arturhoo / como-nos-sentimos Goto Github PK

As they are not put in the Beanstalk tube for geolocation, geotagged tweets don't have weather data attached during the backend's pipeline.

A better solution would be to pass the geotagged tweets to the geocoder, which would then differentiate it from the rest, but still find weather information.

Currently the Map class is used to erase Hasmap structures in the Processing code:

void clearFeelingsOccurrence() {
  Iterator j = feelingsOccurrence.entrySet().iterator();
  while (j.hasNext ()) {
    Map.Entry me = (Map.Entry)j.next();
    feelingsOccurrence.put(me.getKey(), 0);
  }
}

I expect it to be a problem when converting existing code to processingjs

Get template from html5bootstrap

Important sections:

• Filters
  • Only with location
  • Feeling
  • User
  • State
• More stats
• Faq
  • Cities are incorrectly displayed on the map.
  • How does it work?
  • How are the colors chosen?
  • Incorrectly identified feelings, eg.: Parece que hoje fico sem celular, bateria fraca.
• About
• Contact
• Facebook like, Twitter tweet, Github clone/watch (ribbon?)

If, for some reason, no tweet is fetched in a one hour or more timespace, the database will not have a entry for that hour with count=0

There are multiple local_settings.py files around that could be merged.

The files feelings.txt and query_terms.txt should also be put in the root directory as they are relevant to all three components of the system.

Example:

fico nervosa>>>começo a gritar muito

Should be displayed as:

fico nervosa>>>começo a gritar muito

The best solution would be to reload only the canvas.

The use of the template engine to pass the content of the visualization might be a problem.

The checkbox is misplaced.

Ideally, if there is nothing to insert, the geocoder would simply not create the key location in the Mongo document. Example of the issue:

{ 
"_id" : NumberLong("212260966136684544"), 
"author" : { "location" : "Margem Sul", 
             "screen_name" : "Carla1906", 
             "name" : "Carla Caetano" }, 
"text" : "Boa tarde a todos! Estou  a morrer de fome por isso quero jantar.", 
"created_at" : ISODate("2012-06-11T19:12:12Z"), 
"feelings" : [ "fome" ], 
"created_at_local" : ISODate("2012-06-11T16:12:12Z"), 
"location" : null, 
"feelings_size" : 1
}

More or less similar to #7, as generating a separate JS would make it easier to reload the sketch without reloading the webpage.

The JS would need to be generated by the Flask application, still based on the template engine.

Right now, in branch v2, when a tweet has multiple feelings, they are all identified and stored, posing several issues:

• How to represent a tweet with more than one feeling?
• Right now we only represent the first feeling of the array in the DB. When searching for the tweets with specific feelings, we have to specify the size of the array so that we don't get those with multiple feelings, eg:

db.dev03.find({$or: [{'feelings': 'feliz'}, {'feelings': 'triste'}], feelings_size: 1})

The problem, is, although indexes are used, they are not used in the best way possible. Maybe create a new index?

https://github.com/ivaynberg/select2

This list is used, amongst other places, in the feelings select field in the website, and accents positioned after all the "clean" letters. Eg.: ótimo comes after vivo

Consider using S3 or NGINX to serve the file

Quoting processingjs.org:

Font support for the Canvas element varies across browsers. FireFox currently has the best support, but still uses a pre-defined set of fonts. (Documentation to follow soon.) Good alternatives to the browser's Canvas-font functions are available: such as the Glyph method. See also: Canvas Text, BurntToys Bauhaus

We make good use of PFont in our processing work, let's see how we can address that.