This reduces the number of things we have to update in our token parser
and serializer. For payloads, we we have to handle the payload cases
different, but we now have a structure that can deal with that
efficiently.
Signed-off-by: William Casarin <jb55@jb55.com>
In an attempt to make our deck serializer more localized,
comprehensible, and less error-prone, we introduce a new parser
combinator based around string tokens.
This replaces the Selection-based intermediary types so that we have a
more direct serialization style.
This splits notedeck into:
- notedeck
- notedeck_chrome
- notedeck_columns
The `notedeck` crate is the library that `notedeck_chrome` and
`notedeck_columns`, use. It contains common functionality related to
notedeck apps such as the NoteCache, ImageCache, etc.
The `notedeck_chrome` crate is the binary and ui chrome. It is
responsible for managing themes, user accounts, signing, data paths,
nostrdb, image caches etc. It will eventually have its own ui which has
yet to be determined. For now it just manages the browser data, which
is passed to apps via a new struct called `AppContext`.
`notedeck_columns` is our columns app, with less responsibility now that
more things are handled by `notedeck_chrome`
There is still much work left to do before this is a proper browser:
- process isolation
- sandboxing
- etc
This is the beginning of a new era! We're just getting started.
Signed-off-by: William Casarin <jb55@jb55.com>
This splits notedeck into crates, separating the browser chrome and
individual apps:
* notedeck: binary file, browser chrome
* notedeck_columns: our columns app
* enostr: same as before
We still need to do more work to cleanly separate the chrome apis
from the app apis. Soon I will create notedeck-notebook to see what
makes sense to be shared between the apps.
Some obvious ones that come to mind:
1. ImageCache
We will likely want to move this to the notedeck crate, as most apps
will want some kind of image cache. In web browsers, web pages do not
need to worry about this, so we will likely have to do something similar
2. Ndb
Since NdbRef is threadsafe and Ndb is an Arc<NdbRef>, it can be safely
copied to each app. This will simplify things. In the future we might
want to create an abstraction over this? Maybe each app shouldn't have
access to the same database... we assume the data in DBs are all public
anyways, but if we have unwrapped giftwraps that could be a problem.
3. RelayPool / Subscription Manager
The browser should probably maintain these. Then apps can use ken's
high level subscription manager api and not have to worry about
connection pool details
4. Accounts
Accounts and key management should be handled by the chrome. Apps should
only have a simple signer interface.
That's all for now, just something to think about!
Signed-off-by: William Casarin <jb55@jb55.com>
