Final Review

2022/12/13 14:21 PM posted in  CS121   comments
Tags:  #CS-121

5 Testing

Levels of Testing

  1. Unit testing: One component at a time
    • A component could be a method, class, or package
    • If test fails, defect localized to small region
    • Done early in software lifecycle, ideally when/before component is developed, and whenever it changes
  2. Integration/system testing: The whole system together
    • Ensures components work together correctly
    • Possible even if system not complete, as long as there’s some end-to-end slice of its functionality
  3. Other testing terms
    • “Acceptance test” — test system against user requirements
    • “Regression test” — make sure new version of software behaves identically to old version

JUnit

  • A test case in JUnit is just a class
    • Test methods are annotated with @Test
  • Ideally, each test case should check one thing
  • JUnit tests are completely automated
  • Drawback: adds cost
  • Major benefits
    • Tests can be run often
    • Code maintainance and evolution becomes much safer

Regression Testing

Key idea: when you find a bug

  • Always tunthe test when code changes
  • Write a test that exhibits the bug

Developing Test Cases

Black box testing

  • Pros
    • This is what we actually what the program to do
    • Process not influenced by tested component
    • Robust with respect to implementation changes
    • Allows testers to be independent
  • Cons
    • Specs are notoriously complete; specs do not necessarily tell you every place the code could go wrong.

White box testing

  • Pros
    • This is the code we are actually running!
    • Performance optimizations
  • Cons
    • If our code is completely missing come ket property that is in the spec, we might not even know to test it
    • Tests might be overly specific to this particular implementation
  • Actually used glass bos testing earlier.

Coverage Criteria

High coverage means few mistakes remain in program.

Refactoring

Motivation

  • Old-style software design process
    Waterfall Model

  • Good properties

    • Provides structure to the software engineering process
    • Lots of emphasis of careful thought and design early on
  • Bad Properties

    • Requirements often not known in advance
    • Designs often need to be changed
  • Result

    • Strict adherence to waterfall leads to inappropriate designes

Refactoring

New approach to software design

  • Come up with a resonable first design
  • But then be willing to change and evolve design over time
  • Refactoring enables safe design changes
    • Assumption: we have a comprehensive, automatically runnable test suite
    • Then we can divide code changes into two sorts
      • Bug fixes or feature addtions that modify functionality
      • Refactoring
        • Change code design, but do not change behavior
        • implies can rerun all exists tests to ensure change works

Replace Number with Constant

  • New code is more readable
  • Can avoid typos of we reuse magic number several times
  • Might want ro add more digits

Other Refactorings

  • Move method from one class to another

  • Move method from a subclass to a superclass or the reverse

  • Extract code sequence into its own method

  • Replace conditional branching with dunamic dispatch

  • Group together a long parameter list into an object containing those values

  • Ideal time to refactor: When you want to make a change and the current design impedes the change

  • Be cautious of refactoring buggy code

  • Be cautious if you keep refactoring the same code over and over

  • Refactorings tend to be small changes; so they can’t fix every problem

6 Architecture

MVC Arch

Pros and Cons

Pros
  • Seperation between data and interface is key
    • Views can be replaced, changed, customized, expanded
    • As db is read and written, changes reflected in all views
    • Scalable deployment easier, e.g., multiple controller/view instances communicate with one db'
Cons
  • Several potential drawbacks
    • Many kinds of changes to model require changing view and controller
    • Views and controllers are closely coupled.
    • Added complexity

Pipe and Filter

Components are filters, connections are the pipes.

Advantages

  • Overall behavior is a composition of filter behaviors
  • Potentially good reuse by cerating different compositions of filters
  • Can test each filter in isolation
  • Filters can be replaced individually

Disadvantages

  • Not good for interactive use
  • Pipes are narrow. Hard to pass complex data
  • Overhead for parsing/unparsing data when read from/sent to a pipe

Layered Arch

  • Organized as a hierarchy

Advantages

  • Good fit for system with increasing levels of abstraction
  • Changing one layer affects at most two others
  • Can interchange implementations of same layer interface.

Disadvantages

  • May not be able to identify clean layers
  • Might need to jump layers for performance or functionality

Client-Server Arch

  • Server is central point of failure
  • Any client-to-client communication must go through server

Peer-to-Peer Arch

  • Trust between nodes
  • More equal upload/download volume compared to client server
  • Location of data on network not centralized

7 Concurrency

Advantages

Performance!

Disadvantages

  • Harder to think about
  • Concurrency adds overhead

Basic Threads in Java

  1. Subclass Thread and implement run()
  2. Anonymous inner class
  3. Implement Runnable

Ordering

  • within threads, ordering applier
  • across threads, only ordered when
    • One threads starts another
    • One threads waits for another to finish

Reentrant Lock

ReentrantLock allows threads to enter into the lock on a resource more than once

Dead Lock

Synchronized

synchronized(obj) {body}
Obtain lock associated with obj

While loop + signal all = carefully use of signal

Message Passing

  • Pros
    • More natural for many supercomputer architectures and distributed systems
    • No possibilitty of data races
  • Cons
    • Atomicity still problematic
    • Inefficient to exchange large amount of data

8 Debugging

Debuggers

Time Travel Debuggers

Debugging as Experimentation

Simplifying Tests

Delta Debugging

If part of a test is a really big input, simplify it using a binary search-like process called delta debugging

  • Cut input in half, testing each of the halves
  • If one half has the same bug, throw away the other half and repeat
  • If neither half has the bug, increase granularity by removing quarters instead of halves, and repeat

Split the input

Increase Granularity

9 Security

Goal: prevent bad things from happening

Integrity

  • No improper modification of data
    • E.g., account balance updated only by authorized transactions, only you can change your password.
    • Integrity of security mechanism is crucial
    • Enforcement: access control, digital signitures

Confidentiality

  • Protect information from improper release
    • Limit knowledge of data or actions(secrecy)
    • Enforement: access control, encryption
    • Hard to enfore after the fact

Availability

  • System muct respond to requests
    • I.e., do not ensure confidentiality and integrity by unplugging your computer!

10 History