COMP2110 Software Design in 2004 lecture 10 Software Architecture 2 of 2 design lecture 5 of 6

1
COMP2110 Software Design in 2004 lecture 10
Software Architecture 2 of 2 design
lecture 5 of 6

• Goal of this small course modulea descriptive
  understanding(not detailed application of a big
  library of examples).
• Some existing software architectures
• batch and pipeline
• client-server and broker see the notes
• layers see the notes

2
Software Architecture in general

• The decomposition of the design into groups of
  abstract modules (clusters of classes)
• modules are concepts, with names
• account management, playing field, characters
• potential for generality and reuse in other
  projects
• playing field, alarm control, selection list
• coherence and coupling
• stronger/more relationships between classes
  inside the module boundarythan going in and out
  through the boundary

3
Elements of software architecture

• description of elements from which software
  systems are built
• interactions between those elements
• patterns that guide their composition
• constraints on those patterns
• In general, a system is defined in terms of
• a collection of components
• interactions between those components

4
Existing software architectures (just a few)

• dataflow systems
• batch sequential
• pipes and filters
• call-and-return systems
• main program and subroutines
• hierarchical layers
• independent components
• communicating systems
• Model View Controller

example 1lex,parse,generate
notes 2 layers
notes 3 broker
example 2 MVC
5
Example 1Lexical analysis, parsing,
identification (1)

• a widespread class of problems require processing
  relatively complex sequences of character input
  by
• lexical analysis
• parsing
• identification of words (identifiers)
• examples
• programming languages for pretty print,
  syntax-sensitive editing, compilation,
  interpreting
• HTML, XML for many kinds of further processing
  rendering, editing

6
scan parse identify batch architecture (2)

• Batch processing uses the schema
• divide the processing into a number of processing
  stages that can be done in sequence.
• process all of the input in a batch as a whole
  through process 1, to produce a batch of
  intermediate data stage 1
• process all of the intermediate data stage 1
  through process 2... to produce a batch of
  intermediate data stage 2
• in general
• process data stage i through process stage i1
  to produce data stage i1

7
batch architecture
input
data stage 3
data stage 1
data stage 2
process 2
process 3
process 1
abstract syntax tree
AST with tagged meanings
character sequence
token sequence
lexical scan
parse
identify
8
Stage 1 lexical scan
output sequence of tokens
inputHTML source
STARTTAG "p ATTR "color ATTRVAL "black ATTR
"font ATTRVAL "Times DATA "This is it
STARTTAG "pre DATA " no more code! COMMENT "
but then... DATA " no more ENDTAG "pre DATA
"that was it ENDTAG "p"
ltp color"black font"TimesThis is it ltpregt no
more code!lt!-- but then... --gt no morelt/pregt
that was it lt/pgt
lexical analysis
9
Stage 2 parse
input sequence of tokens
phrase structure(syntax tree)
STARTTAG "p ATTR "color ATTRVAL "black ATTR
"font ATTRVAL "Times DATA "This is it
STARTTAG "pre DATA " no more code! COMMENT "
but then... DATA " no more ENDTAG "pre DATA
"that was it ENDTAG "p"
ATTRVAL "black ATTR "color
ATTRVAL "Times ATTR "font ATTRIBSEQ 2
DATA "This is it DATA " no more code!
COMMENT " but then... DATA " no more
ELEMENT "pre" 3 DATA "that was it ELEMENT "p"
4
parse
(this is a syntax tree, in postfix
notationindentation added for human reader)
10
Pipeline architecture

• in a pipeline, the processes produce the same
  input and output as the batch processes but as
  sequences, not as complete batchesThe processes
  operate incrementally later processes can start
  before the earlier ones have finished.
• the batch architecture is a limiting case of the
  pipeline

11
pipeline architecture
input
data stage 3
data stage 1
data stage 2
process 2
process 3
process 1
abstract syntax tree
AST with tagged meanings
character sequence
token sequence
lexical scan
parse
identify
12
pipeline ctd

• the processes are not always in lock step
• the scanner will need to consume a variable
  number of characters for each token
• the parser may need to consume several tokens
  before it can produce any parse tree
• the architectures can be mixed e.g in WebView
• lexical scanner-gt parser
  Pipeline
• -gt renderer
  Batch

13
batch vs pipeline

• performance space (batch big) and time
• responsiveness pipeline has fast first output
• may need whole structure to apply variety of last
  stage processes suggests batch at the last
• may need overview across whole of input
• pipeline may be too complex to program

14
batch vs pipeline

• performance space (batch big) and time
• responsiveness pipeline has fast first output
• may need whole structure to apply variety of last
  stage processes suggests batch at the last
• may need overview across whole of input
• pipeline may be too complex to program

15
example 2Model-View-Controller Architecture

• Recall the example of a spreadsheet program
• Model - the core application part
• underlying data, functional core of application
• registers dependent views and controllers
• notifies views and controllers when data changes
• View - the graphical part of the interface
• creates its controllers
• displays information (about the model) to the
  user
• updates itself when notified that the model
  changes
• retrieves data (gets state) from the model
• Controller - the decision-making part
• accepts user inputs and other events
• translates events into service requests for the
  modelor display requests for the view
• (from comp2100 notes and JMT book)

16
Using an existing architecture in a high-level
design

• How to document the use of an architecture in a
  high level design
• name the architectural model you use
• e.g. Model View Controller
• name which modules play which roles
• e.g. the Model is module CentralControlStates
• the View is module UserInterfaceManager
• the Controller is the ControlLogicManager module
• state any major differences from the architecture