Recent Posts

Several modules of Open Event Frontend involve start and end date-times. While for simple type fields like text, dropdowns or radio buttons, default semantic UI validations are available, which are used inside the app via the form mixin.  Any component which extends the form mixin can make specify the validations required using the getValidationRules hook. For instance, this set of rules will enforce validations on the field called ticket_price which will prohibit it from being left empty, or something other than a real number. getValidationRules() { ticketPrice: { identifier : 'ticket_price', rules : [ { type : 'empty', prompt : this.l10n.t('Please give your ticket a price') }, { type : 'number', prompt : this.l10n.t('Please give a proper price for you ticket') }, { type : 'decimal[0..]', prompt : this.l10n.t('Ticket price should be greater than 0') } ] }, } The validations provided by semantic UI only extend to single fields, and are independent of each other. However, in our use case we have four fields: Start dateStart timeEnd dateEnd time The general requirement is that the DateTime object formed by joining the start date and the start time should be before the DateTime object obtained by joining the end date and time. Also, these four distributed fields exist only on the frontend, on the server they are actually just two fields startsAt and endsAt each carrying UTC time values of DateTime objects. To split them into date and time a new computed function is created which is invoked in the model definitions of various resources. Consider the two following complex fields defined in the model according to the schema on the server. startsAt : attr('moment'), endsAt : attr('moment') In order to split them into two that helper is used as follows: startsAtDate : computedDateTimeSplit.bind(this)('startsAt', 'date', 'endsAt'), startsAtTime : computedDateTimeSplit.bind(this)('startsAt', 'time', 'endsAt'), endsAtDate : computedDateTimeSplit.bind(this)('endsAt', 'date') endsAtTime : computedDateTimeSplit.bind(this)('endsAt', 'time'), These values can then be used inside individual fields. To enhance the user experience jquery Calendar module is used to allow the user to enter the date and time values using a calendar and time picker as shown below. The computedDateTimeSplit helper, takes in the property whose part it is splitting, along with the specification of the part it will split. It also takes an optional endProperty argument, which is passed if it is being called for a start property. This function returns a pair of getters and setters,the get function returns the part of datetime object requested like, date or time where as the setter sets these values each time this function is called. export const computedDateTimeSplit = function(property, segmentFormat, endProperty) { return computed(property, { get() { return moment(this.get(property)).format(getFormat(segmentFormat)); }, set(key, value) { const newDate = moment(value, getFormat(segmentFormat)); let oldDate = newDate; if (segmentFormat === 'time') { oldDate.hour(newDate.hour()); oldDate.minute(newDate.minute()); } else if (segmentFormat === 'date') { oldDate.date(newDate.date()); oldDate.month(newDate.month()); oldDate.year(newDate.year()); } else { oldDate = newDate; } this.set(property, oldDate); } return value; } }); }; With this complex set up it is not possible to use semantic UI validations, hence we…

This blog article will continue the discussions about setting up ember tables on open event frontend. The implementation and design of the search module of the ember tables will be discussed. Open event server supports searching, using filter queries offered by flask-rest-json-api  Leveraging the refreshModel property of the queryParams, we can bind the value of a potential search query of a user to a queryParam. As the user types the query, the queryParam will change, which in turn will refresh the model, and send a new request to the server, giving the impression of a basic search. An icon input field with a search icon is placed on the top right corner of the table wrapped in its own separate component, search-box. The component is rendered within the default table component, which acts as a base for all other tables in the app. Here, searchQuery is binded to the param search defined in the controller mixin for ember tables. <div class="ui small icon input"> {{input type="text" value=searchQuery placeholder="Search ..." }} <i class="search icon"></i> </div> Now in order to generate the query string which will result in a search involves complex array filter manipulations, hence the logic was abstracted into the route mixin to avoid it’s repetition across various routes. Flast-rest-json api’s filters follow the following format for a  basic like styled sql filter: GET /events?filter=[ { "name": "event", "op": "ilike", "val": searchQuery } ] HTTP/1.1 Accept: application/vnd.api+json There are several other query combinations available for comparison, dates and other data-types or for relationships themselves. They needed to be appended to the filter portion of a query and then removed. When the search param is null, the query string has the following form: let queryString = { include : 'tickets,user', filter : filterOptions, 'page[size]' : params.per_page || 10, 'page[number]' : params.page || 1 }; filterOptions are other filters already present, for instance for listing orders on the basis of a dynamic status parameter,  filterOptions can be defined as follows. filterOptions = [ { name : 'status', op : 'eq', val : params.orders_status } ]; Since filterOptions are an array, new filters are either added or appended to the array. applySearchFilters(options, params, searchField) { searchField = kebabCase(searchField); if (params.search) { options.pushObject({ name : searchField, op : 'ilike', val : `%${params.search}%` }); } else { options.removeObject({ name : searchField, op : 'ilike', val : `%${params.search}%` }); } return options; } Thus a method called applySearchFilters is defined in the ember-table-router mixin. It takes three arguments: optionsparamssearchField Options is the filter array, discussed above, params contains the value of the current search param, and the searchField is the target of the search which is defined in the route’s model hook itself. The open-event-api server, expects the name of fields in kebab case, where as the front-end uses camelCase, hence the searchField is explicitly converted into a kebab-case string before manipulating the filter array. If params are present, it appends the filter object to the queryString, else this function will remove the very filter it…

This blog article will continue the discussions about setting up ember tables on open event frontend. The implementation and design of the pagination module of the ember tables will be discussed. Open event server uses JSON:API spec. Whenever the server returns a query, the JSON contains a meta field, that field contains the total number of records on the API server, which satisfy the query. This number, of course is different from the actual number of records returned which depend on the page size, and page number. But this serves as a good starting point for implementing the pagination module. For the open event server, the property inside the meta tag which contains the total count is called count. Using this property we define a computed property called totalContentLength. import Component from '@ember/component'; import { computed, action, get } from '@ember/object'; export default class extends Component { metaItemsCountProperty = 'count'; @computed('metaData') get totalContentLength() { return get(this.metaData, this.metaItemsCountProperty); } } Ember has great support for queryParams, and if in their declaration inside the route, the property refreshModel is set to true, then the model refreshes and reloads the data whenever the query params update. Thus, the current page and page size are maintained in the form of query params.  For computing the total number of pages (pageCount),  that the content will be split in, we have all three required variables, totalContentLength (computed above), currentPage and pageSize (both available as queryParams). Once we have pagesCount, it is easy to decide if moving forward or backwards is possible and can themselves be stored as computed properties. Similarly, in order to move forwards or backwards by a page, or to last or first pages, the queryParam for currentPage can be altered. import Component from '@ember/component'; import { computed, action, get } from '@ember/object'; export default class extends Component { metaItemsCountProperty = 'count'; @computed('metaData') get totalContentLength() { return get(this.metaData, this.metaItemsCountProperty); } @computed('currentPage', 'pageSize', 'totalContentLength') get pageCount() { let totalPages = 1; if (parseInt(this.pageSize) !== 0 && this.pageSize < this.totalContentLength) { totalPages = parseInt(this.totalContentLength / this.pageSize); if (this.totalContentLength % this.pageSize) { totalPages += 1; } } return totalPages; } @computed('currentPage') get moveToPreviousPageDisabled() { return this.currentPage <= 1; } @computed('currentPage', 'pageCount') get moveToNextPageDisabled() { return this.currentPage >= this.pageCount; } @action moveToNextPage() { if (!this.moveToNextPageDisabled) { this.incrementProperty('currentPage'); } } @action moveToPreviousPage() { if (!this.moveToPreviousPageDisabled) { this.decrementProperty('currentPage'); } } @action moveToLastPage() { if (!this.moveToNextPageDisabled) { this.set('currentPage', this.pageCount); } } @action moveToFirstPage() { if (!this.moveToPreviousPageDisabled) { this.set('currentPage', 1); } } } We also need to display, what is the current range of entries being shown. For eg, Showing 20-30 of 100 entries. This can again be computed by using current page, page size and  total number of entries. Using all of these computed properties, we can render the template for a pagination menu, which the user can use in order to navigate through the data.  <div class="ui small pagination menu"> <a role="button" class="item {{if moveToPreviousPageDisabled 'disabled'}}" {{action 'moveToFirstPage'}}> <i class="angle double left icon"></i> </a> <a role="button" class="item {{if moveToPreviousPageDisabled 'disabled'}}"…

This blog article will illustrate how ember tables were set up, reopened as a component, for customization and how the pagination module was implemented. Ember source 3.11 which Open event frontend uses is not compatible with the last release of ember tables, hence the master branch of ember tables which does support the latest ember source was chosen.  To install a dependency from a Github repository link instead of a yarn package, we can use yarn add addepar/ember-table#0aa5637 Ember tables offer no inbuilt theme, and use the default HTML styles. They do have support for styling, but only via CSS selectors in CSS files. In our use case, we needed the styling to be those of semantic UI tables. However, for that the classes had to be added inside the table element, and ember tables by default, don’t allow addition of classes, as the table was under other layers. Even if the classNames property of component had  been specified, it would just append the specified class names to the wrapper of the table element, not the actual table itself. Ember’s reopen feature was used to solve this problem. The reopen method allows ‘reopening’ of the component in the sense that it’s existing properties can be overwritten, or new properties can be added. It is a convention to store the reopened component files in a folder separate from external folder. The definition of the table was inside a component.js file inside the source of ember tables. Hence a new file located at app/extensions/ember-table/component.js was created.  Then the component.js was reopened to change the source of the template file which is used for ember tables. import component from 'ember-table/components/ember-table/component'; import layout from './template'; component.reopen({ layout }); This reopening modifies the ember table component, such that it now searches for the layout file in the new directory at ember-table/components/ember-table/layout.hbs The file layout.hbs now contains  <div class="resize-container"> <table class="ui unstackable table"> {{yield (hash api=api head=(component "ember-thead" api=api) body=(component "ember-tbody" api=api) foot=(component "ember-tfoot" api=api) )}} </table> </div> It is the exact same file which was present in the source of ember tables, with the only modification being the addition of class ui unstackable table to the table element. This class makes the table support semantic UI styling. Ember tables follow a structure similar to ember model tables, wherein the columns are defined inside the controller of the route which will render the table. However, it is not possible to pass the actions defined in the controller to the final cell components for columns without passing them throughout until the very last layer. I.e. actions of controller are not passed to the custom cell components, automatically. Specifying them explicitly results in loss of generalisation, and a significant portion of code will be repeated.  Also, sometimes, we might need to pass more than one valuePaths if we don’t need to pass the entire entity, if a column cell needs only some of the properties. Custom options were a requirement as well. Hence, to generalise these properties, the…

This blog will showcase how test email module was developed for open event.  This required implementing a new API endpoint on the Open event server, and it’s subsequent integration on the front-end. The test email functionality essentially tests if a given mail configuration is successfully able to deliver mail. The user gets an option to enter any email, and then trigger a test mail to that address. The tester may manually verify then if an email has been received. In  app/api/helpers/system_mails.py a field for the test email is added. TEST_MAIL: { 'recipient': 'User', 'subject': u'Test Mail Subject', 'message': ( u"This is a <strong> Test </strong> E-mail." ) } And in the helper file at app/api/helpers/mail.py, a function for sending a test mail is added. def send_test_email(recipient): send_email(to=recipient, action=TEST_MAIL, subject=MAILS[TEST_MAIL]['subject'], html=MAILS[TEST_MAIL]['message'] ) Since this is a trivial route, with no data exchange, we will implement this as a miscellaneous route under the settings API. Also, it is important that the route is restricted for use only by an admin. Decorators are used to ensure that. @admin_misc_routes.route('/test-mail', methods=['POST']) @is_admin def test_email_setup(): recipient = request.json.get('recipient') if not recipient: return UnprocessableEntityError({'source': 'recipient'}, 'Required parameter recipient not found').respond() send_test_email(recipient) return make_response(jsonify(message='Test mail sent, please verify delivery'), 200) The endpoint only accepts POST requests, whose body should contain a field called recipient which contains the email address to which the test mail should be sent. With the API endpoint for test emails ready, a small form needs to be added on the front-end in the admin area. A new form for this field was added within the form of general system settings. sendTestMail() { this.onValid(() => { let payload = { recipient: this.recipientEmail }; let config = { skipDataTransform: true }; this.loader.post('/test-mail', JSON.stringify(payload), config) .then(response => { this.notify.success(response.message); }) .catch(e => { console.warn(e); this.notify.error(this.l10n.t('An unexpected error has occurred')); }); }); } The action sendTestMail sends the necessary POST request to the API endpoint for sending the mail. sendTestMail() { this.onValid(() => { let payload = { recipient: this.recipientEmail }; let config = { skipDataTransform: true }; this.loader.post('/test-mail', JSON.stringify(payload), config) .then(response => { this.notify.success(response.message); }) .catch(e => { console.warn(e); this.notify.error(this.l10n.t('An unexpected error has occurred')); }); }); } Additionally, email validations are added to the field which accepts the email input. This module after it’s implementation is able to handle both cases, if the mail endpoint returns an error status code, this means mail send function on the server is broken, or the configuration is misconfigured in an unexpected manner. If it returns a successful status, and yet mail has not been received on the desired email address, this might point to a deeper problem regarding, the sending mail getting blacklisted or otherwise. Resources  Ember JS-model hook documentationOpen Event API DocsJson api spec patch request docs

This blog article will illustrate how, after significant refactors, eventyay.com was switched from open-event-legacy to the new version of open-event which has a decoupled open-event-API-server and frontend. We will discuss this switch from two aspects - the database migration and the DevOps logistic, as we moved from google cloud to a  more economically feasible platform - hetzner cloud. Downloading a copy of the legacy database from Kubernetes pod on the google cloud platform The first step, was to obtain a copy of the database being used in production. The database was stored inside a pod named Postgres of the Kubernetes cluster. gcloud login gcloud container clusters get-credentials vintage-cluster --zone us-west1-a --project eventyay These commands initialized and authenticated the gcloud sdk with the project eventyay. Next, to gain bash access to the postgres pod, exec of kubectl CLI was used. kubectl exec -it postgres -- /bin/bash The database in both the new and legacy versions was postgresSql. Using the pg_dump functionality, a database can be dumped into a transferable file kubectl exec -it postgres -- /bin/bash However this file still resides on the kubernetes pod’s local storage  itself. The cp utility of kubectl CLI comes in handy to copy that file from pod to local storage.  kubectl cp default/postgres:legacy-24-03-2019.pgsql ~/Downloads/ This command transfers the file to local storage. Now we have a database we can begin to refactor, for this we first need to import it into the postgres instance on the local machine, to take a peek inside the schema. Psql -U postgres Create database legacy_24_03_2019 Create database legacy_24_03_2019_reference postgres legacy_24_03_2019 < ~/Downloads/legacy-24-03-2019.pgsql These commands dump the data inside the  legacy database into the newly created database inside the Postgres instance. The schema and architecture of the new version of server is different from the legacy database, hence DB has to go through migrations. The last migration file shared by legacy and the next gen of server is corresponding to the migration ddaa6df27340. However, it is important to note that the migrations branched from here onwards. Hence to downgrade the databse to the last legacy migration, we need to use the migrations folder of the legacy version of eventyay and not the original migrations directory Assuming that the migrations from the legacy server stiored in a folder called migrations-legacy mv migrations migrations-new mv migrations-legacy migrations Change the config in the local .env inside open-event-server to switch to this newly created database. 9d21de792967 Ticket holder thing python manage.py db downgrade ddaa6df27340 Then we upgrade to the latest instance of the db Switch back the directories (venv) Abhinavs-MacBook-Pro:open-event-server abhinav$ mv migrations migrations-legacy (venv) Abhinavs-MacBook-Pro:open-event-server abhinav$ mv migrations-new migrations Then we upgrade the database  Python manage.py db upgrade These commands migrate the database in principal, and we have a basic system with core functionality using legacy data but the database is not at all ready to be used in production yet. It is rife with bugs. The migrations file don’t cover each change, and some are outright breaking. Some have problems like comma…