So we can act upon the eventual parents of a commit, after fetching a
single commit we need to have the parents of that commit as part of the
json payload.
Signed-off-by: Sebastian Martinez <me@sebastinez.dev>
It turns out that the CRDT formed by the union of Git DAGs
is enough to guarantee everything we need for COBs.
This changes the following things:
* COB operations no longer need to be commutative
* COB histories are traversed in the same deterministic order on all
replicas
* It's now possible to implement RSMs on top of COBs, eg. scripting
* Lamport clocks have been removed
* `radicle-crdt` is no longer a dependency of `radicle`
* COBs are no longer instances of `Semilattice`
* The `Ops` type was removed in favor of having `Op` contain multiple
actions
This improves response time of:
- GET /api/v1/projects/<RID>/tree/<OID>/ from 205.71ms down to 376.4µs
- GET /api/v1/projects/<RID>/tree/<OID>/.nix from 111.06ms down to 351.70µs
Signed-off-by: xphoniex <dj.2dixx@gmail.com>
Instead of using always the repo HEAD for the commit count with
`repo::stats` we should use `repo::stats_from` and pass the `parent`
commit.
This way we allow e.g. a commit history for a peer with lesser commits
to be counted correctly.
Signed-off-by: Sebastian Martinez <me@sebastinez.dev>
We prepare the data for `PeerSelector` and `BranchSelector` in the
`History` and `Browser` instead of `SourceBrowsingHeader`. This makes
the component APIs overall simpler and allows us to remove
`LoadedSourceBrowsingView`.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
There’s no need for the outer `if branches.length > 0`. We already
disable the dropdown unless `branches.length > 1`.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
We had a regression because of introducing marked-linkify-it, which took
links for CONTRIBUTING.md or HACKING.md as domains of a new page instead
than files to be navigating to.
This commit sets the `fuzzyLink` option for marked-linkify-it to false,
which disables this behavior
Also tests that the browser tree expands itself when navigating to a
file inside a directory
Signed-off-by: Sebastian Martinez <me@sebastinez.dev>
We were always calling `replace` and not passing the desired action in
the `navigateInternalOnAnchor`, this was leading to the URL not updating
when e.g. clicking on file URLs in Markdown renders.
By passing the action to the `navigate` fn instead we can also take the
`push` action into account.
Test have been written to guarantee this works
Signed-off-by: Sebastian Martinez <me@sebastinez.dev>
For a single commit view the source browsing header (peer selector,
branch selector, commits, contributor count) does not contain meaningful
information. We remove the header and simply show the commit.
See #938.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
We greatly reduce the numbers of properties that need to be passed to
`BranchSelector` and `PeerSelector` by constructing routes in
`SourceBrowsingHeader`. We’ll lift the logic out of
`SourceBrowsingHeader` in a follow-up.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
The `loadProjectRoute` function has become too big too understand. To
address this, we extract and separate the loading for the `tree` and
`history` views. This also gives us the opportunity to simplify some of
the code and parallelize more HTTP requests.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
Based on a new addition to marked, this commits replaces the imported
marked namespaces with a Marked class, that doesn't change when
`marked.use` introduces new options or extensions.
Also changes the naming slighty to use `markdown` as module, so it
sounds more generic.
Reference: https://github.com/markedjs/marked/pull/2831
Signed-off-by: Sebastian Martinez <me@sebastinez.dev>
We switch the test for `pathToRoute` from a paramterized test to regular
tests. I had a lot of trouble finding the right properties to change
when a test in this group failed. The line number was not helpful at all
and the test description was burried inside an object.
We also rename `pathToRoute` to `urlToRoute` because the function takes
a `URL` as an argument.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
We switch to a project route type that is just a union. This has the
following advantages
* It’s much easier to construct project routes. Fewer properties are
needed and there are no nested objects.
* We eliminate the implicitly shared properties `path`, `peer`,
`revision` and `route` which only belonged to the tree route.
* The structure of the project route is more regular and thus easier to
understand.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
We determine URLs to mock with a URL matcher instead of mocking every
URL and determining the behavior in the route handler. This should be
slightly faster and should result in better debugging experience because
we don’t come across the `route.continue()` event that often.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
This improves UX making it easier to find the tab that I'm looking for
when I have multiple tabs open.
Signed-off-by: Rūdolfs Ošiņš <rudolfs@osins.org>
We simplify some of the choices for `ProjectLoadedView` by reducing the
nesting and removing redundant properties.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>
To avoid the popping up of a terminal input field during the
globalFixture setup, we need to pass the wanted alias as a flag.
Signed-off-by: Sebastian Martinez <me@sebastinez.dev>
We move the `peer` and `revision` properties from the shared
`ProjectLoadedParams` interface to the `tree` and `history` subroutes
where they are needed.
Signed-off-by: Thomas Scholtes <geigerzaehler@axiom.fm>