit SHA: VERGEN_IDEMPOTENT_OUTPUT
Commit Date: None
Commit Branch: None
Cargo Target Triple: x86_64-unknown-linux-gnu
Typst Version: 0.13.1
Typst Source: git+https://github.com/Myriad-Dreamin/typst.git?tag=tinymist/v0.13.10#946ea31fb554bcf62e3215f64ddda87d70b026af

I found out there was a compile error in an old package

[oak quest]

I noticed the warning in new neovim, may be useful information:

The option definition `plugins.lsp.servers.tinymist.rootDir' in `/nix/store/via5xff0d3fbwg5nfxwpy0lijxbg9n3i-test-sources/plugins/lsp/_lsp.nix' no longer has any effect; please remove it.

https://github.com/nix-community/nixvim/pull/3220

This has been converted nicely in official neovim:

https://github.com/neovim/nvim-lspconfig/blob/314b35335cc84bc2a085c84c69da955ba22da163/lsp/tinymist.lua#L55

[white solar]

sorry if im not understanding something correctly
in vscode:

"tinymist.preview.invertColors": "never"

and in helix:

[language-server.tinymist]
command = "tinymist"
config = {preview.background.enabled = true, formatterMode = "typstyle", preview.invertColors = "never" }

it doesnt work in helix tho :(

[hardy grotto]

Sorry I missed your useful message. Indeed the LSP document symbols list seems to have everything I care about. Section appear as namespace and labels appear as constant. Maybe the result of the same LSP request could be post-processed into a ToC. Is this what people do in VSCode? How people do get a ToC when editing their Typst documents in general?

[oak quest]

In vscode it is passed via command arguments. The related code exists since typst-preview era. I think we could reimplement it at the server side to bring consistency to all editors.

[oak quest]

In VS Code, there are two outlines. One is from view of the programming language, that you see in neovim at the moment: the syntax document outline, and ther other is from view of the markup language, the rendered "PDF" document outline.

[vast rock]

I am not sure that is valid toml syntax

Cause I managed to make it work at least for the default preview

[white solar]

i got it (finally)
adding it to the docs would be helfpul :)

[cursive bough]

Do it 😂 👏 👏

[white solar]

ye ill make a pr

[cursive bough]

Epic.

[hardy grotto]

Is the second available through a LSP call?

[oak quest]

No, it is only valid through LSP notifications (the server will streaming push outline changes to the editor). If we would like to support document outline for all editors, might add a new UI page in preview panel.

[oak quest]

I started next big PR to solve the currently opening bug reports forever, https://github.com/Myriad-Dreamin/tinymist/issues?q=is%3Aissue state%3Aopen label%3Abug.

• https://github.com/Myriad-Dreamin/tinymist/pull/1842
• https://github.com/Myriad-Dreamin/tinymist/pull/1843

[flint pollen]

do you know if there's a good way using typlite to:
Have a grid in .typ to put two things side by side (raw typst source and the rendered output side by side)
and export this to markdown, still having it side by side, like in a table?

[flint pollen]

I think I know now, with a custom rule for grids to make html tables.

[oak quest]

I haven't understood your case but love to hear that you solved it.

[flint pollen]

I'm reporting a bug I found that's not directly about grids. You'll see it in a minute

the issue is https://github.com/Myriad-Dreamin/tinymist/issues/1844 and it's about m1verbatim which couldn't be used inside a html.elem for some reason. The use case is typlite .typ → .md

Another thing about grids in typlite is that if I put block elements inside a grid, then the block elements are dropped outside the markdown table syntax. (I have not reported a bug about that).

I created a side-by-side layout using grid successfully using the following (The raw blocks should use triple backtick, but I can't do that here because of discord)
So that demonstrates what I wanted to do.

#let x-target = sys.inputs.at("x-target", default: "pdf")
#let show-if(cond, func) = body => if cond { func(body) } else { body }
#show grid: show-if(x-target == "md", it => {
  let columns = it.columns.len()
  html.elem("table", attrs: (x: "test"), {
    for row in it.children.chunks(columns) {
      html.elem("tr", {
        row.map(cell => html.elem("td", cell)).join()
      })
    }
  })
})

#show raw.where(block: true): show-if(x-target == "md", it => {
  html.elem("pre", html.elem("code", it.text))
})

#grid(
  columns: 2,
  `
  Raw block 1
  `,
  `
  Raw block 2
  `
)

[oak quest]

it looks like a bug if you put markdown syntax verbatim inside <grid/>

[flint pollen]

do you mean that it's a bug in typlite? Or that I can't do that, I shouldn't put markdown verbatim inside a html td element?

[oak quest]

you can't do that.

[flint pollen]

ok, so user error

but the reason I'm trying that is that I need to apply some html rule to the raw block. And wrapping it in pre doesn't actually work for the markdown output

[oak quest]

I didn't consider how m1verbatim should work in md export, so it might not work nice. A subtle thing is that we can output an element in either md or html syntax, and md syntax cannot be nested in html syntax. you hit it in a day we release the typlite 😂

[flint pollen]

I have a challenge for you: I want to create the following markdown document from typst: https://gist.githubusercontent.com/bluss/b06cb65a93a741555670868b104c892c/raw/986a1d108b6d361ce9e6ead5838198c3fd15505f/mdexample.md

As you can see it has a fenced code block inside HTML elements. Why? Because this is what works 🙂 As you can see it renders ok. https://gist.github.com/bluss/b06cb65a93a741555670868b104c892c#file-mdexample-md

[cursive bough]

there is a saying, when it rains, it pours.

[oak quest]

maybe we should keep m1verbatim as is regardless where it is put in a typst document, which is exactly how it is named.

[flint pollen]

I guess for this small thing it's a bit silly to use typlite but it would have solved some problems - makes it easy to have many examples of the package in the readme

[oak quest]

yeah it is good to write typst docs once and show it on github with rendered md, no more duplication by hand.

[flint pollen]

the fallback here would be to write .md directly, have a script that extracts and renders the examples to svg, which is also ok. But for packages if as much as possible can be rendered to md/html it saves precious space in typst/packages

[oak quest]

it is another question that whether we should render and commit example refs. not yet find a perfect solution.

having one or two small examples might be good enough, and as you said it will be expensive to render all of them.

I think full documentation could be rendered in html and deployed on github pages.

and only keep short documentation and highlighted use cases in readme.

[flint pollen]

This can be done using insane hackery and the current verbatim... 😁 Just replace html.elem with custom versions that insert verbatim html tags..

[oak quest]

just like what I show in tex export.

it will take ten year to build typlite to let it fully understands and emits correct tex macros. before that, only verbatim saves all the life.

@flint pollen I think we can let
typst grid element work with raw blocks. I understand you didn't use it for some reason

[flint pollen]

Because without any custom style, typlite drops the raw block before the grid https://gist.github.com/bluss/1843bcfc05318aa51b6ec7b9e375ff9c

similar for any block level element I think

[oak quest]

this is a bug.

[flint pollen]

ok, I can report it

[vast rock]

Since this is resolved due to https://github.com/helix-editor/helix/pull/12527, can we expect :lsp-workspace-command to take the fitting arguments soon like the current buffer file for pinMain? Similar to how it is with neovim? I think I can already add a keybind for this manually but yea

(granted this is only on helix main, not on a release but still)

[hoary stag]

see #1260973804637786224 message

[vast rock]

yep that but built-in

[hoary stag]

I don't think the new helix feature changes what tinymist can do automatically? if it can get the current buffer name without the user adding it themselves to the lsp command, presumably it could have done that earlier already?

I mean for helix tinymist is only the server side, it doesn't ship any client

(I guess that could change once helix finally gets plugins)

[oak quest]

not automatically, but it already makes big progress to allow lsp calls with arguments.

[oak quest]

maybe the docs can be updated once the feature has become steady.

i'm more seeking the lock file as a game changer, that remembers your compilations (export pdfs, or preview) and infers main files for any files smartly. i'm still not sure whether people will like a lock file in roots but it will always be an opt-in feature, by a config per workspace or in global, so I don't quite worry.

[buoyant sand]

I would personally love it

I'd also like having typst.toml to mark these things in general

Any sufficiently complex document should have one

[oak quest]

laurenz personally prefers single-file documents, and modules can be created and maintained as packages. I think it is good as well.

in the cases, we shouldn't and don't have to generate lock files to guess the main files.

[buoyant sand]

I agree for simple documents

But a thesis is not a simple document and the churn of having to reconfigure the root or main file all the time is more annoying than having a typst.toml that lists these

Which is why I said sufficiently complex documents

[autumn sandal]

Hey, I'd like to have a pre-compile hook that would run a script before compiling the typst document.
This is useful for implementing features that are not possible from within the typst document (say perhaps: https://discord.com/channels/1054443721975922748/1387248729169334353)
While I could make a script to do some pre-compile step, then run the typst compiler, it wouldn't work as smoothly as a hook inside tinymist, for previewing within vscode.

Is this already a feature / is there already an issue for this?
If its not already a feature, I'd be happy to implement it.

either a precompile hook, or a way to call a different compiler (which could just be a script that forwards arguments to the actual compiler)

[jaunty kiln]

I don't think creating documents that are compilable only in Tinymist is a good idea

[autumn sandal]

can i not use the tinymist cli to replicate the behavior?

[oak quest]

This is very much like a custom task and we are on the road of making it. The next steps:

• let the server understand vscode tasks, to run them at server side.
• watch and run query tasks and post the query result to your script.

https://myriad-dreamin.github.io/tinymist/feature/export.html#label-Hello World Example (VSCode Tasks)

[oak quest]

I think they will be compiled if you upload the artifacts to webapp. in fact indeed arguable because you won't be able to preview to get final document in webapp.

[autumn sandal]

hm this seems kinda convoluted, but ig it would be more powerful?

[oak quest]

how?

[autumn sandal]

maybe an example would help me to understand. Perhaps this syntax?: ```json
{
"version": "2.0.0",
"tasks": [
{
"type": "typst",
"command": "export",
"export": {
"format": "query",
"query.format": "json",
"query.field": "value",
"query.selector": "<git-tracked-source-file>",
"query.pipeTo": "${workspaceFolder}/my-script.sh"
},
"runOptions": {
"runOn": "typstPreCompile"
}
}
]
}

this would run my-script.sh with the output from the query?

hm ig that could maybe work, since the runOn could be used for any task

lemme see if thats an actual thing you can do in vscode tho

[oak quest]

just let you know that, you may not have to rely on watch query. I ever ran the query tasks to make RPG games using typst.

[autumn sandal]

does not seem like you can contribute to runOn, from an extension

[autumn sandal]

wdym watch query?

[oak quest]

That should be valid, but query.pipeTo may be a potential security risk, because tinymist may run on untrusted workspace, we need be care about such cases.

[oak quest]

So I said this have to be loaded to the server because extension cannot do that.

[autumn sandal]

isnt any task a security risk?

[oak quest]

Instead watch your document and run the query tasks, you can run the query tasks manually.

[oak quest]

I don't see only export PNG introducing a security risk.

[autumn sandal]

I’d assume that running any task requires you to trust the workspace for vscode

[oak quest]

but currently running any typst tasks doesn't require you to trust the workspace. It's all safe.

[autumn sandal]

huh

[autumn sandal]

extension can run a task, it just can't add a new runOn option, no?

[oak quest]

likely

we might not be able to contribute a new runOn option

But may look like this for a watch task:

{
  "version": "2.0.0",
  "tasks": [
    {
      "type": "typst",
      "command": "export",
      "watch": true,
      "input": "main.typ",
      "export": {
        "format": "query",
        "query.format": "json",
        "query.field": "value",
        "query.selector": "<git-tracked-source-file>",
      },
      "runOptions": {
        "runOn": "folderOpen"
      }
    }
  ]
}

[olive bison]

Tinymist works so well in VS code, its really a pleasure to use it. Thanks for your work on this!

[oak quest]

i built the first neovim integration tests. ideally every issue will be made a corresponding regression test.

[keen moon]

You still don't plan to support the extension so that it works remotely, right? 🥹

[oak quest]

"remotely" may refer to work with "ssh-fs" (fore living share) or "vscode-remote". I'm using tinymist with vscode-remote with a remote linux machine. as for working with ssh-fs, it is doable but I should pay more effort to make tinymist more stable.

[keen moon]

Oh, OK! I see! I was just referring to VSCode Remote Server. Thank you very much!

[somber drum]

Has anyone gotten the background preview to work in Zed lately? It doesn't seem to work for me at all anymore

[minor moss]

Hello, I hope this is the right place to ask. Is there a way in VSCode with Tinymist to get a list of local packages when "Initializing a new Typst project based on a template" ?
Right now one needs to know the exact syntax / package name / version and type "@local/packagename:0.1.3".
I see that Tinymist already recognizes local packages in the VSCode interface.

It's not that big of a deal, but it would make the process easier on my side when creating new documents from local templates.

[wheat garnet]

if you type #import "@local/ then I already get a list of packages as autocompletion. do you mean something different?

[minor moss]

I'm thinking about this feature in the command palette.
You select it, provide the name of your package, and it creates a folder and creates a "starter" document based on how the template defined it.

It would be perfect if that box suggested the available local templates, in the same way there is the autocompletion on typing "#import ..."

[minor moss]

👀 😁

[oak quest]

goto the template gallery and favorite a template, and you will see it

I admit I don't know why I did that

but pour down hundreds of templates for completing here is also not good.

[minor moss]

Ok so that is indeed a potential solution, however I'm interested in local templates. (I'm managing a template for my company with some private info I can't put on @preview)

From what I understand, that is not possible in the current state. I let you tell me if you think that could be an interesting feature request. But I do get it if it is not a priority or direction you want to go to !

[oak quest]

I have many other things to do, but I'll love to review if you open a feature request.

We can get list of @local/@preview packages from language server in vscode easily:
https://github.com/Myriad-Dreamin/tinymist/blob/5c6c118b9c73e69a196fa98f6b256aa5bb7cbcc3/editors/vscode/src/features/package.ts#L82-L83

And we can have completion by not using getUserPackageData(context) but using that from language server:
https://github.com/Myriad-Dreamin/tinymist/blob/5c6c118b9c73e69a196fa98f6b256aa5bb7cbcc3/editors/vscode/src/extension.ts#L478

In template gallery, we didn't use a list passed from the vscode, but fetch one from typst universe in place. This looks not food if you maintain @preview/@local packages for you company.
https://github.com/Myriad-Dreamin/tinymist/blob/5c6c118b9c73e69a196fa98f6b256aa5bb7cbcc3/tools/editor-tools/src/features/template-gallery.ts#L331-L333

I notice that you said feature request

instead of pull request

feel free to open an issue and I can make it when I have time, but I don't ensure a ddl.

[minor moss]

So yeah I meant an "issue" requesting a feature. And of course, absolutely no pressure on doing it, or not, at any time ! It really is not a big deal and a small QoL feature.
Since you pointed at some parts of the code, I'll take the time to look into it and see if I can make some not too educated trials on my side.

Thanks for your time !

And thanks for the awesome work !

[empty girder]

Is there a way to debug with Tinymist a “layout did not converge within 5 attempts”? I get that very fast with Alexandria and Glossy, I noticed that the Tinymist preview still manages to compile the document but the Typst CLI does not

[flint pollen]

layout did not converge is a warning, so if they both produce the same warning they should both make some kind of document output..

[oak quest]

you can get a hacker function from the issue

https://github.com/Myriad-Dreamin/tinymist/issues/1307

oh, here https://github.com/typst/typst/issues/4528#issuecomment-2665755004

https://typst.app/universe/package/layout-ltd

[empty girder]

Oh cool, thank you!

[empty girder]

Right, don’t know why I just defaulted mentally to it being an error

[flint pollen]

oh it would be cool to have better tools to debug this, I did a lot of debugging without understanding (and a little bit of using layout-ltd, but it doesn't give so much more info)

Using layout-ltd and mapping out the final value of a state with different iteration counts could be useful..

@oak quest there's no sidechannel to give us debug logs in tinymist, from the document itself? For example to log state values in different iterations 😺

[oak quest]

I expect debugger will be much more useful in the case. I cooked an experimental debugger and see typst stops on a same function in layout iterations.

only having PDF output after each iteration might not be very useful.

[flint pollen]

The thing I learned from my last bout of this is that state values need to converge. In my case it was floating point accuracy problems, worst kind of problem

[oak quest]

For example to log state values in different iterations
Sounds like a good idea.

[flint pollen]

and hover info shows all lengths with 2 decimals but my problems where in the 12th and 13th decimals, so I was kind of blind to it (until Eric helped out) 🙂

[flint pollen]

anyway, this is probably different. The most obvious thing that can happen is if something moves or if something is added to the document as a feedback to a query or state change - and then that again has a feedback effect on the query

[oak quest]

I'm wondering whether we can have a counter updated per layout iterations

and we will be able to sample the values according to the counter, like this:

#if layout-iter() == 0 {
value // hover me
} else if layout-iter() == 3 {
value // hover me to get the value in the 3-rd iteration.
}

[flint pollen]

that's nice

[oak quest]

layout-itd:0.2.0 😁

[flint pollen]

it could maybe be an lsp command? region -> sampled values per iteration? (I don't know what's possible)

[oak quest]

no need for a lsp command. But either a user must implement a layout-iter or the compiler must provide one.

since the second value is only reachable in the 3rd iteration, you surely can see what value is used by hovering on it.

[flint pollen]

layout-ltd uses something close to that, but it's abusing count.final() and then updating it based on the final value to force another iteration. I don't think count.final() can be used for this unfortunately

[flint pollen]

if it's impossible to implement then maybe compiler help or lsp help is needed

[oak quest]

I think I cannot implement a layout-iter at once, because typst layout doesn't work as I imagine.

No, but with layout-limiter(n), you can hover to get the value in the n-th iteration.

[flint pollen]

?r I think this is the whole tree. But we don't get sampled values in the intermediate iterations?

#let mystate = state("teststate")
#mystate.update(0)

#context {
  let empty = [#metadata(none)]
  let x1 = query(<x1>)
  let iteration
  if x1 == () {
    // iteration 1
    iteration = 1
    let _y = mystate.get()
    //_x
  } else {
    let _y = mystate.get()
    
    let x2 = query(<x2>)
    if x2 == () {
      iteration = 2
    } else {

      let x3 = query(<x3>)
      if x3 == () {
        iteration = 3
      } else {
        let x4 = query(<x4>)
        if x4 == () {
          iteration = 4
        } else {
          iteration = 5
        }
        [#empty<x4>]
      }
      [#empty<x3>]
    }

    [#empty<x2>]
  }
  [#empty<x1>]

  [Finished at iteration #iteration]
}

[flint pollen]

Nice, layout-limiter confirms that the iteration count is actually correct as written in that code

[oak quest]

you can only get samples in the last iteration.

not convenient though

[flint pollen]

convenient doesn't matter when we are exploring if it's at all possible

first invent then make convenient 🙂

[flint pollen]

this typst package readme is generated from .typ -> md by typlite https://typst.app/universe/package/tabbyterms
I think it rendered really nicely. But it's using some custom m1verbatim stuff to emit HTML tags to enable tables in tables and raw blocks in tables.
And apart from tables inside tables, there's almost nothing advanced that it needs to show, just that a table was created and using strong emphasis..

[oak quest]

My today dev on README.typ:

= Installation

#let typlite-ver = toml("/crates/typlite/Cargo.toml").package.version
#if type(typlite-ver) == dictionary and typlite-ver.workspace == true {
  typlite-ver = toml("/Cargo.toml").workspace.package.version
}

== Install prebuilt binaries via shell script

#raw(lang: "sh", block: true, {
  "curl --proto '=https' --tlsv1.2 -LsSf https://github.com/Myriad-Dreamin/tinymist/releases/download/v"
  typlite-ver
  "/typlite-installer.sh | sh"
})

[flint pollen]

nice, reading toml is a great trick for .typ files

[oak quest]

I find I can also query shell scripts in the readme to test them in CI:

λ typst query 'docs/tinymist/crates/typlite.typ' --root . --features=html '<shell-script>' --field value
["curl --proto '=https' --tlsv1.2 -LsSf https://github.com/Myriad-Dreamin/tinymist/releases/download/v0.13.14/typlite-installer.sh | sh"]

everyday feels better with typst..

The bad thing was that:

λ typst query ... 'raw.where(lang: "sh")'
error: raw is not locatable

The good thing was that:

show raw.where(lang: "sh"): it => [#it#metadata(it.text)<shell-script>]

May be I can generate this file using typst

https://github.com/Myriad-Dreamin/tinymist/blob/main/editors/neovim/Configuration.md

To fix this error: https://github.com/Myriad-Dreamin/tinymist/issues/1853#issuecomment-3014836261

As well as fix the outdated page: https://myriad-dreamin.github.io/tinymist/configurations.html

luckily all of the package.json and Cargo.toml files can be read by typst

[tall wadi]

hi, all, how to check the version of Typst used bu tinymist?

[oak quest]

?r either check the log to get the repository url at the time of built the tinymist. or use this:

#sys.version

[tall wadi]

Can we use our own compiled Typst binary?

[wheat garnet]

tinymist bundles its own Typst, but if you compiled Typst yourself you could also try to compile Tinymist yourself with the same patches applied. I'm not knowledgeable how to do this though.

[oak quest]

1. cherry picking the three commits from https://github.com/Myriad-Dreamin/typst/commits/tinymist/v0.13.10/ .
2. Change the patch to your typst fork,
   https://github.com/Myriad-Dreamin/tinymist/blob/1478280a0749a8fdd22126c8e8681f99ad0ce5dd/Cargo.toml#L275
3. and trigger a github action, you don't have to build a tinymist binary locally.

A sample PR if you would like to use a nightly typst: https://github.com/Myriad-Dreamin/tinymist/pull/1279/files

The file should contain all information about getting a "released tinymist binary". https://github.com/Myriad-Dreamin/tinymist/blob/main/docs/release-instruction.md
I wrote it in case of one day I'm suddenly away from keyboard.

[pine bloom]

Clicking on a figure in the preview doesn't jump to the text in the editor and viceversa, putting the cursor in a figure doesn't jump to the corresponding place in the preview. Is that a known limitation?

[oak quest]

might be a regression, but I cannot judge it without a concrete case. preview jump is only served in best effort.

[flint pollen]

Is this expected in tinymist? I'm using neovim. If I have some code with a compile error like this:

// This is other.typ
#let test-func(enable: false) = {
  if not enable { }   // error: cannot apply 'not' to integer
}

// This is in main.typ
#test-func(enable: 1)  // I want a hint diagnostic here

The code is in two different files. I would like to have a hint/warning diagnostic on the indicated line in the main file. Because compilation is failing but no diagnostic says why (in the main file). Currently I only see one diagnostic, an error, and it is on the indicated line in other.typ

The webapp does it for this particular test case

I think the information is there, not as a separate diagnostic, but as 'relatedInformation'. So the following nvim bug applies (maybe, I don't have much knowledge): https://github.com/neovim/neovim/issues/19649

[flint pollen]

uh maybe this means that neovim is advertising that it supports relatedInformation but doesn't? https://github.com/neovim/neovim/pull/34454 Not sure I understand. I wonder if I can configure those capabilities

[flint pollen]

The listed workarounds didn't work well.. but I managed to fix it in my nvim config... RIP my sanity, that was tough 🪦

[flint pollen]

🥹

[flint pollen]

as you know I reported an issue for this. My understanding now is that it would be best if tinymist could emit a separate diagnostic for the thing that's in relatedInformation right now. Especially because it's in a different file.

[oak quest]

today i'll prepare for tinymist v0.13.16

[oak quest]

editors/neovim could have diag_spec.lua to confirm the possible bug and prevent regression.

[flint pollen]

also seen in typlite land:

The _term list_<tl> is the following syntax in typst:
....
#show <tl>: link.with("https://typst.app/docs/reference/model/terms/")

we're reStructuredText now 😏

[oak quest]

@flint pollen the main contributor of typlite is busy, so we cannot fix the typlite bug in this release.

[oak quest]

This looks better

?r

#let extern-link(target, href) = body => {
  [#place(hide[#figure(kind: "link", supplement: "link", href)#target])]
  show ref.where(target: target): it => link(href, it.supplement)
  body
}

#show: extern-link(<tl>, "https://typst.app/docs/reference/model/terms/")

The @tl[term list] is the following syntax in typst:
....

[flint pollen]

that looks cool, will experiment with that

[flint pollen]

no rush, don't worry. Do you mean the block elements thing? That seems like a bigger thing or feature request anyway

[oak quest]

Yes. The current lib assumes that no markdown content can be nested in html tags, which is a bit wrong. He need time to change it.

[flint pollen]

I'll hack on something derived from this. I was trying to understand and it works when just deleting the figure 😄 so that one doesn't look necessary?

[oak quest]

the figure is only a hint for auto-completing the reference.

[flint pollen]

ohh

?r the code is getting too complicated (it's not so satisfying) but I would prefer to write a document where it was like this - link targets can go after the link itself

#let linktarget(label, href) = {
  [#metadata((linktarget: true, href: href))#label]
}
#let linkrefs(body) = context {
  let links = query(metadata)
    .filter(mt => type(mt.value) == dictionary and mt.value.at("linktarget", default: false))
  let labels = links.map(lt => lt.at("label", default: none)).filter(l => l != none)
  show selector.or(<__linkrefs_dne>, ..labels.map(l => ref.where(target: l))): it => {
    for link in links {
      if link.has("label") and link.label == it.target {
        let supplement = if it.supplement == auto { "<link>" } else { it.supplement }
        std.link(link.value.href, supplement)
        return
      }
    }
    it
  }
  body
}

#show link: underline
#show: linkrefs

The @tl[term list] is the following syntax in typst; see also @bl[blocks].

#linktarget(<tl>, "https://typst.app/docs/reference/model/terms/")
#linktarget(<bl>, "https://typst.app/docs/reference/layout/block/")

[flint pollen]

using ref syntax is good though. I kind of like the "original" _term list_<tl> too

[flint pollen]

what do you think about the relatedInformation issue? do you think a new separate diagnostic can be added for the help/info part?

[oak quest]

I think the correct direction is to show "..." on the hinted code like rust-analyzer.

while tinymist doesn't show dots in the same way.

[flint pollen]

hm. I don't see how that matches the scenario in the report, where it's across two different files

the diagnostic is placed in an auxiliary (or library) file but the user is most interested in the bug which from the user's perspective was in their own code in the main file, that's where they need to go to fix the problem

That's why it's high impact to fix this, if I understand correctly. The diagnostic is in file B but the user needs to edit the code in file A to fix it. (So it would be great if the diagnostic helped point the user to the right location in file A)
I don't use vscode, maybe the user experience is already good there in the same scenario.

(will test vscode now 🙂 )

[oak quest]

I didn't craft an example that the hinted code are in another crate, but I think you can imagine it.

[flint pollen]

ok testing vscode was fast to set up. I think the same problem is present there as well?

[oak quest]

yes.

and rust-analyzer in fact is placing dots on the hinted code, while tinymist doesn't do it in the same way.

[flint pollen]

ok. sorry for writing too much about this, just want to clear up miscommunication if any

[oak quest]

There is surely some code and behavior we can copy for the r-a

[flint pollen]

rust is not a dynamic language, so I had trouble really finding an equivalent situation in Rust

[oak quest]

I think it isn't about the language themselves, but I found that r-a did use LSP to let vscode show dots instead of not showing anything, so we should be able to do that as well. I don't think we have to duplicate diagnostics to show an unique diagnostic at both error code site and hint code site. But:

• show the diagnostics at the error code site.
• show dots at the hint code site, and hover and see "Hint: ..." and an additional reference to "original diagnostic"

[flint pollen]

hm. what does dots mean more specifically?

[oak quest]

😂

[flint pollen]

aha. So that's not a diagnostic? I wonder what the corresponding thing is in neovim

[oak quest]

yeah, might still be a "relateinformation". but the dots might be too inconspicuous to attract you before.

[flint pollen]

I think the hint here is a separate diagnostic

nvim doesn't support relatedInformation I think

[oak quest]

https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#publishDiagnosticsClientCapabilities

[flint pollen]

Neovim - until development version 3 weeks ago - says that it supports relatedInformation but in reality it doesn't

[oak quest]

https://github.com/Myriad-Dreamin/tinymist/blob/1478280a0749a8fdd22126c8e8681f99ad0ce5dd/tests/fixtures/initialization/neovim-0.9.4.json#L100

but I found this

[flint pollen]

well neovim reported it wrong

[oak quest]

The neovim-0.9.4.json (obtained from my lazyvim client) said it needed "relatedInformation"

[flint pollen]

it was wrong

[oak quest]

ok

[flint pollen]

also I don't want to be too annoying but in this comment: https://github.com/Myriad-Dreamin/tinymist/issues/1870#issuecomment-3035415536 I argue that even with future neovim (which looks like it will maybe support relatedInformation, in neovim 0.12), using relatedInformation will still not solve the issue

I'm just looking at their proposed change and seeing that it does nothing for the cross file scenario

I am too annoying now so I think I should either make suggested PRs for tinymist or be silent

[oak quest]

I read rust-analyzer, and find they split diagnostic:

the analyzer reports a single rust diag with relatedInformation like

RustDiagnostic {
  message: primary_message,
  relatedInformation: [msg1, msg2]
}

It will be split into three lsp diag with relatedInformation:

// primary
LspDiagnostic {
  message: primary_message,
  relatedInformation: [sub1 is here, sub2 is here]
}
// sub1
LspDiagnostic {
  message: msg1,
  relatedInformation: [original is here]
}
// sub2
LspDiagnostic {
  message: msg2,
  relatedInformation: [original is here]
}

[flint pollen]

nice

[oak quest]

I haven't go in depth, and I might be wrong.

https://github.com/rust-lang/rust-analyzer/blob/3816d0ae53c19fe75532a8b41d8c546d94246b53/crates/rust-analyzer/src/diagnostics/to_proto.rs#L285

[flint pollen]

I'm suspecting it's necessary to split the diagnostic, also so that invalidation/refresh of it works correctly.

[oak quest]

I don't think r-a is invalidating diagnostics incorrectly. So this is an enhancement that requires more investigation before fixing it, because we might make a PR that make both vscode and neovim feel good, but after weeks a helix user reports diagnostic issue again. And this is the reason I didn't reply on the issue#1870 quickly.

btw today is my workday, and I'm chatting with you when I'm working 😄

[flint pollen]

I made a neovim workaround where I turn the relatedInformation into extra diagnostics in the correct file and line. And that one has huge problems with invalidation/refresh. Feels like it clashes absolutely with the LSP flow of information

[oak quest]

so your question should be gone (invalidation were wrong) if the extra diagnostics are produced in the language server. Specifically, tinymist updates the diagnostic state and refreshes necessary diagnostic changes per file in the editor.rs.

[hoary glacier]

Quick question, is it possible to set a "main" file from which tinymist would compile? Because I have my project separated into chapter files, but I only want the main file to be exported.
For now I'm using this script:

#!/bin/bash

source_file="${1-src/Mage Conversion.typ}"
basename="${source_file##*/}"
target_file="${2-target/${basename%.*}.pdf}"

typst w --root src --font-path assets -- "$source_file" "$target_file"

As a workaround, but I was wondering if there could be a better way.

[flint pollen]

check out this section on multi-file projects https://myriad-dreamin.github.io/tinymist/frontend/vscode.html#label-Working with Multiple-File Projects it's about pinning a main file

[hoary glacier]

A-ha, thanks, okay.

Thanks.
Is it possible to make it stateful/static by writing in .idea for the workspace or somewhere, by the way?

[autumn sandal]

is there a way to get tinymist to "unravel" the panic so that line 5 gets underlined red as well when previewing a file?
this is particularly annoying for libraries/templates as i have to go to the file containing the panic() to see what happened

as far as i can tell there's not an gh issue for this, which makes me think this is a problem on my part? Somebody would've made an issue for this, surely?

[rough olive]

There isn't, I guess. You can also see the panic message in the problems view below

[autumn sandal]

oh i basically never use the problems view

[autumn sandal]

i'd be happy to work on adding this if giving some guidance.
Would the changes have to be made in the vscode extension, or in the lsp crate?

[buoyant sand]

Just recently @flint pollen has talked about a similar issue where hints are reported at the call site. I suppose there the lsp would have to add new diagnostics with the spanse of the associated trace

So it's probably a thing for the language server, itself not the extension

[flint pollen]

I reported an issue about a similar problem. https://github.com/Myriad-Dreamin/tinymist/issues/1870 the issue seems a bit specific, that was about what I knew at the time. But I would also prefer to solve the general user interface problem rather than any specific technical issue mentioned in the title or the issue. See a little discussion up in the channel, not long ago

and this feature will help in parallel, it's another place the error will be detailed: https://github.com/Myriad-Dreamin/tinymist/pull/1860 🙂

[oak quest]

copilot will undoubtedly be the second active contributor of tinymist. 😅 https://github.com/Myriad-Dreamin/tinymist/pull/1892

[oak quest]

i run out of copilot quotes after requesting 4 prs. copilot will not listen to me until next month. The plan of accepting copilot as a maintainer goes dead in 6 hours.

[urban grove]

At least it was a thing

[harsh wing]

He doesn't have the FOSS spirit, no money — no workey.

[autumn sandal]

i misinterpreted the title thinking it would render placeholder content for errors, which would be pretty neat
ie:

#table(
  columns: 2,
  [A], [B],
  [#panic("error!")], [C]
)

would still create a document, just with a placeholder in place of the panicky content

this might have some not so good implications tho

i dont think you could figure out how far to unravel the call stack before replacing with placeholder content

[cursive bough]

That's hilarious.

[oak quest]

https://github.com/Myriad-Dreamin/tinymist/pull/1904

To complete/analyze too dynamic code, I added some enhancement.

It will work well with some cheap requests (hover, completion, definition, etc.), but some not with expensive requests (rename, reference, etc.). This is already powerful enough I think.

[marsh lance]

Is there some way to inform Tinymist that I'm working on an HTML project and have it not warn about stuff being ignored in paged export?

[oak quest]

https://myriad-dreamin.github.io/tinymist/config/vscode.html#label-tinymist.exportTarget

[marsh lance]

Thanks! I should read the docs next time 😅

[autumn sandal]

this code looks very interesting lol

anyways, is there a way i can add onto the sys.inputs from an export task in vscode?
why?: I'd like to have two different compilation modes:

• one for the digital document
• one for the printed document. this will omit appendices and other extraneous content in favor of providing a QR code to the digital document

[autumn sandal]

ah i suppose i could use two different files

[harsh storm]

Hi, is there a way to activate scroll synchronization between the editor and the preview (like in the Web App)?

[harsh wing]

Flashback to all the news where a help chat bot for some companies starts hallucinating and then they get a backlash.

[oak quest]

If you don't trust llm, you can verify the "search result" by checking the spec markdown file.

[harsh wing]

yeah, here it's not critical

[patent ibex]

Better than deleting prod database lol

[sturdy surge]

What's the current state of Sublime Text support for Tinymist? Has anyone got browser preview working with click to reverse lookup?

I have set up the Tinymist LSP and can separately run tinymist preview from the command line to see a in-browser preview, but they are not integrated

[harsh storm]

hahahaha, thanks!

I even checked https://marketplace.visualstudio.com/items?itemName=myriad-dreamin.tinymist before asking but bot knows more

😂😂

[oak quest]

You could report the concrete bug if it didn't work. I only see few people use sublime text with typst, and they didn't use the browser preview feature.

[oak quest]

the readme doesn't contain everything. I would like to only keep necessary tips in the readme that needed by people newly coming to typst.

[somber drum]

Is deep seek like the Chinese OpenAI? ChatGPT has like a 90% market share around here, does Deepseek have a similar market share in China and is like "the default" chatbot?

[oak quest]

I also use chatgpt when I have some hard tasks. I use deepseek for easy tasks because it is a free and smart chatbot.

[heady prawn]

Quick question, could tinymist feasibly be used as an LSP inside a swift app?

say like a macos app

or ipad

I suppose maybe by using tinymist-query and writing a bit in swift that conforms to lsp

[oak quest]

I remember swift can run and call c functions, so you could build a rust crate that integrates tinymist as git deps and exposes C APIs. As for mobile env, I know tinymist can run in android for emacs android, but I don't know whether ipad provides some more restricted env for applications preventing tinymist from running.

To run tinymist in rust apps with stable interface and least efforts, I tried a bit with fn test_main. The function accepts the clap arguments and uses fn start_project which starts a language service that watches the file system in background. The whole part is still every internal.

[heady prawn]

Ah also, I'm working on adding Typst support to Nova using tinymist for anyone else who uses nova

https://tangled.sh/@adrianlorenzana.ca/Nova-Typst The repository for it is here, and the extension can be installed now in Nova

Next step is integrating tinymist settings into the extension xd

[gentle bay]

Does anyone have an issue with the export pdf on save functionality?

in particular, i see that the pdf file is not being written to, despite the typ file being saved

[oak quest]

Do you see "Project: skip compilation for {id:?} due to harmless vfs changes" in the log? If it is, I think we hit a bug introduced by fixing some other bug.

[gentle bay]

where would i check to see this?

i notice that it does write to the pdf when i cmd + hover over any function or import file

[oak quest]

nvm, I think I have fixed it.

[heady prawn]

Can I ask how tinymist receives settings? Like say when writing settings to languages.toml for Helix, how does tinymist receive those settings?

Eh here: https://github.com/Myriad-Dreamin/tinymist/tree/main/editors/helix

[oak quest]

could you check https://nightly.link/Myriad-Dreamin/tinymist/workflows/ci/main/tinymist-linux-x64.zip ?
This is built by https://github.com/Myriad-Dreamin/tinymist/actions/runs/16607127510#artifacts

[oak quest]

It receives language settings by the standard lsp. When handling the initialization request and workspace/didChangeConfiguration notifcations, it gets the configurations from the editor. The related code is in config.rs (search it on GitHub)

[tardy sparrow]

I'm using Tinymist for Typst in VSCode, and when I use code completion, it automatically inserts (). Is there an option to change this behavior or disable the automatic parentheses insertion?
For example, when I have a word selected and type #strong, the auto-completion result becomes #strong()[word]. I would like to either disable the () insertion completely, or have it auto-complete to #strong[] instead of #strong().

[oak quest]

We complete function calls with two variants:

• #fn| expands to #fn(|). If the fn is known to accept zero positional arguments, the cursor is moved to right of the parens.
• #fn.bracket| expands to #fn[], if a function accepts exactly a content value first. Thus, the fixed and quick way to get #strong[|] is completing#sobk| (strong.bracket)

The above is a sensible default settings imo, and I haven't tuned any functions.

[tardy sparrow]

Thank you for your response!

[oak quest]

I have figured the remaining steps to run tinymist in browser. https://github.com/Myriad-Dreamin/tinymist/pull/1944

First, I need to split the part of code (this is possible but only kind of cumbersome because I built most code of tinymist in typst.ts). Then I need to reimplement the lsp transports for vscode web's web worker. And at last, I need to update the extension code to bootstrap (I finished it tonight).

I'm interested in bringing tinymist's documentation functions to browser. This will be nice, as we could have a docs site that can check docs typst packages at any revision. https://docs.typ/@github/User/repo/branch-or-rev/path/to/package.

• @github/User/repo maps to the github repository (or other git source) that could be fetch from network
• branch-or-rev means we can check documentation at any revision.
• path/to/package is the path to the package. For most packages, this is empty.

[cursive bough]

Yes, it would be a dream come true, as it will be akin to this experiment: Rust Analyzer in the browser, https://ra-wasm.netlify.app/

[oak quest]

which one will be faster to be production ready?🐱

[cursive bough]

Well, ra-wasm has been abandoned for years, and the repository has been archived officially this year (in May), so...

but personally, while it would be cool to have tinymist in the browser, I think like the rust-analyzer guys.. you might realise that it would not have many users. I want to positive here, but it is just.. very cool, but, who will use it?

Oh! You want to build a docs.typ (kind of like docs.rs).

I get it now!

https://tenor.com/view/slap-gif-4810169

[oak quest]

I recently learnt that there is an issue on miri https://github.com/rust-lang/miri/issues/722

at least there is a working rustc.wasm.

[oak quest]

I still would like to avoid complete with webapp. but I think now the situation is better (if I make tinymist work in browser) when I heard webapp was having many active users.

I used tinymist web in github.dev and it looks really useful. it should be good to at least bring some functions to bring convenience to read code on github.dev

My goal is merely make it convenient to read code on github

[cursive bough]

Right.

But aren't there more important feature to pursue this?

I can't see a "symbol overview" when I explore typst packages in vscode. I don't get to look for symbols in the workspace either, so I always use the find-window to navigate typst packages.

[oak quest]

do you mean document outline or syntax symbol outline? they are both available in current tinymist

[cursive bough]

I get this when exploring the touying package.

[oak quest]

It kinda works but not much. tbh I haven't used this feature and the code is from the typst-lsp

[cursive bough]

Okay.

Well, I sort of use that feature when I have to jump from one definition somewhere to another. Especially with touying. It has internal functions, then templates, then examples.

[vital kayak]

I don't think tinymist competes with webapp either way, they serve sep use cases.
Web app is all about collaboration and sharing of docs.

Having tinymist working on web is a big thing for the Typst language, it reduces the friction to get started

It would also be useful for sharing code snippets and embedding runnable examples

[gentle bay]

sorry, i'm unable to check this at the moment

when is the next release planned?

[oak quest]

every weekend

[jaunty kiln]

@oak quest what do you think about the possibility of having numbering appear as inlay hints?

Like in headings and lists and such

[oak quest]

?r Should be good but hard to analyze it. Example:

#enum.item[1]
+ 2
+ 3

[oak quest]

I had some similar thoughts to implement some typst-specific inlay hints, like showing the labeled content as inlay hints. I think we can explore it if this message is upvoted. I might also create an issue to collect votes.

[jaunty kiln]

Yeah probably not easy, but I think inlay hints could be useful in a lot of situations

[cursive bough]

I think this is a good suggestion actually.. In practice, it might not feel important, but for equations, or long winded explanations.. I sometime check the labels, in the pdf, to see if they make sense, and sometimes, I would have used the wrong label. Sometimes it's easy to discern in the code, but other times, the number itself makes it apparent if something is goofed.

The web app shows a little preview of what the label points to. That's super helpful too..

[cursive bough]

https://github.com/Myriad-Dreamin/tinymist/pull/1964

Is this what I asked about? With the symbols?

[hardy grotto]

I see a weird failure with Tinymist preview. If I try
?r

#set text(size: 50pt, 
  fill: gradient.linear(..color.map.rainbow),
  stroke: 1pt + green
)

Test

I don’t see the stroke in preview (but it’s there in the pdf output).

Can anyone reproduce this?

[flint pollen]

I see the same thing.

[hardy grotto]

Which same thing? Do you see the stroke?

[oak quest]

I can reproduce it, but I might not fix it quickly. My next focus is to fix some old bugs, especially the tinymist.lock bugs that could enhance QoL of neovim users significantly (from the feedback I collected).

[hardy grotto]

Oh I 100% vote for neovim fixes!

[oak quest]

I completed the last bug reported in 2024 and there are 14 bugs left in 2025. I have already set up automated and e2e tests for neovim, and I hope this can ensure stable support. I will add some random improvements to make myself happy (I don't want to see the bug fixes fulfilling the changelog)

[hardy grotto]

I only wanted to check whether I had something broken on my side.

And I’m extremely grateful for your work on all this, don’t worry!

[flint pollen]

I reproduced your issue

[oak quest]

I just forgot it. It would be great if you could form a bug report and submit it on GitHub. You don't have to attach logs, because the code snippet has reproduced the bug.

[flint pollen]

When I work on other projects now (not typst) I miss the sampled values on hover.

[oak quest]

the sampled values on hover are usually called "Inline Values" information. I view it a killing feature but hard to implement in other languages. Debuggers could implement it: when you stop the program and values can be get on hovering. Some special languages may also implement it (nix might be one).

[flint pollen]

yeah it's like a constant live debug

by the way the neovim work you are doing - the canonical client - I'm just wondering about a clarification - what's the use of that, is it relevant for tinymist users at all, will we use it? Or is it just for automatic or interactive testing of tinymist + editor during development?

[oak quest]

As I mentioned in the PR, I would like to add comments and tests to the neovim client. The neovim client may not be really used by users, though I hope it could, but the code must be continuously and sufficiently tested. The extra exposed lsp commands must be heavily documented and help developers plant features in other client (like jetbrains). Some features that I don't want to document in the https://myriad-dreamin.github.io/tinymist/ but the canonical client includes:

• how preview instances are created and maintained by clients.
• how to get extended notifications from the server: compile status, document outline, and so on.
• how to use the features other than previewing: profiling typst compilations or lsp server, get debug information about a typst document like font usages.

The neovim client may not be really used by users, though I hope it could
Since I found many neovim users are satisfied with only having standard lsp features, that provided by nvim-lspconfig. Installing a language-specific plugins seems to be excessive. I think it understandable and I also prefer to implement the standarized features whenever I can.

[wheat garnet]

Some special languages may also implement it (nix might be one).

Iirc, Google sheets does too. Seems natural, as spreadsheets are basically data flow languages without side effects. Very useful indeed, I find myself missing it greatly in Python...

[flint pollen]

Having the pin main command is very useful. When starting I only used lspconfig and typst watch and it's good that it works so well with only that. And typst watch is something I come back to when using/testing development versions of typst anyway (only sometimes)

[hardy grotto]

I created https://github.com/Myriad-Dreamin/tinymist/issues/1981

[hardy grotto]

I would be very happy to install a language specific plugin. I already have typst-preview.nvim

[flint pollen]

neovim also ships with a bit of typst support already built in (indent settings, syntax colors, commentstring and such settings)

the vim world is messy. Three sources for syntax colors (legacy and treesitter and lsp)

[oak quest]

I suddenly have some idea that replacing the pin main command. https://github.com/Myriad-Dreamin/tinymist/issues/1982

The previous pinMain('entry.typ') is as same as pushCompilationHistory(['typst compile entry.typ']). With this I think I can deprecate the pinMain and unpinMain. I don't know whether people will be bothered by not having unpinMain in future.

[flint pollen]

aha

[solemn coral]

I am getting bad server configuration after updating tinymist

bad server configuration: failed to deserialize ""preview": Error("invalid type: boolean `false`, expected a string or an object with image and rest fields", line: 0, column: 0)

[oak quest]

please check the errors in the configuration. there should be one that was not detected previously

[solemn coral]

How do I do that?

[rotund flame]

I've opened a PR for this. https://github.com/Myriad-Dreamin/tinymist/pull/1987

[solemn coral]

Thank you. Changing invertColors fixed it

[rotund flame]

However, it's strange that you got a boolean false while I got a null.

I'm not very sure about VS Code's default behavior

[solemn coral]

It might have been a previous configuration I set

when resetting to the default, it goes to null - and setting it to "auto", "never", or "always" makes the warning go away

[rotund flame]

Ok. That makes sense then.

[solemn coral]

I believe it worked with true/false on previous versions

Also, why does it not use a drop-down on settings editor? I would rather not edit the JSON manually if I don't have too

[crimson blade]

Same here.
I have been looking for a Typst-specific nvim plugin, better with a lot of predefined text objects.
For example, VimTeX provides useful text objects for LaTeX based on syntax analysis, which I found very useful.

[flint pollen]

that's a good idea. What text objects are you thinking of? I've noticed that the function text object (tree sitter based) is not even working in typst, that's not good

(does it work for you, in neovim?)

[hardy grotto]

in my ~/.config/nvim/after/queries/typst/textobjects.scm I currently have

(math) @typst_math.outer
(math (_) @typst_math.inner)
(code _ @typst_function.outer) ; This is needed for the next query to jump back.
(call item: (ident) @typst_function.name (_))

The math is really the most needed one. It allows my to yank formulas.

The function ones are also nice when I want to turn a lemma into a proposition or theorem.

This is assuming you use treesitter.

Of course I also have Typst snippets. This is much less needed than for LaTeX because there is a lot less boilerplate code. But it’s still mandatory if you want to use fletcher for instance.

I can share more details if this is not enough.

[flint pollen]

thanks for sharing! I've never touched treesitter configuration at all, I just use standard configs. Will try this out

[hardy grotto]

You will also need https://github.com/nvim-treesitter/nvim-treesitter-textobjects

[flint pollen]

that one I have

[hardy grotto]

Then in my treesitter config I have

    textobjects = {
      select = {
        enable = true,
        lookahead = true, -- Automatically jump forward to textobj, similar to targets.vim
        keymaps = {
          ['aa'] = '@parameter.outer',
          ['ia'] = '@parameter.inner',
          ['af'] = '@function.outer',
          ['if'] = '@function.inner',
          ['ac'] = '@class.outer',
          ['ic'] = '@class.inner',
          ['i$'] = '@typst_math.inner',
          ['a$'] =  '@typst_math.outer',
          ['fn'] = '@typst_function.name',
        },
      },

as explained at https://github.com/nvim-treesitter/nvim-treesitter-textobjects?tab=readme-ov-file#text-objects-select

With this, I can put my cursor anywhere in a formula and type ya$ to copy the whole formula,including dollars, and yi$ if I don’t want to include the $.

[buoyant sand]

Does neovim not come with delimiter matching?

Ah wait I remember this feature in helix was inspired by easy motion I believe

[flint pollen]

if you mean like matching {} and () and jumping between then, yes it comes with that, % jumps between them. And there are default built in text objects for the regions inside {}, () and so on

[buoyant sand]

Ah but it won't do that for any delimiters defined for a language then?

Just a default set?

[flint pollen]

in other programming languages I have for example yaf to copy the whole current function which doesn't work with typst for some reason

yaf is (yank all function)

[flint pollen]

seems like it.

[hardy grotto]

There are probably plugins that extend this default set. But treesitter is already around, so it’s easy to use it.

[buoyant sand]

Yeah that makes sense

I don't actually know if helix needs the tree sitter queries for this or if it can just do this from the simple language delimiters config

[flint pollen]

what's a delimiter in that sense? So it starts from #let ... and captures until the final } ?

[buoyant sand]

They're configured in the languages config and are used for this and stuff like auto closing

[flint pollen]

this one is the only thing that's working for me 😆 How do you usually debug this, I can't seem to use the typst_function thing at all. How does the af work for you (since it doesn't use typst_function)?

[buoyant sand]

So for math it would be the dollar signs

[flint pollen]

dollar sign matching already works in neovim without change

I regularly use it

[hardy grotto]

Maybe I broke something, let me check it still works.

[buoyant sand]

Ah right you guys don't select things before you yank it

I forgot

[flint pollen]

oh and that means that when I said that's the only thing that works above, that's because it's a default thing

[hardy grotto]

I don’t understand what you mean.

The af has nothing to do with typst, the one about typst is fn.

Sorry I should have removed the bits that have nothing to do with typst.

[flint pollen]

ok, so the af text object does not select anything in typst for you either

that's basically one of my questions

[hardy grotto]

Do you see the tree with InspecTree?

[flint pollen]

yes

[hardy grotto]

The way I use fn is to swith between different function names. Say I have

#lemma[
  $1 + 1 = 2$.
]

and I decide I want to upgrade this to a theorem. I have my cursor anywhere in the statement and I type cfn (for “change function”) and the “lemma” gets deleted and my cursor is there.

[flint pollen]

nice

ok, I have no way to debug what's going on so I'll have to slowly read more about this

[hardy grotto]

I’d be really happy to help, but my family is calling me.

sorry

[gentle bay]

seems like problem persists in 0.13.18

[2025-08-04T19:47:45Z INFO  sync_ls::server] handled  textDocument/formatting - (939) in 498.67µs
[2025-08-04T19:47:45Z INFO  sync_ls::server] notifying textDocument/didSave
[2025-08-04T19:47:45Z INFO  sync_ls::server] notify textDocument/didSave succeeded in 12.71µs
[2025-08-04T19:47:47Z INFO  sync_ls::server] handling workspace/executeCommand - (940)

[flint pollen]

I made something, this is the first working definition of @function.outer for typst in treesitter. (Sorry for all the neovim/treesitter talk..)
(let pattern: (call _) value: _ @function.inner) @function.outer
Only thing is I'd like to include the # before the let if possible I think. If it exists

[hardy grotto]

I’m back.