-
Notifications
You must be signed in to change notification settings - Fork 19
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
chore: react-router-dom
migration
#980
base: main
Are you sure you want to change the base?
Conversation
…mpat` - update routes - remove unused code - add `NotFound` page (wip)
- fix type issues - remove redundant variable - fix tests (most of them)
- fix bug in `useSearchQueryParametersState`
- downgrade packages to match grafana `grafanaDependency` version
- use `generateRoutePath` - remove done TODOs - update tests for `ProbeCard` - update tests for `NewProbe` - update tests for `EditProbe` - remove debug content from `TestRouteInfo`
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
self-review ✅
src/page/EditCheck/__tests__/ApiEndPointChecks/HTTPCheck/Submit.test.tsx
Outdated
Show resolved
Hide resolved
src/page/EditCheck/__tests__/ApiEndPointChecks/TCPCheck/Submit.test.tsx
Outdated
Show resolved
Hide resolved
- update not found usage for `EditProbe`
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is so good and is paying off a ton of debt. I didn't even find the 79 files that daunting because of how much good work it was doing 🙏🙏🙏
Main asks are a few of the mundane dev tasks: tests and documentation.
The only thing I didn't get a chance to test before I wrap up for the week (and go on PTO for another week) is what happens with all the existing routes and if the redirects are handled correctly. I saw some legacy redirect logic but didn't get a chance to test it.
package.json
Outdated
"@grafana/scenes": "5.1.0", | ||
"@grafana/schema": "11.0.0", | ||
"@grafana/ui": "11.0.0", | ||
"@grafana/ui": "^11.0.0", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd make the argument we should go the other way and get rid of all of the caret flag syntax in package.json
and be explicit so this acts more closely as the source of truth rather than having to grep search the yarn.lock
file to check what version is actually being used. (@grafana/ui 11.3.0 in this case).
Also, in this case we have to update the grafanaDependency
in plugin.json
to >=11.3.0
to be on the safe side (and remember to check if we should bump it every time we do yarn install
).
I also don't trust developers enough to respect npm semvar... (I mean just look at our plugin releases).
Happy to debate if you think we should keep caret syntax, though.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unfortunately, 11.3.0
is required for the tests not to blow up.
So, unless it's absolutely crucial that we stay on an older version of @grafana/ui
, I'd rather bump the dep version to 11.3.0
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sorry, I should have been clearer. I am all for updating the packages -- little and often is great and if we can continually update the Grafana packages especially so they are as close to the runtime as possible that's great.
My concern is in utilising the ^
caret syntax for the dependencies. I'm not a fan of it because package.json
becomes opaque and you have to delve into yarn.lock
to know what you are really using. It also assumes that every package you utilise accurately follows semver and I have trust issues after this being a source of bugs in other projects I have worked on previously 😅
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I can see why pinning @grafana/*
versions makes sense (even though it doesn't mean that the code will run with that version in the wild), but pinning "other" packages for no other reason than "what if..." seems like a false sense of security. Many of our dependencies have their own dependencies that will install minor
and patch
, meaning that we have no actual control over potential side-effects by installing a minor/patch.
Also, when two (including our package) or more packages rely on the same major version, they may share the same package instead of installing their specific version.
Note: is bundleGrafanaUI: false
something we want to use? I think it ties into this.
TL;DR: I could live with (but, wouldn't recommend) pinned versions with ~
(so that patches are installed).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Alright, let's agree to pin the runtime @grafana
packages and update the plugin.json grafanaDependency to match.
We haven't encountered any problems with the caret syntax and you are right, it does help manage the dependency tree better.
Note: is bundleGrafanaUI: false something we want to use? I think it ties into this.
No, I don't think we do. It just creates a different set of problems and goes against the grain of what everyone else is doing. Because we have SLOs with CVEs now we should naturally get a pretty good rhythm of keeping our dependencies updated.
@@ -31,9 +32,11 @@ export const ProbeEditor = ({ | |||
probe, | |||
submitText, | |||
supportingContent, | |||
readOnly, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we add a few tests as a safety and implicit documentation for the readOnly
prop?
My worry is reading this in the future it is confusing there is a readOnly
prop and also a useCanEditProbe
hook so it looks like something is redundant and can be removed.
I think much of the confusion stems from the component names. Should we make a ViewProbe
component to add to the router? Maybe we can tackle that when we do #848?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
- updated the prop name
forceViewMode
- added test for the new prop
This is an intermediate solution for the lack of a ViewProbe
component.
path={route} | ||
element={ | ||
<> | ||
<TestRouteInfo /> |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we document what this is doing, please. It took me a while to work out its purpose and had to look at some of the tests. Would help me out on getting started on this ticket, too: #875 🙏
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This looks really great 🎉🎉🎉
Thank you for putting the time into this! I know fixing the tests wasn’t trivial 🙏
I’ve left a few comments and questions, but overall, everything looks fantastic. I tried to make the routes break, but I wasn't successful 💯
const { checkTypeGroup } = useParams<CheckFormPageParams>(); | ||
|
||
if (check) { | ||
const checkType = getCheckType(check.settings); | ||
return CHECK_TYPE_OPTIONS.find(({ value }) => value === checkType)?.group; | ||
} | ||
|
||
return checkTypeGroup; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
const { checkTypeGroup } = useParams<CheckFormPageParams>(); | |
if (check) { | |
const checkType = getCheckType(check.settings); | |
return CHECK_TYPE_OPTIONS.find(({ value }) => value === checkType)?.group; | |
} | |
return checkTypeGroup; | |
if (check) { | |
const checkType = getCheckType(check.settings); | |
return CHECK_TYPE_OPTIONS.find(({ value }) => value === checkType)?.group; | |
} | |
const { checkTypeGroup } = useParams<CheckFormPageParams>(); | |
return checkTypeGroup; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@VikaCep are you we want to do this?
We'd have to add // eslint-disable-next-line react-hooks/rules-of-hooks
before the conditional hook.
ESLint: React Hook "useParams" is called conditionally. React Hooks must be called in the exact same order in every component render. Did you accidentally call a React Hook after an early return?(react-hooks/ rules-of-hooks)
import { CHECK_TYPE_GROUP_OPTIONS } from 'hooks/useCheckTypeGroupOptions'; | ||
import { CheckForm } from 'components/CheckForm/CheckForm'; | ||
import { getRoute } from 'components/Routing.utils'; | ||
|
||
import { PluginPageNotFound } from '../NotFound/NotFound'; | ||
|
||
export const NewCheck = () => { | ||
const { checkTypeGroup } = useParams<CheckFormPageParams>(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should we use const checkTypeGroup = useFormCheckTypeGroup();
here instead?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
IMO, using useFormCheckTypeGroup
doesn't add anything. The use of useParams
is a fallback mechanism in useFormCheckTypeGroup
(for when check
is undefined
).
<QueryClientProvider client={getQueryClient()}> | ||
<MetaContextProvider meta={{ ...SM_META, ...meta }}> | ||
<FeatureFlagProvider> | ||
<SMDatasourceProvider> | ||
<PermissionsContextProvider> | ||
<MemoryRouter initialEntries={history.entries}> | ||
<CompatRouter> | ||
<Routes> | ||
<Route path={PLUGIN_URL_PATH}> | ||
<Route path="*" element={children} /> | ||
</Route> | ||
</Routes> | ||
</CompatRouter> | ||
</MemoryRouter> | ||
</PermissionsContextProvider> | ||
</SMDatasourceProvider> | ||
</FeatureFlagProvider> | ||
</MetaContextProvider> | ||
</QueryClientProvider> | ||
); | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
extensions: { | ||
exposedComponents: [], | ||
}, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just wondering, but why was this needed?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Updated types require this.
- lock grafana dependencies on 11.3.0 - add 404 page instead of redirect to Home
# Conflicts: # package.json # yarn.lock
What this PR does / why we need it:
Update code base to use
react-router-dom@v6
APIs withreact-router-dom-v5-compat
.v5 is obsolete and Grafana now bundles with v6 (opt-in, actually
react-router-dom-v5-compat
).With these changes, SM removes a lot of code debt tied to
react-router-dom
.This PR also updates routing with new paths and some and a
NotFound
component.Route changes
From
/checks/edit/<id>/<checkTypeGroup>
to/checks/<id>/edit
From
/checks/<id>/dashboard
to/checks/<id>
From
/probes/edit/<id>
to/probes/<id>/edit
(edit) |/probes/<id>
(view)Not found page
/checks/2230/not-a-valid-route
Note: An invalid check ID in the route will keep the old (redirect) behavior (for now).
/checks/new/not-a-valid-check-type-or-group
/probes/<invalid-id>
View mode for private probes
There is now a View mode for the private probe edit page
Those who can edit can click the edit button to switch to edit mode. Those without sufficient rights to edit, will not see the button.
Breadcrumbs
Attempts have been made to improve breadcrumbs (mainly checks and probes routes).
Which issue(s) this PR fixes:
Resolves: #854
Resolves: #975
Special notes for your reviewer:
Getting tests to work again was a battle. Pay extra attention to strange test setup etc (very likely there are some exploratory code in the code base).
Try to make the routes break! 🙇🏻