How to implement Full Text Search in Meteor/Telescope - meteor

I have attempted implementing search in Telescope using pure javascript, since it looks like FTS is a while off for Meteor to implement and I couldn't get 2.4 playing nicely with Meteor yet.
I'm using the existing pagination model that is already implemented in Telescope to display the Top/New/Best posts, plus a Session variable for the search keyword that is set in the Router when you navigate to e.g. /search/foobar.
However, it doesn't quite seem to be working; when I have, say, 100 posts, the regular paginated subscription only comes back with 25 of these and my search results only show the posts in the first 25.
I've been banging my head against a wall for days trying to debug this one: sometimes it works, sometimes it doesn't!
Here's the code (I've included all additional search code for reference):
app.js:
var resultsPostsSubscription = function() {
var handle = paginatedSearchSubscription( 10, 'searchResults' );
handle.fetch = function() {
return limitDocuments( searchPosts( Session.get( 'keyword' ) ), handle.loaded() );
};
return handle;
};
var resultsPostsHandle = resultsPostsSubscription();
paginated_sub.js:
I duplicated the existing paginatedSubscription because I can't pass a Session var in as an arg; it needs to be dynamic. I'll probably refactor later.
paginatedSearchSubscription = function (perPage/*, name, arguments */) {
var handle = new PaginatedSubscriptionHandle(perPage);
var args = Array.prototype.slice.call(arguments, 1);
Meteor.autosubscribe(function() {
var subHandle = Meteor.subscribe.apply(this, args.concat([
Session.get( 'keyword' ), handle.limit(), function() { handle.done(); }
]));
handle.stop = subHandle.stop;
});
return handle;
}
search.js: (new file, in /common directory)
// get all posts where headline, categories, tags or body are LIKE %keyword%
searchPosts = function( keyword ) {
var query = new RegExp( keyword, 'i' );
var results = Posts.find( { $or: [ { 'headline': query }, { 'categories': query }, { 'tags': query }, { 'body': query } ] } );
return results;
};
publish.js:
Meteor.publish( 'searchResults', searchPosts );
posts_list.html:
<template name="posts_results">
{{> posts_list resultsPostsHandle}}
</template>
posts_list.js:
Template.posts_results.resultsPostsHandle = function() {
return resultsPostsHandle;
};
router.js:
there's a search bar in the nav that redirects to here
posts_results = function( keyword ) {
Session.set( 'keyword' , keyword );
return 'posts_results';
};
Meteor.Router.add({
...
'/search/:keyword':posts_results,
...
})
Any help would be greatly appreciated!

A little late but here is a full write up on how to implement full text search in meteor.
"The simplest way without editing any Meteor code is to use your own mongodb."

Related

Meteor publications/subscriptions not working as expected

I have two publications.
The first pub implements a search. This search in particular.
/* publications.js */
Meteor.publish('patients.appointments.search', function (search) {
check(search, Match.OneOf(String, null, undefined));
var query = {},
projection = {
limit: 10,
sort: { 'profile.surname': 1 } };
if (search) {
var regex = new RegExp( search, 'i' );
query = {
$or: [
{'profile.first_name': regex},
{'profile.middle_name': regex},
{'profile.surname': regex}
]
};
projection.limit = 20;
}
return Patients.find(query, projection);
});
The second one basically returns some fields
/* publications.js */
Meteor.publish('patients.appointments', function () {
return Patients.find({}, {fields: {'profile.first_name': 1,
'profile.middle_name': 1,
'profile.surname': 1});
});
I've subscribed to each publication like so:
/* appointments.js */
Template.appointmentNewPatientSearch.onCreated(function () {
var template = Template.instance();
template.searchQuery = new ReactiveVar();
template.searching = new ReactiveVar(false);
template.autorun(function () {
template.subscribe('patients.appointments.search', template.searchQuery.get(), function () {
setTimeout(function () {
template.searching.set(false);
}, 300);
});
});
});
Template.appointmentNewPatientName.onCreated(function () {
this.subscribe('patients.appointments');
});
So here's my problem: When I use the second subscription (to appointments.patients), the first one doesn't work. When I comment the second subscription, the first one works again. I'm not sure what I'm doing wrong here.
The issue here is you have two sets of publications for the same Collection. So when you refer to the collection in client there is now way to specify which one of the publication it has to refer too.
What you can do is, publish all data collectively i.e. all fields you are going to need and then use code on client to perform queries on them.
Alternatively, the better approach will be to have two templates. A descriptive code:
<template name="template1">
//Code here
{{> template2}} //include template 2 here
</template>
<template name="template2">
//Code for template 2
</template>
Now, subscribe for one publication to template one and do the stuff there. Subscribe to second publication to template 2.
In the main template (template1) include template2 in it using the handlebar syntax {{> template2}}

Jasmine - Testing links via Webdriver I/O

I have been working on a end-to-end test using Webdriver I/O from Jasmine. One specific scenario has been giving me significant challenges.
I have a page with 5 links on it. The number of links actually challenges as the page is dynamic. I want to test the links to see if each links' title matches the title of the page that it links to. Due to the fact that the links are dynamically generated, I cannot just hard code tests for each link. So, I'm trying the following:
it('should match link titles to page titles', function(done) {
client = webdriverio.remote(settings.capabilities).init()
.url('http://www.example.com')
.elements('a').then(function(links) {
var mappings = [];
// For every link store the link title and corresponding page title
var results = [];
for (var i=0; i<links.value.length; i++) {
mappings.push({ linkTitle: links.value[0].title, pageTitle: '' });
results.push(client.click(links.value[i])
.getTitle().then(function(title, i) {
mappings[i].pageTitle = title;
});
);
}
// Once all promises have resolved, compared each link title to each corresponding page title
Promise.all(results).then(function() {
for (var i=0; i<mappings.length; i++) {
var mapping = mappings[i];
expect(mapping.linkTitle).toBe(mapping.pageTitle);
}
done();
});
});
;
});
I'm unable to even confirm if I'm getting the link title properly. I believe there is something I entirely misunderstand. I am not even getting each links title property. I'm definately not getting the corresponding page title. I think I'm lost in closure world here. Yet, I'm not sure.
UPDATE - NOV 24
I still have not figured this out. However, i believe it has something to do with the fact that Webdriver I/O uses the Q promise library. I came to this conclusion because the following test works:
it('should match link titles to page titles', function(done) {
var promise = new Promise(function(resolve, reject) {
setTimeout(function() { resolve(); }, 1000);
});
promise.then(function() {
var promises = [];
for (var i=0; i<3; i++) {
promises.push(
new Promise(function(resolve, reject) {
setTimeout(function() {
resolve();
}, 500);
})
);
}
Promise.all(promises).then(function() {
expect(true).toBe(true)
done();
});
});
However, the following does NOT work:
it('should match link titles to page titles', function(done) {
client = webdriverio.remote(settings.capabilities).init()
.url('http://www.example.com')
.elements('a').then(function(links) {
var mappings = [];
// For every link store the link title and corresponding page title
var results = [];
for (var i=0; i<links.value.length; i++) {
mappings.push({ linkTitle: links.value[0].title, pageTitle: '' });
results.push(client.click(links.value[i])
.getTitle().then(function(title, i) {
mappings[i].pageTitle = title;
});
);
}
// Once all promises have resolved, compared each link title to each corresponding page title
Q.all(results).then(function() {
for (var i=0; i<mappings.length; i++) {
var mapping = mappings[i];
expect(mapping.linkTitle).toBe(mapping.pageTitle);
}
done();
});
})
;
});
I'm not getting any exceptions. Yet, the code inside of Q.all does not seem to get executed. I'm not sure what to do here.
Reading the WebdriverIO manual, I feel like there are a few things wrong in your approach:
elements('a') returns WebElement JSON objects (https://code.google.com/p/selenium/wiki/JsonWireProtocol#WebElement_JSON_Object) NOT WebElements, so there is no title property thus linkTitle will always be undefined - http://webdriver.io/api/protocol/elements.html
Also, because it's a WebElement JSON object you cannot use it as client.click(..) input, which expects a selector string not an object - http://webdriver.io/api/action/click.html. To click a WebElement JSON Object client.elementIdClick(ID) instead which takes the ELEMENT property value of the WebElement JSON object.
When a client.elementIdClick is executed, the client will navigate to the page, trying to call client.elementIdClick in the next for loop cycle with next ID will fail, cause there is no such element as you moved away from the page. It will sound something like invalid element cache.....
So, I propose another solution for your task:
Find all elements as you did using elements('a')
Read href and title using client.elementIdAttribute(ID) for each of the elements and store in an object
Go through all of the objects, navigate to each of the href-s using client.url('href'), get the title of the page using .getTitle and compare it with the object.title.
The source I experimented with, not run by Jasmine, but should give an idea:
var client = webdriverio
.remote(options)
.init();
client
.url('https://www.google.com')
.elements('a')
.then(function (elements) {
var promises = [];
for (var i = 0; i < elements.value.length; i++) {
var elementId = elements.value[i].ELEMENT;
promises.push(
client
.elementIdAttribute(elementId, 'href')
.then(function (attributeRes) {
return client
.elementIdAttribute(elementId, 'title')
.then(function (titleRes) {
return {href: attributeRes.value, title: titleRes.value};
});
})
);
}
return Q
.all(promises)
.then(function (results) {
console.log(arguments);
var promises = [];
results.forEach(function (result) {
promises.push(
client
.url(result.href)
.getTitle()
.then(function (title) {
console.log('Title of ', result.href, 'is', title, 'but expected', result.title);
})
);
});
return Q.all(promises);
});
})
.then(function () {
client.end();
});
NOTE:
This fails to solve your problem, when the links trigger navigation with JavaScript event handlers not the href attributes.

Reactive cursor without updating UI for added record

I am trying to make a newsfeed similar to twitter, where new records are not added to the UI (a button appears with new records count), but updates, change reactively the UI.
I have a collection called NewsItems and I a use a basic reactive cursor (NewsItems.find({})) for my feed. UI is a Blaze template with a each loop.
Subscription is done on a route level (iron router).
Any idea how to implement this kind of behavior using meteor reactivity ?
Thanks,
The trick is to have one more attribute on the NewsItem Collection Say show which is a boolean. NewsItem should have default value of show as false
The Each Loop Should display only Feeds with show == true and button should show the count of all the items with show == false
On Button click update all the elements in the Collection with show == false to show = true
this will make sure that all your feeds are shown .
As and when a new feed comes the Button count will also increase reactively .
Hope this Helps
The idea is to update the local collection (yourCollectionArticles._collection): all articles are {show: false} by default except the first data list (in order not to have a white page).
You detect first collection load using :
Meteor.subscribe("articles", {
onReady: function () {
articlesReady = true;
}
});
Then you observe new added data using
newsItems = NewsItems.find({})
newsItems.observeChanges({
addedBefore: (id, article, before)=> {
if (articlesReady) {
article.show = false;
NewsItems._collection.update({_id: id}, article);
}
else {
article.show = true;
NewsItems._collection.update({_id: id}, article);
}
}
});
Here is a working example: https://gist.github.com/mounibec/9bc90953eb9f3e04a2b3.
Finally I managed it using a session variable for the current date /time:
Template.newsFeed.onCreated(function () {
var tpl = this;
tpl.loaded = new ReactiveVar(0);
tpl.limit = new ReactiveVar(30);
Session.set('newsfeedTime', new Date());
tpl.autorun(function () {
var limit = tpl.limit.get();
var time = Session.get('newsfeedTime');
var subscription = tpl.subscribe('lazyload-newsfeed', time, limit);
var subscriptionCount = tpl.subscribe('news-count', time);
if (subscription.ready()) {
tpl.loaded.set(limit);
}
});
tpl.news = function() {
return NewsItems.find({creationTime: {$lt: Session.get('newsfeedTime')}},
{sort: {relevancy: -1 }},
{limit: tpl.loaded.get()});
},
tpl.countRecent = function() {
return Counts.get('recentCount');
},
tpl.displayCount = function() {
return Counts.get('displayCount');
}
});
Template.newsFeed.events({
'click .load-new': function (evt, tpl) {
evt.preventDefault();
var time = new Date();
var limit = tpl.limit.get();
var countNewsToAdd = tpl.countRecent();
limit += countNewsToAdd;
tpl.limit.set(limit);
Session.set('newsfeedTime', new Date());
}
});

How to return number of items in collection?

I'm new to Meteor and I want to create a slideshow with items from a collection, in this case simple words. The slideshow should be controlled by back and forward buttons and replace the current word.
In JavaScript/jQuery I would create an array of objects and a control index, with limits via if-statements, so the index never can drop below zero or overflow the length of the array.
See fiddle for working example:
http://jsfiddle.net/j0pqd26w/8/
$(document).ready(function() {
var wordArray = ["hello", "yes", "no", "maybe"];
var arrayIndex = 0;
$('#word').html(wordArray[arrayIndex]);
$("#previous").click(function(){
if (arrayIndex > 0) {
arrayIndex -= 1;
}
$('#word').html(wordArray[arrayIndex]);
});
$("#next").click(function(){
if (arrayIndex < wordArray.length) {
arrayIndex += 1;
}
$('#word').html(wordArray[arrayIndex]);
});
});
Meteor
I'm curious how to implement this in regards to best practice in meteor and abide to the reactive pattern as I'm still trying to wrap my head around this interesting framework. My first hurdle is to translate the
if (arrayIndex < wordArray.length)
// to
if (Session.get("wordIndex") < ( (((length of collection))) )
According to the docs I should do a find on the collection, but I have only manage to return an empty array later with fetch. Sorry if this got long, but any input would be appreciated to help me figure this out.
collection.find([selector], [options])
cursor.fetch()
This is the code I have so far:
Words = new Mongo.Collection("words");
if (Meteor.isClient) {
// word index starts at 0
Session.setDefault("wordIndex", 0);
Template.body.helpers({
words: function () {
return Words.find({});
},
wordIndex: function () {
return Session.get("wordIndex");
}
});
Template.body.events({
"submit .new-word": function (event) {
// This function is called when the word form is submitted
var text = event.target.text.value;
Words.insert({
text: text,
createdAt: new Date() //current time
});
// Clear form
event.target.text.value = "";
// Prevent default form submit
return false;
},
'click #previous': function () {
// decrement the word index when button is clicked
if (Session.get("wordIndex") > 0) {
Session.set("wordIndex", Session.get("wordIndex") - 1);
}
},
'click #next': function () {
// increment the word index when button is clicked
if (Session.get("wordIndex") < 10 ) {
Session.set("wordIndex", Session.get("wordIndex") + 1);
}
}
});
}
if (Meteor.isServer) {
Meteor.startup(function () {
});
}
.count() will return the number of documents in a collection.
`db.collection.count()`
There is something called Collection helpers, which works similar to other helpers (eg., template, etc.,). More elaborate explanation is covered here: https://medium.com/space-camp/meteor-doesnt-need-an-orm-2ed0edc51bc5

How to DRY up Template Helpers in meteor?

My Template helper has duplicated code:
Template.foodMenu.helpers({
breakfast: function() {
var breakfastItems = EatingTimes.find(// query for breakfast items);
// function to sort breakfastItems in here (code duplication)
},
lunch: function() {
var lunchItems = EatingTimes.find(// query for lunch items);
// function to sort lunchItems in here (code duplication)
},
dinner: function() {
var dinnerItems = EatingTimes.find(// query for dinner items);
// function to sort breakfastItems in here (code duplication)
}
);
I would like to DRY it up:
Template.foodMenu.helpers({
breakfast: function() {
var breakfastItems = EatingTimes.find(// query for breakfast items);
sortFoodItems(breakfastItems);
},
lunch: function() {
var lunchItems = EatingTimes.find(// query for lunch items);
sortFoodItems(lunchItems);
},
dinner: function() {
var dinnerItems = EatingTimes.find(// query for dinner items);
sortFoodItems(dinnerItems);
}
);
Where do I place this function so I can DRY up? How do I name space it so I can call it properly? I am using Iron Router if that makes a difference.
var sortFoodItems = function (foodItems) {
// code to sort out and return foodItems to particular method that calls it
};
Just define you function before helpers in the same file
var sortFoodItems = function (foodItems) {
// code to sort out and return foodItems to particular method that calls it
};
Template.foodMenu.helpers({
breakfast: function() {
var breakfastItems = EatingTimes.find(/* query for breakfast items */);
sortFoodItems(breakfastItems);
},
lunch: function() {
var lunchItems = EatingTimes.find(/* query for lunch items */);
sortFoodItems(lunchItems);
},
dinner: function() {
var dinnerItems = EatingTimes.find(/* query for dinner items */);
sortFoodItems(dinnerItems);
}
});
If you want to use sortFoodItems function in multiple files, create folder with name lib and put the function in file functions.js without var keyword to make it global. For example:
//lib/functions.js
sortFoodItems = function (foodItems) {
// code to sort out and return foodItems to particular method that calls it
};
You need to understand how Meteor reads your project directories. Read more about structuring your application in Meteor docs.
You could also create a global helper for this using Template.registerHelper(name, function) which would look like:
Template.registerHelper('menuItems', function(eatingTime, sortCriteria) {
//do some checking of your arguments
eatingTime = eatingTime || '/* your default eating time */';
sortCriteria = sortCriteria || {/* your default sort criteria */};
check(eatingTime, String);
check(sortCriteria, Object);
//find and sort your items in one mongo query or you could do separate find and sort if you want
menuItems = EatingTimes.find({time: eatingTime}, sortCriteria);
return menuItems;
});
This would be the most meteoric way of DRYing up your code.
and then in your template, you could call it as:
{{#each menuItems 'time args' 'sort arg'}}

Resources