Home

Awesome

GitHub release (latest SemVer) Build Status codecov

Cylc UI

Installation

Install the UI Server which bundles the UI.

Copyright and Terms of Use

Copyright (C) 2018-<span actions:bind='current-year'>2024</span> NIWA & British Crown (Met Office) & Contributors.

Cylc is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

Cylc is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with Cylc. If not, see GNU licenses.

Development

Install & Build

# Project setup
yarn install

# start dev server in offline mode (uses mock data, auto-updates the browser page on change)
yarn run serve

# pass options to vite (e.g. to use a different port or expose host)
VITE_OPTIONS='--host myhost' yarn run serve

# build for production
yarn run build

# build for development (rebuilds on change)
yarn run build:watch
# and launch using
cylc gui --ui-build-dir=<cylc-ui-path>/dist/

# start dev server in offline mode, using the build instead of source files
yarn run preview

Note the incremental rebuild is quite slow so an alternative to yarn run build:watch is to run the Vite development server while using the Cylc UI Server live data:

# First launch the gui to authenticate with the URL token
cylc gui --port=3000 --ServerApp.allow_origin='http://localhost:5173'
# Close that tab once it's loaded
# Now launch using
yarn run serve:vue --mode development
# (you must access via http://localhost:5173)

Tests

There are three groups of tests:

For coverage:

yarn run coverage:unit
yarn run coverage:e2e

Mocked Data

The "offline" mode (aka yarn run serve) which is also used for the end to end tests is powered by a "mock" data server.

You can find the index of mocked data here: src/services/mock/json/index.cjs

Mock data is automatically loaded when the subscription/query issued matches an entry in that file.

Code Style

See .eslintrc.cjs for style. To test, run:

yarn run lint

Or to lint a particular file/directory:

yarn eslint path/to/file

Project Setup

We are using Vue. The project was originally created with vue-cli, but has switched to Vite with the upgrade from Vue 2 to 3.

The configuration for how the app is served and built is defined in vite.config.js.

We are currently using the Vuetify component library. Its configuration is defined in src/plugins/vuetify.js.

We use concurrently for concurrently running the mock data json-server and the Vite dev server, and also Cypress. This is configured in scripts/concurrently.cjs.

Browser compatibility

There are two aspects of browser compatibility:

The former is handled by Vite. It uses esbuild to transform instances of newer syntax when building.

However, new APIs are not handled and must be polyfilled if deemed necessary.

We define a specification for browser compatibility in .browserslistrc. See https://github.com/browserslist/browserslist.

Remember it is not just our source code that must meet our back-compat specification, but our bundled dependencies (e.g. Vuetify) too! Vite/esbuild handles syntax for bundled dependencies.

However the bottom line is that as of 2023, browser support is much less of an issue than it was even a couple of years ago, due to the proliferation of evergreen browsers. The only real concern is bleeding-edge API calls creeping into our source code or runtime dependencies. To catch this, we are using eslint-plugin-compat in CI to scan the build for any such API calls.

Integration with the backend Cylc UI server

Running yarn run build[:watch] outputs the build into the ./dist/ folder. When running the Cylc Hub, you must remember to point the static files directory to the location of your ./dist folder.

This way with both Cylc Hub and Cylc UI running, you can work on either - or both - projects. Changes done in your Tornado application should reflect immediately or upon process restart. While the changes done in your Vue.js application will be automatically handled by your build:watch command.

Internationalization

[!NOTE] Internationalization is only partly implemented at the moment.

This project utilizes vue-i18n for internationalization. While this project is not part of Vue.js, it is maintained by one of the Vue.js core developers.

Messages for internationalization are kept in JSON files. Look at src/lang/ for each locale. For example, for British English, the message files are kept under src/lang/en-GB.

The locale is defined by a variable $i18n, which is accessible in each component. So in a component you should be able to change the locale - if necessary - by calling this.$i18n.locale = 'pt-BR'.

Accessibility

After applying changes to the code, might be a good idea to pass the new version of the application through an accessibility tool such as WAVE.

There is also a browser extension which makes testing the development version much easier.

TypeScript

TypeScript is most likely the future for us. It can be adopted gradually. At the moment we only have JSDoc comments which can provide type information in your IDE.

How The Data Is Provisioned

The Cylc UI connects to the GraphQL endpoint provided by the Cylc UI Server using a websocket.

GraphQL Queries

Here's the "Hello World!" of Cylc GraphQL queries, it returns the ID of every workflow (under ~/cylc-run):

query {
  workflows {
    id
  }
}

GraphQL Subscriptions

To keep data up to date, we use subscriptions, a subscription is essentially a repeating query. The way we've set it up, the server will only send new responses when the data changes.

This subscription will send back the ID of every workflow, any time the list of workflows changes (i.e. when you install a new workflow or clean an old one):

subscription {
  workflows {
    id
  }
}

GraphQL Delta-Subscriptions

To avoid sending the entire list of workflow IDs every time the list changes we subscribe to special "delta" objects. These allow us to track changes in the list which is useful for efficiently synchronising data between the server and the web app.

This subscription will notify us when workflows are added, updated or removed:

subscription {
  deltas {
    added {
      workflows {
        id
      }
    }
    updated {
      workflows {
        id
      }
    }
    pruned {
      id
    }
  }
}

How Views Request Data

The Cylc "views" (e.g. the Tree, Table and Graph views) define a subscription which defines all of the data they require.

This subscription is automatically registered with the WorkflowService when the view is loaded.

The WorkflowService will then issue and manage this subscription on your behalf. The data you requested will become available in the data store when it arrives. The data store will be kept up to date whenever this data changes.

The subscription will be cancelled when the view is closed.

Subscription Merging

When a new view is opened, the WorkflowService will take the subscription for this view merge it with any other active subscriptions from other views.

E.G. If view-a has this subscription:

subscription {
  deltas {
    added {
      taskProxies {
        id
        name
        status
        firstParent
      }
    }
  }
}

And view-b has this subscription:

subscription {
  deltas {
    added {
      taskProxies {
        id
        status
        isHeld
        isRunahead
        isQueued
      }
    }
  }
}

Then the WorkflowService will merge these subscriptions into:

subscription {
  deltas {
    added {
      taskProxies {
        id
        name
        status
        firstParent
        isHeld
        isRunahead
        isQueued
      }
    }
  }
}

This is how we avoid requesting duplicate information about the same things for different views.

Each view can request whatever data it needs, however, filtering cannot be performed in the subscription because that filtering would apply to all merged subscriptions. If two subscriptions cannot be merged (e.g. different filtering options) then an error will be raised. Perform filtering within the view where appropriate.

When the UI Server sends deltas back to the WorkflowService, they are used to maintain the data store.

Note, ensure you request the id for all objects, the data store needs this to operate.

The Data Store

The central data store contains all of the information requested by all of the views. The WorkflowService keeps this up to date by applying the deltas it receives from the UI Server to the store in order. Each delta is timestamped which allows us to detect transmission errors, the store will be rebuilt in the event of error.

The data store is currently VueX but will probably be migrated to Pina in the future.

The data store contains an entry for every object requested where an object might be a user, workflow, cycle-point, task or job. Every object has a unique ID.

For example, a workflow in the data store might look like this:

{
  // the unique object ID as a string
  id: '~me/my-workflow',
  // the parsed ID as an object
  tokens: {user: 'me', workflow: 'my-workflow, ...},
  // the kind of node that this is
  type: 'workflow',
  // the last part of the ID
  name: 'my-workflow',
  // any data that has been requested
  data: {
    status: 'running',
    host: 'myhost',
    port: '1234',
  },
  // read on...
  children: [],
}

You can access these objects via one of two ways, the index or the tree.

The Index

This is a mapping which contains every object listed by its ID.

E.G. something along the lines of:

$index = {
  '~me': Object,
  '~me/my-workflow': Object,
  '~me/my-workflow//cycle': Object,
  '~me/my-workflow//cycle/task': Object,
  '~me/my-workflow//cycle/task/job': Object,
}

This is useful if you know the ID of the object you want to retrieve.

The Tree

For convenience, these nodes are also arranged into a hierarchy allowing you to walk/iterate over them.

E.G. the tree structure might look like this:

Family Hierarchies

Some views (e.g. the Tree view) want access to the family hierarchy of tasks.

E.G. for this workflow:

[runtime]
  [A]
  [[a1, a2]]
    inherit = A

Defines this hierarchy:

If you need to walk the family hierarchy down to a task (like the Tree view does), then add these fields to your TaskProxy subscription:

ancestors
FirstParent

The data store will now automatically construct a family tree for you to iterate in every Cycle object.

Access this using the FamilyTree property.

Edges & Namespaces

Some views may require graph edges (i.e. the arrows in the Graph view).

Some views may require namespaces (i.e. task definitions).

These are available via the $edges and $namespaces indexes which are available on workflow objects.

Example View

For documentation on how to write a view see src/views/SimpleTree.vue which contains a simple implementation of a minimal tree view.