Skip to content
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

[Security Solution][Detections] Rule forms cleanup #76138

Merged
merged 43 commits into from
Sep 4, 2020

Conversation

rylnd
Copy link
Contributor

@rylnd rylnd commented Aug 27, 2020

Summary

In preparation for new rule types/fields, and in response to improvements to the form library, this adds types to the majority of our form code.

Besides the bugfixes,the main behavioral change here is that it is no longer possible to navigate between invalid tabs on the Rule Edit page, and you must instead make a tab valid before moving away. This is how the Create page currently behaves, and so behavior (and implementation) are now consolidated between the two pages. edit: I've reverted this behavior but changed the invalid callout to display on navigation between tabs (previously it was only shown when you clicked Save. Once we have UX for displaying validation errors to the user on both Create and Edit, we can persist invalid data in a consistent manner.

"Features":

  • adds types our form state/helpers
    • (we had a lot of implicit anys and ass in there)
  • adds regression assertions to an existing "create rule" cypress test (7fa8bca)
  • Removes intermediate state in our form components
    • Summary: the parent holds the data to be persisted, while the form lib holds the data as the user interacts with the form; most of the local state was there to work around issues with the form lib (e.g. memoization, which is now fixed)
    • We persist form hooks (RuleStepsFormHooks) in the parent so that we can validate and store the active step's data
    • We persist VALID form data (RuleStepsFormData) in the parent in order to repopulate a form that has previously unmounted
      • If a form is unmounted or a form's field is unmounted, the form hook loses knowledge of the form/field
  • Removes uses of FormDataProvider
    • These were used to react to form changes, and typically used local state to track the old value to compare.
    • In (almost) all of our current uses, we only care about current and default values, never the "previous value," and so we were able to reimplement the same functionality with useFormData and useEffect.

"Bugfixes":

  • Fixes moving away from the Actions step (b174270)
  • Fixes bug moving between steps of the Edit form (4995e81)
    • Consequence here is that if a user navigates away from an invalid step, their modifications will be lost
  • Fixes "link" between default severity and default risk score, where changing the former updated the latter (0972e16)
  • Removes some unnecessary ML API calls (aac72c9)

Unaddressed issues/notes

Checklist

For maintainers

Digging through the git history, it looks like this was replaced with
the isUpdateView prop at some point. There's a small chance that we're
indirectly leveraging the effect of this value being changed, but I
think we're safe.
We have lots of anys and unknowns in here, this is my attempt to fix
that.

I started by defining better types for the state/refs in our parent
component; everything else mostly flowed out of that:

* Step components now type their form hook for their step's data
* Removes lots of unneeded `as` casts
* Renames uses of `accordionId` with `step` and `activeStep`, since they
  are also values of our `RuleStep` enum
* Step components now export their default values
  * The data flow is simpler when the parent passes these values in,
    rather than trying to merge props with some internal defaults.
  * The internal defaulting is still there, but I think it'll be
    unnecessary once I audit the `edit` forms.

I've only done this work for the Define step for now, the rest are next.
Now that the create step is passing in the default values, we no longer
need to merge with internal state.

The one exception is the default indexes; since we need that data for
our "reset to default indexes" behavior, we'll keep that functionality
within our DefineStep component.
We don't gain much by forcing the parent to pass in default values. The
slightly cleaner types are not worth the burden to the parent; instead,
we add a type guard to be used in our parent to ensure our values are
present before working with them. Previously, we were circumventing this
logic with an `as` cast.
These are arrays of strings, so a shallow comparison should suffice
here. Also reorders conditions to short-circuit on simple booleans
first.
There's currently a bug on master where returning to a previous form
step does not populate its previous values.

After some investigation it appears that this is due to form values
being reset on submission (form.reset()). Previously, we kept a separate
copy of data in each step's state, and had a useEffect that would
repopulate the form's values if they ever became out of sync. Once that
was removed I believe this bug was introduced.

For now the fix is effectively reimplementing the above behavior, albeit
a little more elegantly with `reset()`. If we can restructure the form
logic to only require the form data at the end (big if), then we can
remove the need to "repopulate" these values to the form.

For now, this does remove the local copy of data in the step component
as I believe that is no longer needed. Data is stored in the parent,
copied/modified to the form, and pushed back up when one clicks on to
the next step.
The linter was complaining because it didn't think that `typedUseKibana`
was a hook. But it is, and we should name it as such.
Things still aren't quite working, state gets lost when moving through
steps but I believe this is addressed in an outstanding PR so I'm not
sweating it right now.

* Removes as much state in Step components as possible
  * We shouldn't need this as the form holds all the state as well. If
    we need to "watch" for a change, we can subscribe to the form's
    observable to replace FormDataProvider and local state (TODO)
* Removes setting of default values in form components
  * I believe that this is redundant with defaultValues provided to
    useForm, but I need to verify.
* Removes redundant uses of field's defaultValue
  * Most are redundant with the form's defaultValue; added calls to
    field.setValue in cases where the default is generated internally
* Removes calls to reset() after submitting
  * These were needed due to a bug in the form lib; once elastic#75796 is
    merged these will no longer be necessary.
This exists identically earlier in the component; I'm guessing it was
the result of a bad merge conflict.
* Makes data structures more similar to rule creation form
* Adds type guards for asserting which step is "active"
* Simplifies logic around the active tabe/step/form
* Removes use of wait() in favor of act()
* Fixes mock call assertion now that we're no longer setting our data to
  null (which was a now-unnecessary form lib workaround).
@rylnd rylnd added Team:SIEM v8.0.0 release_note:skip Skip the PR/issue when compiling release notes v7.10.0 labels Aug 27, 2020
@rylnd rylnd self-assigned this Aug 27, 2020
We never send our actions data back down to the actions form, so it was
lost if you went to a previous step.

Since the actions UI still had any connectors you created, you merely
had to reselect the throttle and the connector, but this prevents you
from having to do that.
Asserts that our rule form repopulates with the provided values when
going back to a previous step. This is to cover a regression that was
not caught by CI (but which has now been fixed).
* Validation and data collection are performed in the parent, not the
  step component
* Step component provides a form ref and notifies the parent when it's
  being submitted; the rest is the parent's responsibility
* Renames some internal helper functions to be more declarative:
  submitStep, editStep, etc.
If the active step form is invalid we will receive no data, so we must
not persist that lest the form blows up on absent values when we later
navigate back.
These exercise functionality that was moved into the parent, so they
need a new home.
@rylnd
Copy link
Contributor Author

rylnd commented Sep 1, 2020

@patrykkopycinski @sebelga I'm seeing some odd behavior, described in the first "Unaddressed issues/notes" above. I haven't yet determined whether it's a bug with the form lib or merely with our forms, but it's present on both this branch and master. If either of you has a moment to 👀 that issue and provide some guidance it would be much appreciated.

@sebelga
Copy link
Contributor

sebelga commented Sep 1, 2020

@rylnd Having a look...

@sebelga
Copy link
Contributor

sebelga commented Sep 1, 2020

@rylnd I saw what it is. I introduced a regression in my last PR and will revert it asap! Sorry for this and thanks for raising it 👍

@rylnd rylnd marked this pull request as ready for review September 3, 2020 17:04
@rylnd rylnd requested review from a team as code owners September 3, 2020 17:04
@elasticmachine
Copy link
Contributor

Pinging @elastic/siem (Team:SIEM)

* Form hooks now _always_ return the form's data, regardless of validity
* Edit Rule now:
  * persists invalid data
  * submits both the active form and the destination form on navigation.
    This is necessary to refresh validations on the destination form,
    since the form lib assumes a newly-mounted form is valid
* simplifies "invalid tab" logic to be derived from our persisted data
If the rule is immutable, they can only edit actions.
@spong
Copy link
Member

spong commented Sep 4, 2020

No need to address here (so much goodness already! :), but wanted to comment on the standardization of the disabled/loading state of fields. Looks like we're closer than before, but there are still a few fields that don't disable when loading/saving. This may be particularly hard to get the Actions component to match since we're pulling in the alerting component full sail, but we should be able to get the others to match since we own those components. In the screenshots below, the arrows point to the fields that don't disable on load/save:

Definition

About

Schedule (As above, probably not much we can do here unless the alerting component supports the loading/disabled state)

Actions

@@ -9,7 +9,7 @@ import { EuiHorizontalRule, EuiFlexGroup, EuiFlexItem, EuiButton } from '@elasti
import * as RuleI18n from '../../../pages/detection_engine/rules/translations';

interface NextStepProps {
onClick: () => Promise<void>;
onClick: () => void;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice catch!

 Conflicts:
	x-pack/plugins/security_solution/public/detections/pages/detection_engine/rules/create/helpers.ts
	x-pack/plugins/security_solution/public/detections/pages/detection_engine/rules/create/index.tsx
	x-pack/plugins/security_solution/public/detections/pages/detection_engine/rules/edit/index.tsx
Comment on lines +95 to +98
const [{ throttle: formThrottle }] = (useFormData({
form,
watch: ['throttle'],
}) as unknown) as [Partial<ActionsStepRule>];
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These always get a double-take from me when I see them, but understand with the removal of FormDataProvider.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, I've discussed this with @sebelga and I'm hopeful that we can get a generic type for our use case (where form data F is identical to output data T.

Since the default severity is 'low,' these two defaults now coincide.
Copy link
Member

@spong spong left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Checked out, tested locally, and performed code review. So much goodness here @rylnd, fantastic work with the cleanup effort! 🙂 Transforming that tech debt into a tech investment with all the type and control flow enhancements you packed in here. I've left a few nits (mostly about starting to use our common types if we can), but nothing worth holding this PR up. LGTM! 👍 🚀 🎉

@rylnd
Copy link
Contributor Author

rylnd commented Sep 4, 2020

Good catch on the SeverityValue redundancy. Fixed in b24742c.

@kibanamachine
Copy link
Contributor

💚 Build Succeeded

Build metrics

@kbn/optimizer bundle module count

id value diff baseline
securitySolution 1951 +1 1950

async chunks size

id value diff baseline
securitySolution 9.9MB +1.7KB 9.9MB

History

To update your PR or re-run it, just comment with:
@elasticmachine merge upstream

@rylnd rylnd merged commit 1b9030f into elastic:master Sep 4, 2020
@rylnd rylnd deleted the rule_form_cleanup_and_prep branch September 4, 2020 21:18
rylnd added a commit to rylnd/kibana that referenced this pull request Sep 4, 2020
* Remove unused isNew field

Digging through the git history, it looks like this was replaced with
the isUpdateView prop at some point. There's a small chance that we're
indirectly leveraging the effect of this value being changed, but I
think we're safe.

* WIP: Making rule form type safe

We have lots of anys and unknowns in here, this is my attempt to fix
that.

I started by defining better types for the state/refs in our parent
component; everything else mostly flowed out of that:

* Step components now type their form hook for their step's data
* Removes lots of unneeded `as` casts
* Renames uses of `accordionId` with `step` and `activeStep`, since they
  are also values of our `RuleStep` enum
* Step components now export their default values
  * The data flow is simpler when the parent passes these values in,
    rather than trying to merge props with some internal defaults.
  * The internal defaulting is still there, but I think it'll be
    unnecessary once I audit the `edit` forms.

I've only done this work for the Define step for now, the rest are next.

* Make defaultValues a required prop of the define step

Now that the create step is passing in the default values, we no longer
need to merge with internal state.

The one exception is the default indexes; since we need that data for
our "reset to default indexes" behavior, we'll keep that functionality
within our DefineStep component.

* Refactor rule creation forms to not require default values

We don't gain much by forcing the parent to pass in default values. The
slightly cleaner types are not worth the burden to the parent; instead,
we add a type guard to be used in our parent to ensure our values are
present before working with them. Previously, we were circumventing this
logic with an `as` cast.

* Remove unnecessary "deep" comparison

These are arrays of strings, so a shallow comparison should suffice
here. Also reorders conditions to short-circuit on simple booleans
first.

* Make StepRuleDescription generic on its schema

* Fixes bug introduced by form lib updates

There's currently a bug on master where returning to a previous form
step does not populate its previous values.

After some investigation it appears that this is due to form values
being reset on submission (form.reset()). Previously, we kept a separate
copy of data in each step's state, and had a useEffect that would
repopulate the form's values if they ever became out of sync. Once that
was removed I believe this bug was introduced.

For now the fix is effectively reimplementing the above behavior, albeit
a little more elegantly with `reset()`. If we can restructure the form
logic to only require the form data at the end (big if), then we can
remove the need to "repopulate" these values to the form.

For now, this does remove the local copy of data in the step component
as I believe that is no longer needed. Data is stored in the parent,
copied/modified to the form, and pushed back up when one clicks on to
the next step.

* Rename typed hook to obviate eslint exception

The linter was complaining because it didn't think that `typedUseKibana`
was a hook. But it is, and we should name it as such.

* WIP: Fixing type errors in the other form steps

Things still aren't quite working, state gets lost when moving through
steps but I believe this is addressed in an outstanding PR so I'm not
sweating it right now.

* Removes as much state in Step components as possible
  * We shouldn't need this as the form holds all the state as well. If
    we need to "watch" for a change, we can subscribe to the form's
    observable to replace FormDataProvider and local state (TODO)
* Removes setting of default values in form components
  * I believe that this is redundant with defaultValues provided to
    useForm, but I need to verify.

* More form cleanup

* Removes redundant uses of field's defaultValue
  * Most are redundant with the form's defaultValue; added calls to
    field.setValue in cases where the default is generated internally
* Removes calls to reset() after submitting
  * These were needed due to a bug in the form lib; once elastic#75796 is
    merged these will no longer be necessary.

* Fix some leftover type errors

* Remove duplicated useEffect hook

This exists identically earlier in the component; I'm guessing it was
the result of a bad merge conflict.

* Fix Rule edit form

* Makes data structures more similar to rule creation form
* Adds type guards for asserting which step is "active"
* Simplifies logic around the active tabe/step/form

* Fixes About Step jest tests

* Removes use of wait() in favor of act()
* Fixes mock call assertion now that we're no longer setting our data to
  null (which was a now-unnecessary form lib workaround).

* Fix bug with going to a previous step after editing actions

We never send our actions data back down to the actions form, so it was
lost if you went to a previous step.

Since the actions UI still had any connectors you created, you merely
had to reselect the throttle and the connector, but this prevents you
from having to do that.

* Add assertions to our rule creation test

Asserts that our rule form repopulates with the provided values when
going back to a previous step. This is to cover a regression that was
not caught by CI (but which has now been fixed).

* Simplify Rule Creation logic

* Validation and data collection are performed in the parent, not the
  step component
* Step component provides a form ref and notifies the parent when it's
  being submitted; the rest is the parent's responsibility
* Renames some internal helper functions to be more declarative:
  submitStep, editStep, etc.

* Don't persist empty form data when leaving a form step

If the active step form is invalid we will receive no data, so we must
not persist that lest the form blows up on absent values when we later
navigate back.

* Skip About Step tests for now

These exercise functionality that was moved into the parent, so they
need a new home.

* Remove unnecessary calls to setValue

* Instead of setting our kibana url after the form is created, we add it
  to the form's default state
* We do not need to set the throttle field value, the field component
  already does this

* Style: logic cleanup

* Prevent users from navigating away from an invalid step on rule edit

We can go against the form lib conventions and persist this invalid data
ourselves on transition, but for now this brings the create/edit forms
into alignment so that adding the aforementioned behavior should be
nearly identical on both.

* Display callout if attempting to navigate away from an invalid tab

We already do this if you try to _submit_ the form, but not when you
click on another tab.

* Persist our form submit() rather than the entire form

Making the entire stateful form a useEffect dependency was likely
causing unnecessary render cycles.

This may also have been part of why both the hooks and the data are refs
instead of normal state; to prevent these rerenders.

* Replace FormDataProvider with useFormData hook

We have to do a type cast here because the hook's data is not typed, but
other than that this is a solid win and cleans things up immensely; the
side effects that result from these values changing are much more
apparent (IMO).

* Move fetch of fields data _after_ form initialization

This ensures that our first fetch of fields will use the index patterns
on the form, not necessarily the default ones.

* Replace FormDataProvider on About step

* This fixes a bug where changing the default severity no longer updated
  the default risk score. It looks like this was broken when the
  severity/riskScore overrides were added, and the values of these
  fields changed from primitives to objects.

* Replace local state with useFormData

By watching the value directly from the form we no longer have a need
for local state, as we were just using it to determine whether the
throttle had changed from the default.

* Types cleanup

Remove some unneeded casts, add some needed ones.

* Rewrite About Step tests

Rather than asserting that the form is invalid through the UI, we
instead retrieve data out of the form hook and assert on that instead.

* Add memoization back to StepRuleDescription

I'm not sure that it's necessary, but best to leave it until we have
time to audit/remove multiple of these.

* Do not fetch ML Jobs if StepRuleDescription is not rendering ML Data

We were incorrectly invoking the useSecurityJobs hook any time the
StepRuleDescription component was rendered, regardless of whether we
actually needed that data.

This moves the useSecurityJobs hook into the component that uses it,
MlJobDescription. If we end up having multiple of these on the page we
can evaluate caching/sharing this data somehow, but for now this
prevents:

* 3x3=9 unnecessary ML calls on the Rule Create page.
* 1x3=3 unnecessary ML calls on Rule Details
* 1x3=3 unnecessary ML Calls on the Rule Edit page.

* Fix bug where revisiting the About step could modify the user's Risk Score

With the severity/risk score link back in place, there was a bug where a
user who had previously set a manual risk score would have it rewritten
on edit (or edit during rule creation).

This was due to a poorly-written useEffect that basically said "if there
is a severity, set a risk score." This has now been amended to say "if
the user changes the severity, set a risk score."

* Clean up About Step tests

* We don't need act(), it's not doing anything.
* We don't need to click Continue since we're just talking to the form
  hook

* Fix local form data when form isn't mounted

If the form isn't on the page (e.g. if we're read-only), then
useFormData will return no values. In these cases, we can simply fall
back to the initialState values, as they'll either be: the default
values on a new form, or: the current values on an active create/edit
form.

Updates the manual type of useFormData to reflect this "maybe" fact.

* Allow user to navigate between invalid tabs on Edit Rule

* Form hooks now _always_ return the form's data, regardless of validity
* Edit Rule now:
  * persists invalid data
  * submits both the active form and the destination form on navigation.
    This is necessary to refresh validations on the destination form,
    since the form lib assumes a newly-mounted form is valid
* simplifies "invalid tab" logic to be derived from our persisted data

* Fix logical error

If the rule is immutable, they can only edit actions.

* Remove unneeded eslint exception

Fixed by upstream elastic#76471

* Make 21 the default risk score for a new rule

Since the default severity is 'low,' these two defaults now coincide.

* Remove duplicated type in favor of common one
rylnd added a commit that referenced this pull request Sep 5, 2020
* Remove unused isNew field

Digging through the git history, it looks like this was replaced with
the isUpdateView prop at some point. There's a small chance that we're
indirectly leveraging the effect of this value being changed, but I
think we're safe.

* WIP: Making rule form type safe

We have lots of anys and unknowns in here, this is my attempt to fix
that.

I started by defining better types for the state/refs in our parent
component; everything else mostly flowed out of that:

* Step components now type their form hook for their step's data
* Removes lots of unneeded `as` casts
* Renames uses of `accordionId` with `step` and `activeStep`, since they
  are also values of our `RuleStep` enum
* Step components now export their default values
  * The data flow is simpler when the parent passes these values in,
    rather than trying to merge props with some internal defaults.
  * The internal defaulting is still there, but I think it'll be
    unnecessary once I audit the `edit` forms.

I've only done this work for the Define step for now, the rest are next.

* Make defaultValues a required prop of the define step

Now that the create step is passing in the default values, we no longer
need to merge with internal state.

The one exception is the default indexes; since we need that data for
our "reset to default indexes" behavior, we'll keep that functionality
within our DefineStep component.

* Refactor rule creation forms to not require default values

We don't gain much by forcing the parent to pass in default values. The
slightly cleaner types are not worth the burden to the parent; instead,
we add a type guard to be used in our parent to ensure our values are
present before working with them. Previously, we were circumventing this
logic with an `as` cast.

* Remove unnecessary "deep" comparison

These are arrays of strings, so a shallow comparison should suffice
here. Also reorders conditions to short-circuit on simple booleans
first.

* Make StepRuleDescription generic on its schema

* Fixes bug introduced by form lib updates

There's currently a bug on master where returning to a previous form
step does not populate its previous values.

After some investigation it appears that this is due to form values
being reset on submission (form.reset()). Previously, we kept a separate
copy of data in each step's state, and had a useEffect that would
repopulate the form's values if they ever became out of sync. Once that
was removed I believe this bug was introduced.

For now the fix is effectively reimplementing the above behavior, albeit
a little more elegantly with `reset()`. If we can restructure the form
logic to only require the form data at the end (big if), then we can
remove the need to "repopulate" these values to the form.

For now, this does remove the local copy of data in the step component
as I believe that is no longer needed. Data is stored in the parent,
copied/modified to the form, and pushed back up when one clicks on to
the next step.

* Rename typed hook to obviate eslint exception

The linter was complaining because it didn't think that `typedUseKibana`
was a hook. But it is, and we should name it as such.

* WIP: Fixing type errors in the other form steps

Things still aren't quite working, state gets lost when moving through
steps but I believe this is addressed in an outstanding PR so I'm not
sweating it right now.

* Removes as much state in Step components as possible
  * We shouldn't need this as the form holds all the state as well. If
    we need to "watch" for a change, we can subscribe to the form's
    observable to replace FormDataProvider and local state (TODO)
* Removes setting of default values in form components
  * I believe that this is redundant with defaultValues provided to
    useForm, but I need to verify.

* More form cleanup

* Removes redundant uses of field's defaultValue
  * Most are redundant with the form's defaultValue; added calls to
    field.setValue in cases where the default is generated internally
* Removes calls to reset() after submitting
  * These were needed due to a bug in the form lib; once #75796 is
    merged these will no longer be necessary.

* Fix some leftover type errors

* Remove duplicated useEffect hook

This exists identically earlier in the component; I'm guessing it was
the result of a bad merge conflict.

* Fix Rule edit form

* Makes data structures more similar to rule creation form
* Adds type guards for asserting which step is "active"
* Simplifies logic around the active tabe/step/form

* Fixes About Step jest tests

* Removes use of wait() in favor of act()
* Fixes mock call assertion now that we're no longer setting our data to
  null (which was a now-unnecessary form lib workaround).

* Fix bug with going to a previous step after editing actions

We never send our actions data back down to the actions form, so it was
lost if you went to a previous step.

Since the actions UI still had any connectors you created, you merely
had to reselect the throttle and the connector, but this prevents you
from having to do that.

* Add assertions to our rule creation test

Asserts that our rule form repopulates with the provided values when
going back to a previous step. This is to cover a regression that was
not caught by CI (but which has now been fixed).

* Simplify Rule Creation logic

* Validation and data collection are performed in the parent, not the
  step component
* Step component provides a form ref and notifies the parent when it's
  being submitted; the rest is the parent's responsibility
* Renames some internal helper functions to be more declarative:
  submitStep, editStep, etc.

* Don't persist empty form data when leaving a form step

If the active step form is invalid we will receive no data, so we must
not persist that lest the form blows up on absent values when we later
navigate back.

* Skip About Step tests for now

These exercise functionality that was moved into the parent, so they
need a new home.

* Remove unnecessary calls to setValue

* Instead of setting our kibana url after the form is created, we add it
  to the form's default state
* We do not need to set the throttle field value, the field component
  already does this

* Style: logic cleanup

* Prevent users from navigating away from an invalid step on rule edit

We can go against the form lib conventions and persist this invalid data
ourselves on transition, but for now this brings the create/edit forms
into alignment so that adding the aforementioned behavior should be
nearly identical on both.

* Display callout if attempting to navigate away from an invalid tab

We already do this if you try to _submit_ the form, but not when you
click on another tab.

* Persist our form submit() rather than the entire form

Making the entire stateful form a useEffect dependency was likely
causing unnecessary render cycles.

This may also have been part of why both the hooks and the data are refs
instead of normal state; to prevent these rerenders.

* Replace FormDataProvider with useFormData hook

We have to do a type cast here because the hook's data is not typed, but
other than that this is a solid win and cleans things up immensely; the
side effects that result from these values changing are much more
apparent (IMO).

* Move fetch of fields data _after_ form initialization

This ensures that our first fetch of fields will use the index patterns
on the form, not necessarily the default ones.

* Replace FormDataProvider on About step

* This fixes a bug where changing the default severity no longer updated
  the default risk score. It looks like this was broken when the
  severity/riskScore overrides were added, and the values of these
  fields changed from primitives to objects.

* Replace local state with useFormData

By watching the value directly from the form we no longer have a need
for local state, as we were just using it to determine whether the
throttle had changed from the default.

* Types cleanup

Remove some unneeded casts, add some needed ones.

* Rewrite About Step tests

Rather than asserting that the form is invalid through the UI, we
instead retrieve data out of the form hook and assert on that instead.

* Add memoization back to StepRuleDescription

I'm not sure that it's necessary, but best to leave it until we have
time to audit/remove multiple of these.

* Do not fetch ML Jobs if StepRuleDescription is not rendering ML Data

We were incorrectly invoking the useSecurityJobs hook any time the
StepRuleDescription component was rendered, regardless of whether we
actually needed that data.

This moves the useSecurityJobs hook into the component that uses it,
MlJobDescription. If we end up having multiple of these on the page we
can evaluate caching/sharing this data somehow, but for now this
prevents:

* 3x3=9 unnecessary ML calls on the Rule Create page.
* 1x3=3 unnecessary ML calls on Rule Details
* 1x3=3 unnecessary ML Calls on the Rule Edit page.

* Fix bug where revisiting the About step could modify the user's Risk Score

With the severity/risk score link back in place, there was a bug where a
user who had previously set a manual risk score would have it rewritten
on edit (or edit during rule creation).

This was due to a poorly-written useEffect that basically said "if there
is a severity, set a risk score." This has now been amended to say "if
the user changes the severity, set a risk score."

* Clean up About Step tests

* We don't need act(), it's not doing anything.
* We don't need to click Continue since we're just talking to the form
  hook

* Fix local form data when form isn't mounted

If the form isn't on the page (e.g. if we're read-only), then
useFormData will return no values. In these cases, we can simply fall
back to the initialState values, as they'll either be: the default
values on a new form, or: the current values on an active create/edit
form.

Updates the manual type of useFormData to reflect this "maybe" fact.

* Allow user to navigate between invalid tabs on Edit Rule

* Form hooks now _always_ return the form's data, regardless of validity
* Edit Rule now:
  * persists invalid data
  * submits both the active form and the destination form on navigation.
    This is necessary to refresh validations on the destination form,
    since the form lib assumes a newly-mounted form is valid
* simplifies "invalid tab" logic to be derived from our persisted data

* Fix logical error

If the rule is immutable, they can only edit actions.

* Remove unneeded eslint exception

Fixed by upstream #76471

* Make 21 the default risk score for a new rule

Since the default severity is 'low,' these two defaults now coincide.

* Remove duplicated type in favor of common one
gmmorris added a commit to gmmorris/kibana that referenced this pull request Sep 7, 2020
* master:
  [APM] Use observer.hostname instead of observer.name (elastic#76074)
  Legacy logging: fix remoteAddress being duplicated in userAgent field (elastic#76751)
  Remove legacy deprecation adapter (elastic#76753)
  [Security Solution][Detections] Rule forms cleanup (elastic#76138)
  Adds back in custom images for reporting + tests (elastic#76810)
  [APM] Fix overlapping transaction names (elastic#76083)
  [Ingest Pipelines] Add descriptions for ingest processors A-D (elastic#75975)
@MindyRS MindyRS added the Team: SecuritySolution Security Solutions Team working on SIEM, Endpoint, Timeline, Resolver, etc. label Sep 23, 2021
@elasticmachine
Copy link
Contributor

Pinging @elastic/security-solution (Team: SecuritySolution)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
release_note:skip Skip the PR/issue when compiling release notes Team: SecuritySolution Security Solutions Team working on SIEM, Endpoint, Timeline, Resolver, etc. Team:SIEM v7.10.0 v8.0.0
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants