In this post, we are documenting how we used Google’s TensorFlow to build this image recognition engine. We’ve used Inception to process the images and then train an SVM classifier to recognise the object. Our aim is to build a system that helps a user with a zip puller to find a matching puller in the database. This piece will also cover how the Inception network sees the input images and assess how well the extracted features can be classified. Continue reading
A list of command line tips to optimize your terminal experience. All the recommended features are linked to their appropriate sources. Continue reading
Ever wanted a simple, affordable and feature-rich hosting for your custom domain? AWS S3 + Cloudfront combined is a great static site hosting solution, but can be troublesome to setup. You need to setup the S3 correctly, configurable Cloudfront, worry about best practices such as site redirection, sync file and invalidate Cloudfront for update, etc. These are just too troublesome for hosting a static sites, so we open-sourced our tool AWS-site-manager. AWS Site Manager is a simple command-line tool which makes hosting a static website with AWS S3 and CloudFront in best practices really simple.
Last week, while people were thinking about the weekend, 4 Ourskyers got together to join the Software Testing World Cup 2016 (STWC) Asia Competition, one of the leading events for the software testing community. Testers from all over the world will team up and join the competition to train up their skills and take home the crown. Roy, Oliver, Ten and I (Joyz) come from different positions in Oursky formed a 1-day team to join this exciting opportunity. I’d like to share with you the things we learned during the 3 hours we had against the rest of the world to test a mobile app.
The Internet seems to be always available nowadays. Thanks to the mobility of phones and tablets, mobile app developers can design their apps with a variety of features that assume the Internet connection is stable and let users experience tithe unleash the power of the convenient and efficient network services. Social media apps such as Facebook, WhatsApp and Instagram are good examples. However, how would theses apps function when there is no stable internet connection?
“No Network” bugs are hard to dig out.
When an app puts cloud services as its major selling point, it would be often running into various Network Connection Errors. You will see a lot of these issues from the bug report from testers.
The problem only scales up with the project size. And – it makes developers headache – it is hard to gid out all “No Network” related bugs. Often, apps are developed at a company that serves ultra-high performance network.
Developers cannot react quick to fix problem will have to receive “1-star rating App Reviews” on App Store or Play Store hopelessly.
Do not assume the Internet is always reachable.
Start building an app with the Online-First pattern does speed up the progress in early stage development, yet it might become a crisis in a later stage. If your apps rely heavily on the Internet connection, you should be more proactively take network instability problems into account earlier. Hence, the concept “offline-first” was introduced.
Editor’s Note: Chung is an Oursky Intern. After participating in our Smart Contract consulting project, he wrote this introductory article to catalog his learnings about the Ethereum network.
You must have heard of Bitcoin, the distributed currency system.
Bitcoin utilizes a technique called blockchain which stores all the transaction records. Blockchain also ensures that no malicious modifications could be applied to the records.
As a distributed system, Bitcoin utilizes machines on the whole network to compute and verify the changes on the records.
What if we go went beyond storing transaction records into program states? This makes the whole network a computer for general purpose computations with program states that no one can modify illegally. Ethereum does this.
Let me explain below.
What is missing from Bitcoin?
Bitcoin blockchain architecture has several problems for general purpose computations. They include network state dependency, Turing-completeness and block time.
Bitcoin only stores the transactions of coins in the blockchain. Therefore, the state of the network (implicitly constructed by transaction records) is just the amount of coins for each account. For general purpose computations, we need general purpose program states.
The egg dropping puzzle was one of the key interview questions in Google
Egg Dropping Puzzle (2-egg, 100-floor version)
“Figure out the highest floor of a 100-floor building an egg can be dropped without breaking, given two eggs”
The Egg Dropping Puzzle is a mathematical puzzle that has been around the internet for some time now, which is known to be adopted in interviews of major companies like Google, Microsoft, Accenture and even Hewlett Packard.
You are to determine the minimum number of attempts required in the worst case scenario to find the critical floor.
For QA in an agile development team, SPEED is everything. It is a game on a high-speed roundabout, developers introduce new features and fixes every day, and you better test everything before the next iteration. Besides the new features or fixes, you still need to handle a full UI regression smoke test, how is that possible? Hire more staff? Test automation is a way out for smart teams, and that’s what we do at Oursky.
The basic aim of Test Automation is to replace repetitive human work with computers.
So, to get started, all you need to do is no more than wrapping up your existing test plans, and to decide which parts are to be automated. If you already have a comprehensive test plan, with step-by-step test cases and well-defined expected results, congratulations! You are perfectly ready for automation.
Or else, some preparation are needed.
Python is known as a dynamic, strong-typed language. Most developers love it but some feel mad without type checking or type-hinted auto-completion. In Python3.5, Type Hints is introduced to further delight developers who want those features.
Type Hints offers type checking on function parameters, return values and class attributes, as if it’s static-typed. If you pass something does not match the expected type, a warning will be given.
According to The Theory of Type Hints, here’s an example showing how the rules work out in practice:
Say there is an
Employee class, and a subclass
class Employee: ...
class Manager(Employee): ...
Let’s say variable e is declared with type
e = Employee() # type: Employee
Now it’s OK to assign a
Manager instance to e:
It’s not OK to assign an
Employee instance to a variable declared with type
m = Manager() # type: Manager
m = Employee() # Fails static check
Now, suppose we have a variable whose type is
a = some_func() # type: Any
It’s OK to assign
Of course it’s also OK to assign
Employee e to
We’ve been dealing with many many different apps. App quality and usability are always our core values we never settle for less.
There are some common UX mistakes that engineers often overlooked. Getting these discovered and handled, one can build greater apps.
Get ready, here are our top 10:
10. Long text is too long in a text field
You’ve probably met this before. Long text can be really loooooooooooooooooooooooooong.
It’s not the text’s fault.
It’s the text field couldn’t handle that.
Yet, there are ways to deal with that:
- Truncate the text.
- Overflow the text field, make it scroll.
- Expand the text field.
- Limit the text length!