can compile both the react application statically so it can be served by
nginx
and then builds a docker container with nginx for easy deployment to
something like fly.io
I mean since it's static I could probably host it for free on a website
like netlify or something, but I still wanted a docker container for
now.
Notably I did have to move the generated javascript parser into the
mast-react-vite folder so that `buildNpmPackage` was scoped to the
mast-react-vite folder. I'm not sure how to get it working in a truly
monorepo style build working for it yet, but I'm not exactly inclined to
make it super shiny right now either.
basically at MVP now
the react app can now add, complete and reactively filter tasks
The filter command doesn't have a enter action, it should probably do
something. maybe I do need the default to be select, instead of filter,
and filter should _actually filter_.
Syncing is next. I'll need to look at the vcln starter template to
understand how they set up, and then maintain different databases for
each user.
What needs to be done is setting up rooms, and the backend such that
everybody gets their own lil sql database, and the syncer can identify
them. we don't really have any notion of security, which is uhhh, not
great.
After that we can look at some malleable features that lend themself to
discoverability. ideally, I don't need to have a tutorial at all.
reactivity needs to exist for almost every command. I want to save the
command, to a dirty state sql table, and then restore it on load. that
should cause all the elements to re-render the proposed change.
this actually unlocks a thought for me: executing filters, sets them as
a "context". a context makes up a view. reports are literally just
saved contexts then. this allows the tanstack table to handle virtual
ids for us!
Then, for the malleable input, all we need some kind of triple scroll
wheel that can be typed, for each of the three segments. ideally with
desired autocomplete 🤔
i think i'm reimagining the CLI for the web almost!
It's not fully hooked up to the react app, and it's not quite complete
either, but it's a start.
Currently, it will parse filter implicitly, and there's also the done
and add command. For now, all commands must place the command name in
the right location.
Additionally, I've hooked up the react table's selection state to the
command parser. Filter based on _ids_ explicitly does work. Manually
selecting an ID will also place it in the input field for
discoverability. It does not remove ids that were manually deselected
yet, nor does filter based on any other attribute work.
The command parser also parses times, but it does so very poorly
currently. More work will need to go into that, for now it's basically a
stub. I'll also need to convert the parser to Go.
The react table conviently handles virtual tables based on it's internal
`row.id`. I'm still wondering how to implement virtual ids, but at least
on the web app, they are constantly updated since the view is always
visible, which is different to the CLI (though perhaps the same as a
TUI?)
Notably, there's no semantics for the difference between selection and
filtering. I think perhaps the implicit filter command should default to
selection, but the explicit filter command should be handled
differently, and actually execute filtering. I haven't yet looked into
how the TanStack Table handles filtering, but I am aware that it does.
I think for an MVP, once done is implemented, I can deploy this, the web
socket server, and try and start using it.
colors...
I stole a bunch of things from
[this](https://github.com/danielagg/shadcn-data-table-resize-columns)
repository I found online, and that seems to have helped.
I'm still having trouble with colors and tailwind and shadcn...
The buttons also only misalign on firefox, not chrome...
