Alex Mishyn's blog @amishyn-blog - Tumblr Blog

Product defined in Excel

First time it happened to me when I was building one pet product for fun. The idea was simple, get data from device, crunch some numbers and Voila! product is ready. But, apparently it is not real world case, data appears to be bad quality and need to be cleaned before processing.

So I decided to use algorithms and started looking into space of target language, iOS swift, — no implementations found, then looked at Ruby situation is much better but still decided to think. And then I realised that R has full set of algorithms to play with. So I went to R and failed. Failed because I didn’t know my data well enough to understand what should I do with it and what algorithms to apply.

So I ended with excel spreadsheet and did all the job manually: got rid of garbage, unified and pre-processed data. In excel. I ended with description of algorithm how to clean data. And this is all I need.

Next step, was implement this algorithm. I did it in Ruby, because it is faster to develop and debug. Tweaked it till I get expected results on sample set. And only after prototype succeed I sat down and coded algorithm in swift and finally launched application on iOS device. This story is open as I haven’t launched app yet. But lesson learned, know your data, and do prototype in the fastest way.

The second case was when I decided analyze my personal finances. Shame on me, I don’t accurately track them in one of those apps. So I manually exported transactions, categorised and started playing with structured data. Comparing categories, sorting, making charts. And all this in order to find an answer got to optimise my family spendings.

Only here I realised how much faster I can iterate in excel rather than implementing features only to find out I need more iteration to understand feature is not useful. Obviously iterate in excel is faster than iterate by code. So this is the right way, build prototype manually and then automate if it succeeded.

Some time ago I was a little bit involved in similar web product which put on hold current;y. So there was a struggle with what to do with collected data and how to help customer with processing. What features user needs and so on. How to visualize it properly. How to build habit of user to optimise spednings etc.

Now I can say that most efficient mvp should be done manually first, got proven and only then automated. And excel is a good tool.

PS: I’ve used apple numbers, but since excel is an industry standard in spreadsheets I’ve used it’s name

#mvp prototype

Comparing speed of RABL and Active Model Serializers

I’ve been using RABL to run JSON api for quite long time. There are good and bad things.

But for me key moment always is product. So here from product perspective most interesting is performance and time spent on development. Here I would like to talk about performance.

I’ve expected to just google for comparison of rabl vs active model serialisers, read an article and decide what to do. I did so, but unfortunately nothing found. So I decided to do my own research.

I picked my small project, which is easy to migrate from rabl to active_model_serialzers and run performance tests against it. So I made selective rendering based on input parameter and deployed code to heroku. And ran jmeter to measure endpoint results.

I’ve picked heavy request with a lot of json generated and put equivalent load to rabl and active model serialisers.

Environment is next: 1 heroku dyno with Rails 4.0 deployed on ruby 2.1. and puma web server.

Here are the results:

Small load

1 threads, 3 sec ramp-up and 10 times to repeat.

Average for rabl is: 406ms

Average for am is : 369ms

More load

3 threads, 3 sec ramp-up and 50 times to repeat.

Average for rabl is: 1000ms

Average for am is : 600ms

Bigger load

5 threads, 5 sec ramp-up and 50 times to repeat.

Average for rabl is: 1260ms

Average for am is : 980ms

So under small load results are pretty similar. But under heavier load rabl slower than active model serializers by 40%. I’m moving my project to active model serializers.

Links:

https://github.com/nesquena/rabl

https://github.com/rails-api/active_model_serializers

Today faced a problem where Devise authenticate_... method doesn't work. I have custom controller for Registrations, and inside it I to have protected actions.

class CandidatesController < Devise::RegistrationsController before_filter :authenticate_candidate!, :only => [:show] def show end

But protection is not working for show action. To find out the problem, I've user debugger First, I redefined the method.

def authenticate_candidate! debugger super end

And then step in, and see the problem

45 class_eval <<-METHODS, __FILE__, __LINE__ + 1 46 def authenticate_#{mapping}!(opts={}) 47 opts[:scope] = :#{mapping} => 48 warden.authenticate!(opts) if !devise_controller? || opts.delete(:force) 49 end

So adding custom methods into custom devise controller may be tricky.

iOS 7 UI changes brief summary

iOS 7 was publicly presented at WWDC 2013 keynote. iOS sessions covers all new UI and API. There are a lot videos to see. Each video is about 50 minutes length, some of them overlap each other a little bit. Here I want to give you a brief overview of new concepts and provide links to original WWDC sessions.

First impression of iOS 7 was based on screenshots with that weird icons and great parallax effect. Same stuff but flattened and coloured in acid would someone say. But it is not true. This is just visible part of changed concept. Furthermore, phrase "Parallax, blurs and translucency take the place of gradients, shadows and cutlines" is also a description of noticeable changes. The main concept is changed.

Focus on content

Is the new concept of iOS. And three big ideas behind it.

Clarity

"Show whats relevant. Present data to user in a way that makes sense. Prove features which are self explanatory and discoverable"

Deference

"get the chrome out of the way"

Depth

"Parallax, blurs and translucency take the place of gradients, shadows and cutlines"

Examples and tools

Full screen content

Buttons without borders

Typography

Text styles

Text Dynamic type

All this info you can find at 101 "Platforms State of the Union" as well as all other news from Apple. And more details in the following sessions:

210 - Building interface for iOS7

206 - UIKit Dynamics

208 - What's ne in iOS interface desing

218 - Custom transitions using view controllers

226 - Implementing engaging ui

225 - Best practices for great iOS design

Focus on content is great but users need something more. So the sauce is animations.

animations

218 "Custom Transitions Using View Controllers"

pinch and transition between collections ( #218 40:10) Animated transition between two different views

collection view transition (#218 16:35) This is a technical example for previous example.

example of tab bar transition (#218 31:27) From now, you can add such transitions even to standard controllers. In this case it is tab bar.

206 "Getting Started with UIKit Dynamics"

Motion effects (#218 17:06) Tries to be pretend as real world animation

dragging photos (#206 5:00 Drag corner and object will rotate it, like in real life

bouncing of the view (#206 8:30) Its a gravity effect with colliding

A video for those animations:

All videos as well as PDF's you can find at: https://developer.apple.com/wwdc/videos/

Preload associations in rails

Preloading associations is well known, include used to do it

Post.include(:user, :comments).limit(10)

What if you need to run it on already fetched collection? ActiveRecord::Associations::Preloader helps.

ActiveRecord::Associations::Preloader.new( collectionsofobjects, associations )

Where collectionsofobjects is array of ActiveRecord objects and associations is specifies one or more associations that you want to preload. Same syntax as used in include (Here more info) [http://apidock.com/rails/ActiveRecord/Associations/Preloader/new/class]

Usage:

posts = Post.limit(10) #… ActiveRecord::Associations::Preloader.new( posts, [ :user, :comments, { :location => [:city, :country]} ] ).run

#rails #preload associations

Cache when you need not when you can

Put rails cache is a piece of cake.

Which one you want ? Page, fragment ? Templates ? Http ? Rails cache guide

But all this options supposed to make app faster. And to really do it you should identify slow pieces first. Guessing is not for professionals. If you are pro, you should have monitoring tools such as new relic where you can identify real problems. A nd then you should leverage all that caching options. And again measure it to ensure positive result.

Do not put cache “not be worse”, it’s not true. It creates illusion of defense, but in reality it just tricks you.

Examples:

Collection cached, but associations used in templates doesn’t. As a result you have lots of queries.

Cached resources which should not be cached at all. User owned stuff for instance. Its frustrating when you have no control over you resources.

When caches such big that they not fit into memcache.

Optimize code without caching first. Then cache.

#caching #optimization #rails

Production usage of ActiveRecord::IdentityMap was short

I was thinking to try identity map in production. I had some negative feedback and actually comments on the net was not so positive. A lot of notes that its not production ready, because of association issues and so on. Here you can find more

But I want to speed up just read action While reading source code, I found that it has method :use which accepts block as parameter. So I’ve created around filter to use IdentityMap for particular action

around_filter :with_identity_map, :only => :index def with_identity_map ActiveRecord::IdentityMap.use do yield end end

Thats all. Results was positive. It decreased page load by two times from 1600 ms to 800ms. Which is still unacceptable. It looks like quick hack to fix not “perfect code”

Few weeks everything was ok. But after some changes IdentityMap starts conflicting with caching.

Errors was like “has_many association returns nil”. I disabled identity map. And then we refactored code as a result of refactorign page load downs to ~200 ms.

That was journey with identity_map.

#active_record #rails #identity_map #production

Custom smtp settings for Rails Mailers

I have to define different smtp settings per mailer. Lets say, I want to use different SendGrid accounts to send emails. Rails Mailers use global ActionMailer::Base.smtp_settings. If you want custimize smtp settings per mailer you cant just do MyMailer.smtp_settings = {...}And got work done. Actually you can, but it will set this settings globally. The easiest way, is just to redefine smtp settings for particular mailer

class MyMailer < ActionMailer::Base def self.smtp_settings { ... } end end

And if I want to set different smtp settings for few mailers and make them easy configurable.

I can use next approach: Load configuration for yaml file for each mailer.

Code for it:

Create parent mailer, lets say ApplicationMailer

class ApplicationMailer def self.smtp_settings Hash[SmtpSettings.send(self.to_s.underscore)].symbolize_keys rescue Settingslogic::MissingSetting ActionMailer::Base.smtp_settings end end

class MyMailer < ApplicationMailer end

And create settings:

class SmtpSettings < Settingslogic source Rails.root.join("config/smtp_settings.yml") namespace Rails.env end

smtp_settings.yml:

common: &default_settings my_mailer: address: smtp.sendgrid.net port: 25 authentication: :plain development: <<: *default_settings staging: <<: *default_settings test: <<: *default_settings production: <<: *default_settings

Thats all. Now its easy to configure smtp settings per mailer and per environment. Set mailtrap.io account for dev and stagign and use SendGrid for prod.

#rails3 #action_mailer #smtp_settings

Rails 3 resource renaming

Rails 2 way:

There was :as option in routes. Which rename routes.

Assume that you have next in routes.rb

resources :blogs # /blogs

If you want to rename it from urls perspective you can:

resources :blogs, :as => 'posts' # /posts

But, in rails 3 this doesnt work.

As workaround I can propose next:

resources :posts, :as => 'blogs', :controller => 'Blogs'

Rails 3 and mongodb through mongoid

I start playing with rails 3 and mongo. Found lot of articles on this theme, but only some steps works for me. Most useful is post by Nicola Racco.

I've created simple web app with authentication and attachments photo/video.

As js framework I use jQuery 1.4.2. For integration use smth like:

http://github.com/CodeOfficer/jquery-helpers-for-rails3

I use http://gist.github.com/399682, that found somewhere on github.

For authentication: devise

for attachements: carrierwave

ORM: mongoid

Paginations: Mongoid paginate + will_paginate for views

Here is my steps:

Install rails 3 beta 3

gem install rails --pre

Creating app

rails myproject --skip-activerecord

ensure that in config/application.rb

require 'action_controller/railtie' require 'action_mailer/railtie' require 'active_resource/railtie' require 'rails/test_unit/railtie'

instead of:

require 'rails/all'

As ORM I choose Mongoid. Beacuse in devise 1.1.rc1 orm adapter only for mongoid and there are no adapter for mongo_mapper

gem "mongoid", "2.0.0.beta4"

Now lets start implement app logic

Authentication:

gem 'devise', :git => 'git://github.com/plataformatec/devise.git'

I've installed device 1.1.rc1,

bundle install

Next install devise

rails generate devise_install

Rails will generate the file "config/initializers/devise.rb", responsable of the Devise initialization. Opening the file you'll see several options that you can configure as you like to customize the framework behavior. In this article the task is to make Devise to work with Mongoid, so we don't care of these options. Let's change the line:

require 'devise/orm/active_record'

in:

require 'devise/orm/mongoid'

Next steps following devise docs:

class User include Mongoid::Document include Mongoid::Timestamps devise :database_authenticatable, :registerable, :recoverable, :rememberable, :trackable#, :validatable has_many_related :posts end

Pagination:

Post.all.paginate(:page => params[:page], :per_page => 10)

for views simply use will_paginate

Attachements:

http://github.com/jnicklas/carrierwave

Uploader placed in app/uploaders

lass PhotoUploader < CarrierWave::Uploader::Base include CarrierWave::Compatibility::Paperclip include CarrierWave::RMagick storage :file def store_dir ":rails_root/public/photos" end def paperclip_path ":rails_root/public/photos/:style/:basename.:extension" end version :thumb do process :resize_to_fit => [800,600] end end

Model

require 'carrierwave/orm/mongoid' class Photo include Mongoid::Document include Mongoid::Timestamps field :title, :type => String field :permalink, :type => String field :content, :type => String # field :hit, :type => Integer, :default => 0 field :photo, :type => String field :photo_file_name, :type => String belongs_to_related :map mount_uploader :photo, PhotoUploader, :mount_on => :photo_file_name end

Controller

class PhotosController < ApplicationController def create photo = Photo.create(params[:photo]) redirect_to(:back) end

And views:

<% form_tag videos_path, {:multipart => true} do %> <p> <label>Photo</label> <%= file_field_tag 'video[video]' %> </p> <%=submit_tag 'Upload'%> <% end %>

#rails3 #mongoid #authentication #attachements #devise #carrierwave

AR save(false) used less memory than save

When having lot of updates ( ~ 7983)

Memory used by daemon permanently growth upto 800Mb

THe problem is next

When saving record without validation save(false) max mem usage was 161 Mb

But there are no validations for this object

When I go through source code. Found that this issue because of valid? method.

Going more deeply I found that issue with new_record?

def new_record?

@new_record

end

Strange )

Tested on MRI 1.8.7, 64-bit

AR 1.15.6

Ruby hash.with_indifferent_access

Different behavior in 1.8.7 and 1.8.6 with Rails 1.2.6

>> {:sent => 1, "sent" => 0}.with_indifferent_access

=> {"sent"=>0}

>> RUBY_VERSION

=> "1.8.6"

>> RAILS_GEM_VERSION

=> "1.2.6"

>> {:sent => 1, "sent" => 0}.with_indifferent_access

=> {"sent"=>1}

>> RUBY_VERSION

=> "1.8.7"

>> RAILS_GEM_VERSION

=> "1.2.6"

trulala

amishyn-blog

markoblogo:

Где деньги Юль? (via trulala)

ld100

amishyn-blog

ld100:

Весь мак-софт должен полставляться в таком виде!

Nice Infrared palms wallpaper on my desktop

#desktop #wallpaper

ld100

IE6 вышел в 2001 году, с тех пор успели разбомбить ВТЦ, захватить Афганистан, Ирак, выбрать по два раза Жору Куста с дядей Вовой, и по одному разу Обаму с Медведевым. За этот же срок прокатились всякие оранжевые революции, придумали Ладу Калину и Приору, телефоны проделали путь от 3310 до IPhone’ов и Nexus One, опять таки за это же время появились и достигли сегодняшних высот facebook, twitter, vkontakte и ещё уйма всего. А интернет технологии что, на месте что ли по-вашему стоят? Хороша ложка к обеду, IE6 своё давно отжил.

С ХабраХабра про Internet Explorer 6 (via ld100)

Screencast in background on iPhone

I've got couple of screencasts from some conferences. At some moment I want to listen it going to work. So I need play it in background but regular control will stop screencast after locking iPhone or exiting iPod.

But next trick is working. Run screencast then exit the iPod app and it will be stopped. Next you should just push play button on garniture (on headphones) and screencast plays in background!

#iphone #screencast #background

Trending Blogs

Recently Viewed Blogs

Alex Mishyn's blog