We have recently discovered that we were using ID type in place of UUID
type in many place in the code.
We have merged #4895 but this introduced bugs as we forgot to replace it
everywhere
- Implemented dataloader package on metadata graphql server
- Implemented a dataloader for relation metadata module
---------
Co-authored-by: Jérémy M <jeremy.magrin@gmail.com>
We have discovered that GraphQL inputs for fields of type ids in create
/ update input where using a more permissive ID type than the type used
in FilterInput in queries.
This PRs fixes it and make sure that all Input are using UUID graphql
scalar types
## Context
Calendar scope was too broad, this PR updates it to events only.
Also changing "Cannot connect Google account to demo workspace" error to
a 404 to avoid having a 500 for something expected
We've introduced in PR #4373 standard ids to be able to rename standard
fields and objects.
Fields part was working properly, but objects part was not yet
implemented.
This PR is adding the missing parts to make it work.
Foreign tables should be created using migrations, as we do for standard
tables.
Since those are not really generated from the object metadata but from
the remote table, those migrations won't live in the object metadata
service.
This PR:
- creates new types of migration : create_foreign_table and
drop_foreign_table
- triggers those migrations rather than raw queries directly
- moves the logic to fetch current foreign tables into the remote table
service since this is not directly linked to postgres data wrapper
- adds logic to unsync all tables before deleting
---------
Co-authored-by: Thomas Trompette <thomast@twenty.com>
This PR:
- creates the query to delete a connection
- creates the hook that triggers the query
- triggers the hook function when clicking on remove + get back to
connection page
---------
Co-authored-by: Thomas Trompette <thomast@twenty.com>
- Select component was adding a duplicate useListenClickOutside already
present in useDropdown for closing dropdown.
- Added debug logs for hotkeys scopes
This PR is dropping the column `targetColumnMap` of fieldMetadata
entities.
The goal of this column was to properly map field to their respecting
column in the table.
We decide to drop it and instead compute the column name on the fly when
we need it, as it's more easier to support.
Some parts of the code has been refactored to try making implementation
of composite type more easier to understand and maintain.
Fix#3760
---------
Co-authored-by: Charles Bochet <charles@twenty.com>
Endpoint is broken since we now use `Remote` as a suffix for remote
table names.
This PR:
- creates a common function to calculate the name of the remote table
- use it in the `findAvailableRemotePostgresTables` to know if a table
has been synced or not
Co-authored-by: Thomas Trompette <thomast@twenty.com>
## Removing repetitive queries (impacting performance on page load)
We have recently introduced the capability to detect schema version
mismatch. To do that, we add a new header in all our queries. On page
load, this header is added once we know the currentUser and especially
its currentWorkspace and related schema version.
However, applying this header to apollo client will re-trigger all
queries that have been already performed (GetClientConfig and
GetCurrentUser). To avoid re-triggering them, I'm introducing two new
"isLoaded" states and skip the query if the query has already been
performed
## Fixing Relation Detail not displaying data on show page
Small bug introduced in a previous PR
In this PR:
- fix empty list placeholder positionning
- prevent user from erasing custom address field as composite types
removal is not supported yet @ijreilly FYI
- fix show page relation error
- Implement address filter
## Context
We are now removing Messaging V2 feature flag to use it everywhere.
## Implementation
- renaming FetchWorkspaceMessagesCommandsModule to
MessagingCommandModule to make it more generic since it it hosts all
commands related to the messaging module
- creating a crons folder inside commands and jobs crons should be named
with xxx.cron.command.ts instead of xxx.command.ts. Same for jobs, jobs
should be named with xxx.cron.job.ts. In a future PR we should make sure
those CronJobs implement a CronJob interface since it's a bit different
(a CronJob does not contain a payload compared to a Job)
- Cron commands have been renamed to "cron:$module:command" so
`fetch-all-workspaces-messages-from-cache:cron:start` has been renamed
to `cron:messaging:gmail-fetch-messages-from-cache`. Also having to
create a command to stop the cron is a bit painful to maintain so I
removed them for now, this can be easily done manually with pg-boss or
bull-mq
- Removing full-sync and partial-sync commands as they were there for
testing only, we might put them back at some point but we will have to
adapt the code anyway.
- Feature flag has been removed from the MessageChannel standard object
to make sure those new columns are created during the next sync-metadata
## Context
Calendar tables are behind a featureFlag, they do not exist if the
feature flag is off which means we should not use them for the same
reason. I'm adding a check on the featureFlag before calling the
repository.
```
Error executing raw query for workspace 20202020-1c25-4d02-bf25-6aeccf7ea419: relation "workspace_1wgvd1injqtife6y4rvfbu3h5.calendarEventParticipant" does not exist
```
## Test
locally with and without featureflag
**Fixed the following build issue:**
```
./src/app/contributors/api/fetch-issues-prs.tsx
12:3 Error: Type boolean trivially inferred from a boolean literal, remove type annotation. @typescript-eslint/no-inferrable-types
./src/app/contributors/api/search-issues-prs.tsx
12:3 Error: Type boolean trivially inferred from a boolean literal, remove type annotation. @typescript-eslint/no-inferrable-types
```
Co-authored-by: Ady Beraud <a.beraud96@gmail.com>
**Fixed different issues** :
- Multiple CSS fixes: font-size, colors, margins, z-index ...
- Fixed hover on contributor avatars
- Added link to contributors in footer
- Made the year in the footer dynamic (2023 --> 2024)
- Added name of contributor in "Thank you" section of Contributor page
- Added footer in small screens
- Made Activity Log Responsive
- Fixed bug in "saving issues to DB", title was null everywhere. I
needed to implement an "upsert" behaviour to update the existing
database on init
**To be noted :**
There is the following bug on production happening on mobile when you
refresh a second time :
<img width="1440" alt="Screenshot 2024-04-05 at 01 30 58"
src="https://github.com/twentyhq/twenty/assets/102751374/b935b07a-63dc-463d-8dcb-070ad4ef6db0">
It seems to be related to the following issue on mdx :
[https://github.com/hashicorp/next-mdx-remote/issues/350](https://github.com/hashicorp/next-mdx-remote/issues/350)
I added the following code that fixed this bug for me in development
(this needs to be tested in production) :
```
const serialized = await serialize(content, {
mdxOptions: {
development: process.env.NODE_ENV === 'development',
}
})
```
---------
Co-authored-by: Ady Beraud <a.beraud96@gmail.com>
Re-enables no-console eslint rule in stories and tests files:
- In stories, use `action` from '@storybook/addon-actions' or `fn` from
'@storybook/test' instead of console.
- In tests, console methods can be mocked like this:
`global.console.error = jest.fn()`.
In this PR, I'm fixing two things on the ViewPicker in Create mode:
- if the Dropdown has no max height, it should not be scrollable (which
is causing issue with inner dropdowns being cut by overflow: hidden
- if the user has changed the icon, the type or the name of the view,
consider the create form as isDirty and prevent its value to be
overriden by re-renders (cache updates for example)
## Context
When running a command, the process should end normally however it stays
hanging due to the open connection with redis client (when
CACHE_STORAGE_TYPE=redis)
This PR adds the necessary logic to gracefully close the connection once
the module is destroyed. Thanks to that, the command process now
properly ends once executed.
related to #4749.
Adjusted the max height for window so that it cuts the last option a
little.
I wanted to test the menu from the second pic in the issue with the
multi select option but could not figure out where it was in the
application.
---------
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
updatedAt and deletedAt field changed to string like createdAt field. On
file upload updatedAt field will be given date same as createdAt.
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
Backend: Adding a new util function that throw an error if the
objectMetadata is remote
Frontend: hiding the save button when remote
Also renaming `useObjectMetadataItemForSettings` since this hook is used
in other places than settings and is not in the settings repo. Name can
definitely be challenged!
---------
Co-authored-by: Thomas Trompette <thomast@twenty.com>
Split from https://github.com/twentyhq/twenty/pull/4518
- Setup `@ui/*` as an internal alias to reference `twenty-ui/src`.
- Configures twenty-front to understand the `@ui/*` alias on development
mode, so twenty-ui can be hot reloaded.
- When building on production mode, twenty-front needs twenty-ui to be
built beforehand (which is automatic with the `dependsOn` option).
- Configures twenty-front to understand the `@ui/*` alias when launching
tests, so there is no need to re-build twenty-ui for tests.
---------
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
Several fixes for remote objects:
- labels are now displayed in title case. Added an util for this.
- Ids are often integers but the foreign keys on the relations were
uuid. Sending the id type to the object metadata service so it can
creates the foreign key accordingly
- Graphql comments are override when several remote objects are
imported. Building a function that fetch the existing comment and update
it
---------
Co-authored-by: Thomas Trompette <thomast@twenty.com>
We used to not type properly the return of getRecordFromCache before the
work from last week.
The getViewFromCache function was patching this temporarily in a "dirty"
way.
Now that this is cleaner, I'm removing the typescript patch. This fixing
several behaviors on viewBar
Split from https://github.com/twentyhq/twenty/pull/4518
Part of #4766
Adds a script to auto-generate twenty-ui exports in `index.ts` files.
---------
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
Split from https://github.com/twentyhq/twenty/pull/4518
Related to #4766
Mutualizes eslint config between projects.
I didn't include `twenty-server` in this PR as this was causing too many
lint errors.
We've seen a few cascading errors (e.g. comment.activityId would be non
nullable but cascade behavior is set to "set null"). I think it's safer
if we have to explicitly chose the deletion behavior it every time.
Especially since Postgres default to "No action" while we defaulted to
"Set Null", which is confusing.
In the future we will most likely introduce a second param
`onSoftDelete` in the decorator
**Context**
cf. feature request
[#4597](https://github.com/twentyhq/twenty/issues/4597)
Enables deletion of custom fields that aren't active nor of type
relation
Also
1. renamed a misnamed file
2. deleted redundant hook BeforeDeleteOneField as it seemed best to move
the logic to the resolver instead
**How was it tested?**
Did not write unit tests as code is to be migrated (discussed with
@Weiko).
Locally tested.
---------
Co-authored-by: Marie Stoppa <mariestoppa@MacBook-Pro-de-Marie.local>
Closes#4556
- Renames some pages and components after discussion about terminology
with @thomtrp.
- Creates the Settings/Integrations/Database/Connection page.
Having 2 different dev setups caused confusion, let's remove the Docker
local setup and recommend people install yarn locally.
Also simplified some docs by merging pages together, the recommend
self-hosting option is now the docker-compose / to adapt the
docker-compose.
Foreign table id cannot be a foreign key of a base table. But the
current code use foreign keys to link object metadata with activities,
events... So we will:
- create a column without creating a foreign key
- add a comment on the table schema so pg_graphql sees it as a foreign
key
This PR:
- refactor a bit object metadata service so the mutation creation is
separated into an util
- adds the mutation creation for remote object relations
- add a new type of mutation to create a comment
---------
Co-authored-by: Thomas Trompette <thomast@twenty.com>
## Context
SyncExternalId should be renamed because this won't always represent an
id. For example, microsoft API does not use ids but dates for their
sync. Also we think external is a bit redundant so we are removing it.
Note: this field is not used by the front-end (and will probably never
be)
Fix: (#4204)
The issue was that when that panel was opened its content would wrap
instead of maintaining its desired structure. I fixed this bug by adding
a minimum width to the panel's contents so that they would stay
correctly formatted throughout the opening transition.
---------
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
We were missing `JsDom` dependencies in the package.json generated by nx
while running `twenty-server`: `yarn nx build:packageJson`
Detailed explanation:
- we are currently using nx paradigm which is to put dependencies of all
projets at root, which enables global package migrations for the whole
monorepo
- for production containers, we only want specific project dependency to
be added. This is done by running `yarn nx build:packageJson` on
`twenty-server`. Nx is statically analyzing twenty-server dependencies
and generating a tailored package.json that production containers can
later use.
- However, `nx` static analysis is not flawless and is missing some
packages. We are going to stop using it as the value is not there yet
but the burden for developers is high. The guideline is to put back
project dependencies into specific package `package.json`
- Therefore, I'm adding `jsdom` to twenty-server `package.json`
- If sync fails we set authFailedAt
- This information is displayed in the frontend in accounts with a `Sync
Failed` pill
- The user can reconnect his account in the dropdown menu
- A new OAuth flow is triggered
- The account is synced
## Context
A new ADDRESS field type has been introduced and the company object has
been updated to use this new type however this introduced a few
regressions.
The good strategy would be to introduce a new field and rename the old
one.
This PR revert that change to fix the issue.
The tests were broken. It turns out that it is not easy to mock two
apolloClients as apollo only provides "MockedProvider" component to mock
apollo in tests.
MockedProvider Api does not allow us to mock on client level.
For now, I'm defaulting to the base ApolloClient in test mode to avoid
the issue
Split from https://github.com/twentyhq/twenty/pull/4518
- Upgrades dependencies and applies automatic config migrations with the
command: `npx nx migrate nx` (see
https://nx.dev/nx-api/nx/documents/migrate)
- Fixes lint errors after upgrading `@typescript-eslint`
Note: it was not possible (for now) to migrate Nx to the latest stable
version (v18.2.1) because it upgrades Typescript to v5.4.3, which seems
to cause a bug on install when Yarn tries to apply its native patches.
Might be a bug on the Yarn side.
When writing to the normalized cache (record), it's crucial to use _refs
for relationships to avoid many problems. Essentially, we only deal with
level 0 and generate all fields to be comfortable with their defaults.
When writing in queries (which should be very rare, the only cases are
prefetch and the case of activities due to the nested query; I've
reduced this to a single file for activities
usePrepareFindManyActivitiesQuery 🙂), it's important to use queryFields
to avoid bugs. I've implemented them on the side of query generation and
record generation.
When doing an updateOne / createOne, etc., it's necessary to distinguish
between optimistic writing (which we actually want to do with _refs) and
the server response without refs. This allows for a clean write in the
optimistic cache without worrying about nesting (as the first point).
To simplify the whole activities part, write to the normalized cache
first. Then, base queries on it in an idempotent manner. This way,
there's no need to worry about the current page or action. The
normalized cache is up-to-date, so I update the queries. Same idea as
for optimisticEffects, actually.
Finally, I've triggered optimisticEffects rather than the manual update
of many queries.
---------
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
* default value boolean fixed
* fixed creation, fixed updating a value to false
* fixed default value for default value if boolean
* fixed tests
---------
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
* feat: implement confirmation prompt
* feat: remove prop drilling and introduce recoil state
* chore: fix eslint issues
* feat: set record text according to length of records
* chore: fix eslint issues
* refactor: made changes according to code review
* fix: show delete according to singular and plural records.
* fix: eslint issues
* feat: show number of selected records
* style: fix positioning of actionbar
* feat: display ConfirmationModal seperately
* chore: remove recoil state and use usestate instead
* chore: minor change
---------
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
* adding back new navbar structure
* adding back new home
* adding back objects
* adding back 4 pages
* adding back some pages
* added 3 videos for api, webhook and tasks
* feat: add more dependencies check, randomize postgres admin password, tail logs of server container
* feat: improve retro compatibility
* feat: comment POSTGRES_ADMIN_PASSWORD as it will be generated by the one liner
